1493 lines
56 KiB
Rust
Raw Normal View History

First refactor for WebIDL bindings This commit starts the `wasm-bindgen` CLI tool down the road to being a true polyfill for WebIDL bindings. This refactor is probably the first of a few, but is hopefully the largest and most sprawling and everything will be a bit more targeted from here on out. The goal of this refactoring is to separate out the massive `crates/cli-support/src/js/mod.rs` into a number of separate pieces of functionality. It currently takes care of basically everything including: * Binding intrinsics * Handling anyref transformations * Generating all JS for imports/exports * All the logic for how to import and how to name imports * Execution and management of wasm-bindgen closures Many of these are separable concerns and most overlap with WebIDL bindings. The internal refactoring here is intended to make it more clear who's responsible for what as well as making some existing operations much more straightforward. At a high-level, the following changes are done: 1. A `src/webidl.rs` module is introduced. The purpose of this module is to take all of the raw wasm-bindgen custom sections from the module and transform them into a WebIDL bindings section. This module has a placeholder `WebidlCustomSection` which is nowhere near the actual custom section but if you squint is in theory very similar. It's hoped that this will eventually become the true WebIDL custom section, currently being developed in an external crate. Currently, however, the WebIDL bindings custom section only covers a subset of the functionality we export to wasm-bindgen users. To avoid leaving them high and dry this module also contains an auxiliary custom section named `WasmBindgenAux`. This custom section isn't intended to have a binary format, but is intended to represent a theoretical custom section necessary to couple with WebIDL bindings to achieve all our desired functionality in `wasm-bindgen`. It'll never be standardized, but it'll also never be serialized :) 2. The `src/webidl.rs` module now takes over quite a bit of functionality from `src/js/mod.rs`. Namely it handles synthesis of an `export_map` and an `import_map` mapping export/import IDs to exactly what's expected to be hooked up there. This does not include type information (as that's in the bindings section) but rather includes things like "this is the method of class A" or "this import is from module `foo`" and things like that. These could arguably be subsumed by future JS features as well, but that's for another time! 3. All handling of wasm-bindgen "descriptor functions" now happens in a dedicated `src/descriptors.rs` module. The output of this module is its own custom section (intended to be immediately consumed by the WebIDL module) which is in theory what we want to ourselves emit one day but rustc isn't capable of doing so right now. 4. Invocations and generations of imports are completely overhauled. Using the `import_map` generated in the WebIDL step all imports are now handled much more precisely in one location rather than haphazardly throughout the module. This means we have precise information about each import of the module and we only modify exactly what we're looking at. This also vastly simplifies intrinsic generation since it's all simply a codegen part of the `rust2js.rs` module now. 5. Handling of direct imports which don't have a JS shim generated is slightly different from before and is intended to be future-compatible with WebIDL bindings in its full glory, but we'll need to update it to handle cases for constructors and method calls eventually as well. 6. Intrinsic definitions now live in their own file (`src/intrinsic.rs`) and have a separated definition for their symbol name and signature. The actual implementation of each intrinsic lives in `rust2js.rs` There's a number of TODO items to finish before this merges. This includes reimplementing the anyref pass and actually implementing import maps for other targets. Those will come soon in follow-up commits, but the entire `tests/wasm/main.rs` suite is currently passing and this seems like a good checkpoint.
2019-05-23 09:15:26 -07:00
//! The polyfill for the WebIDL bindings proposal in wasm-bindgen.
//!
//! This module contains the polyfill (or at least the current state of and as
//! closely as we can match) the WebIDL bindings proposal. The module exports
//! one main function, `process`, which takes a `walrus::Module`. This module is
//! expected to have two items:
//!
//! * First it contains all of the raw wasm-bindgen modules emitted by the Rust
//! compiler. These raw custom sections are extracted, removed, decoded, and
//! handled here. They contain information such as what's exported where,
//! what's imported, comments, etc.
//! * Second, the `descriptors.rs` pass must have run previously to execute all
//! the descriptor functions in the wasm module. Through the synthesized
//! custom section there we learn the type information of all
//! functions/imports/exports in the module.
//!
//! The output of this function is then a new `walrus::Module` with the previous
//! custom sections all removed and two new ones inserted. One is the webidl
//! bindings custom section (or at least a close approximate) and the second is
//! an auxiliary section for wasm-bindgen itself. The goal is for this auxiliary
//! section to eventually be empty or inconsequential, allowing us to emit
//! something that doesn't even need a JS shim one day. For now we're still
//! pretty far away from that, so we'll settle for using webidl bindings as
//! aggressively as possible!
use crate::decode;
Second large refactor for WebIDL bindings This commit is the second, and hopefully last massive, refactor for using WebIDL bindings internally in `wasm-bindgen`. This commit actually fully executes on the task at hand, moving `wasm-bindgen` to internally using WebIDL bindings throughout its code generation, anyref passes, etc. This actually fixes a number of issues that have existed in the anyref pass for some time now! The main changes here are to basically remove the usage of `Descriptor` from generating JS bindings. Instead two new types are introduced: `NonstandardIncoming` and `NonstandardOutgoing` which are bindings lists used for incoming/outgoing bindings. These mirror the standard terminology and literally have variants which are the standard values. All `Descriptor` types are now mapped into lists of incoming/outgoing bindings and used for process in wasm-bindgen. All JS generation has been refactored and updated to now process these lists of bindings instead of the previous `Descriptor`. In other words this commit takes `js2rust.rs` and `rust2js.rs` and first splits them in two. Interpretation of `Descriptor` and what to do for conversions is in the binding selection modules. The actual generation of JS from the binding selection is now performed by `incoming.rs` and `outgoing.rs`. To boot this also deduplicates all the code between the argument handling of `js2rust.rs` and return value handling of `rust2js.rs`. This means that to implement a new binding you only need to implement it one place and it's implemented for free in the other! This commit is not the end of the story though. I would like to add a mdoe to `wasm-bindgen` that literally emits a WebIDL bindings section. That's left for a third (and hopefully final) refactoring which is also intended to optimize generated JS for bindings. This commit currently loses the optimization where an imported is hooked up by value directly whenever a shim isn't needed. It's planned that the next refactoring to emit a webidl binding section that can be added back in. It shouldn't be too too hard hopefully since all the scaffolding is in place now. cc #1524
2019-06-10 07:09:51 -07:00
use crate::descriptor::{Descriptor, Function};
First refactor for WebIDL bindings This commit starts the `wasm-bindgen` CLI tool down the road to being a true polyfill for WebIDL bindings. This refactor is probably the first of a few, but is hopefully the largest and most sprawling and everything will be a bit more targeted from here on out. The goal of this refactoring is to separate out the massive `crates/cli-support/src/js/mod.rs` into a number of separate pieces of functionality. It currently takes care of basically everything including: * Binding intrinsics * Handling anyref transformations * Generating all JS for imports/exports * All the logic for how to import and how to name imports * Execution and management of wasm-bindgen closures Many of these are separable concerns and most overlap with WebIDL bindings. The internal refactoring here is intended to make it more clear who's responsible for what as well as making some existing operations much more straightforward. At a high-level, the following changes are done: 1. A `src/webidl.rs` module is introduced. The purpose of this module is to take all of the raw wasm-bindgen custom sections from the module and transform them into a WebIDL bindings section. This module has a placeholder `WebidlCustomSection` which is nowhere near the actual custom section but if you squint is in theory very similar. It's hoped that this will eventually become the true WebIDL custom section, currently being developed in an external crate. Currently, however, the WebIDL bindings custom section only covers a subset of the functionality we export to wasm-bindgen users. To avoid leaving them high and dry this module also contains an auxiliary custom section named `WasmBindgenAux`. This custom section isn't intended to have a binary format, but is intended to represent a theoretical custom section necessary to couple with WebIDL bindings to achieve all our desired functionality in `wasm-bindgen`. It'll never be standardized, but it'll also never be serialized :) 2. The `src/webidl.rs` module now takes over quite a bit of functionality from `src/js/mod.rs`. Namely it handles synthesis of an `export_map` and an `import_map` mapping export/import IDs to exactly what's expected to be hooked up there. This does not include type information (as that's in the bindings section) but rather includes things like "this is the method of class A" or "this import is from module `foo`" and things like that. These could arguably be subsumed by future JS features as well, but that's for another time! 3. All handling of wasm-bindgen "descriptor functions" now happens in a dedicated `src/descriptors.rs` module. The output of this module is its own custom section (intended to be immediately consumed by the WebIDL module) which is in theory what we want to ourselves emit one day but rustc isn't capable of doing so right now. 4. Invocations and generations of imports are completely overhauled. Using the `import_map` generated in the WebIDL step all imports are now handled much more precisely in one location rather than haphazardly throughout the module. This means we have precise information about each import of the module and we only modify exactly what we're looking at. This also vastly simplifies intrinsic generation since it's all simply a codegen part of the `rust2js.rs` module now. 5. Handling of direct imports which don't have a JS shim generated is slightly different from before and is intended to be future-compatible with WebIDL bindings in its full glory, but we'll need to update it to handle cases for constructors and method calls eventually as well. 6. Intrinsic definitions now live in their own file (`src/intrinsic.rs`) and have a separated definition for their symbol name and signature. The actual implementation of each intrinsic lives in `rust2js.rs` There's a number of TODO items to finish before this merges. This includes reimplementing the anyref pass and actually implementing import maps for other targets. Those will come soon in follow-up commits, but the entire `tests/wasm/main.rs` suite is currently passing and this seems like a good checkpoint.
2019-05-23 09:15:26 -07:00
use crate::descriptors::WasmBindgenDescriptorsSection;
use crate::intrinsic::Intrinsic;
use failure::{bail, Error};
use std::borrow::Cow;
use std::collections::{HashMap, HashSet};
use std::path::PathBuf;
use std::str;
use walrus::{ExportId, FunctionId, ImportId, Module, TypedCustomSectionId};
use wasm_bindgen_shared::struct_function_export_name;
Second large refactor for WebIDL bindings This commit is the second, and hopefully last massive, refactor for using WebIDL bindings internally in `wasm-bindgen`. This commit actually fully executes on the task at hand, moving `wasm-bindgen` to internally using WebIDL bindings throughout its code generation, anyref passes, etc. This actually fixes a number of issues that have existed in the anyref pass for some time now! The main changes here are to basically remove the usage of `Descriptor` from generating JS bindings. Instead two new types are introduced: `NonstandardIncoming` and `NonstandardOutgoing` which are bindings lists used for incoming/outgoing bindings. These mirror the standard terminology and literally have variants which are the standard values. All `Descriptor` types are now mapped into lists of incoming/outgoing bindings and used for process in wasm-bindgen. All JS generation has been refactored and updated to now process these lists of bindings instead of the previous `Descriptor`. In other words this commit takes `js2rust.rs` and `rust2js.rs` and first splits them in two. Interpretation of `Descriptor` and what to do for conversions is in the binding selection modules. The actual generation of JS from the binding selection is now performed by `incoming.rs` and `outgoing.rs`. To boot this also deduplicates all the code between the argument handling of `js2rust.rs` and return value handling of `rust2js.rs`. This means that to implement a new binding you only need to implement it one place and it's implemented for free in the other! This commit is not the end of the story though. I would like to add a mdoe to `wasm-bindgen` that literally emits a WebIDL bindings section. That's left for a third (and hopefully final) refactoring which is also intended to optimize generated JS for bindings. This commit currently loses the optimization where an imported is hooked up by value directly whenever a shim isn't needed. It's planned that the next refactoring to emit a webidl binding section that can be added back in. It shouldn't be too too hard hopefully since all the scaffolding is in place now. cc #1524
2019-06-10 07:09:51 -07:00
use wasm_webidl_bindings::ast;
First refactor for WebIDL bindings This commit starts the `wasm-bindgen` CLI tool down the road to being a true polyfill for WebIDL bindings. This refactor is probably the first of a few, but is hopefully the largest and most sprawling and everything will be a bit more targeted from here on out. The goal of this refactoring is to separate out the massive `crates/cli-support/src/js/mod.rs` into a number of separate pieces of functionality. It currently takes care of basically everything including: * Binding intrinsics * Handling anyref transformations * Generating all JS for imports/exports * All the logic for how to import and how to name imports * Execution and management of wasm-bindgen closures Many of these are separable concerns and most overlap with WebIDL bindings. The internal refactoring here is intended to make it more clear who's responsible for what as well as making some existing operations much more straightforward. At a high-level, the following changes are done: 1. A `src/webidl.rs` module is introduced. The purpose of this module is to take all of the raw wasm-bindgen custom sections from the module and transform them into a WebIDL bindings section. This module has a placeholder `WebidlCustomSection` which is nowhere near the actual custom section but if you squint is in theory very similar. It's hoped that this will eventually become the true WebIDL custom section, currently being developed in an external crate. Currently, however, the WebIDL bindings custom section only covers a subset of the functionality we export to wasm-bindgen users. To avoid leaving them high and dry this module also contains an auxiliary custom section named `WasmBindgenAux`. This custom section isn't intended to have a binary format, but is intended to represent a theoretical custom section necessary to couple with WebIDL bindings to achieve all our desired functionality in `wasm-bindgen`. It'll never be standardized, but it'll also never be serialized :) 2. The `src/webidl.rs` module now takes over quite a bit of functionality from `src/js/mod.rs`. Namely it handles synthesis of an `export_map` and an `import_map` mapping export/import IDs to exactly what's expected to be hooked up there. This does not include type information (as that's in the bindings section) but rather includes things like "this is the method of class A" or "this import is from module `foo`" and things like that. These could arguably be subsumed by future JS features as well, but that's for another time! 3. All handling of wasm-bindgen "descriptor functions" now happens in a dedicated `src/descriptors.rs` module. The output of this module is its own custom section (intended to be immediately consumed by the WebIDL module) which is in theory what we want to ourselves emit one day but rustc isn't capable of doing so right now. 4. Invocations and generations of imports are completely overhauled. Using the `import_map` generated in the WebIDL step all imports are now handled much more precisely in one location rather than haphazardly throughout the module. This means we have precise information about each import of the module and we only modify exactly what we're looking at. This also vastly simplifies intrinsic generation since it's all simply a codegen part of the `rust2js.rs` module now. 5. Handling of direct imports which don't have a JS shim generated is slightly different from before and is intended to be future-compatible with WebIDL bindings in its full glory, but we'll need to update it to handle cases for constructors and method calls eventually as well. 6. Intrinsic definitions now live in their own file (`src/intrinsic.rs`) and have a separated definition for their symbol name and signature. The actual implementation of each intrinsic lives in `rust2js.rs` There's a number of TODO items to finish before this merges. This includes reimplementing the anyref pass and actually implementing import maps for other targets. Those will come soon in follow-up commits, but the entire `tests/wasm/main.rs` suite is currently passing and this seems like a good checkpoint.
2019-05-23 09:15:26 -07:00
const PLACEHOLDER_MODULE: &str = "__wbindgen_placeholder__";
Second large refactor for WebIDL bindings This commit is the second, and hopefully last massive, refactor for using WebIDL bindings internally in `wasm-bindgen`. This commit actually fully executes on the task at hand, moving `wasm-bindgen` to internally using WebIDL bindings throughout its code generation, anyref passes, etc. This actually fixes a number of issues that have existed in the anyref pass for some time now! The main changes here are to basically remove the usage of `Descriptor` from generating JS bindings. Instead two new types are introduced: `NonstandardIncoming` and `NonstandardOutgoing` which are bindings lists used for incoming/outgoing bindings. These mirror the standard terminology and literally have variants which are the standard values. All `Descriptor` types are now mapped into lists of incoming/outgoing bindings and used for process in wasm-bindgen. All JS generation has been refactored and updated to now process these lists of bindings instead of the previous `Descriptor`. In other words this commit takes `js2rust.rs` and `rust2js.rs` and first splits them in two. Interpretation of `Descriptor` and what to do for conversions is in the binding selection modules. The actual generation of JS from the binding selection is now performed by `incoming.rs` and `outgoing.rs`. To boot this also deduplicates all the code between the argument handling of `js2rust.rs` and return value handling of `rust2js.rs`. This means that to implement a new binding you only need to implement it one place and it's implemented for free in the other! This commit is not the end of the story though. I would like to add a mdoe to `wasm-bindgen` that literally emits a WebIDL bindings section. That's left for a third (and hopefully final) refactoring which is also intended to optimize generated JS for bindings. This commit currently loses the optimization where an imported is hooked up by value directly whenever a shim isn't needed. It's planned that the next refactoring to emit a webidl binding section that can be added back in. It shouldn't be too too hard hopefully since all the scaffolding is in place now. cc #1524
2019-06-10 07:09:51 -07:00
mod bindings;
mod incoming;
mod outgoing;
pub use self::incoming::NonstandardIncoming;
pub use self::outgoing::NonstandardOutgoing;
/// A nonstandard wasm-bindgen-specific WebIDL custom section.
///
/// This nonstandard section is intended to convey all information that
/// wasm-bindgen itself needs to know about binding functions. This means that
/// it basically uses `NonstandardIncoming` instead of
/// `IncomingBindingExpression` and such. It's also in a bit easier to work with
/// format than the official WebIDL bindings custom section.
///
/// Note that this is intended to be consumed during generation of JS shims and
/// bindings. There it can be transformed, however, into an actual WebIDL
/// binding section using all of the values it has internally.
First refactor for WebIDL bindings This commit starts the `wasm-bindgen` CLI tool down the road to being a true polyfill for WebIDL bindings. This refactor is probably the first of a few, but is hopefully the largest and most sprawling and everything will be a bit more targeted from here on out. The goal of this refactoring is to separate out the massive `crates/cli-support/src/js/mod.rs` into a number of separate pieces of functionality. It currently takes care of basically everything including: * Binding intrinsics * Handling anyref transformations * Generating all JS for imports/exports * All the logic for how to import and how to name imports * Execution and management of wasm-bindgen closures Many of these are separable concerns and most overlap with WebIDL bindings. The internal refactoring here is intended to make it more clear who's responsible for what as well as making some existing operations much more straightforward. At a high-level, the following changes are done: 1. A `src/webidl.rs` module is introduced. The purpose of this module is to take all of the raw wasm-bindgen custom sections from the module and transform them into a WebIDL bindings section. This module has a placeholder `WebidlCustomSection` which is nowhere near the actual custom section but if you squint is in theory very similar. It's hoped that this will eventually become the true WebIDL custom section, currently being developed in an external crate. Currently, however, the WebIDL bindings custom section only covers a subset of the functionality we export to wasm-bindgen users. To avoid leaving them high and dry this module also contains an auxiliary custom section named `WasmBindgenAux`. This custom section isn't intended to have a binary format, but is intended to represent a theoretical custom section necessary to couple with WebIDL bindings to achieve all our desired functionality in `wasm-bindgen`. It'll never be standardized, but it'll also never be serialized :) 2. The `src/webidl.rs` module now takes over quite a bit of functionality from `src/js/mod.rs`. Namely it handles synthesis of an `export_map` and an `import_map` mapping export/import IDs to exactly what's expected to be hooked up there. This does not include type information (as that's in the bindings section) but rather includes things like "this is the method of class A" or "this import is from module `foo`" and things like that. These could arguably be subsumed by future JS features as well, but that's for another time! 3. All handling of wasm-bindgen "descriptor functions" now happens in a dedicated `src/descriptors.rs` module. The output of this module is its own custom section (intended to be immediately consumed by the WebIDL module) which is in theory what we want to ourselves emit one day but rustc isn't capable of doing so right now. 4. Invocations and generations of imports are completely overhauled. Using the `import_map` generated in the WebIDL step all imports are now handled much more precisely in one location rather than haphazardly throughout the module. This means we have precise information about each import of the module and we only modify exactly what we're looking at. This also vastly simplifies intrinsic generation since it's all simply a codegen part of the `rust2js.rs` module now. 5. Handling of direct imports which don't have a JS shim generated is slightly different from before and is intended to be future-compatible with WebIDL bindings in its full glory, but we'll need to update it to handle cases for constructors and method calls eventually as well. 6. Intrinsic definitions now live in their own file (`src/intrinsic.rs`) and have a separated definition for their symbol name and signature. The actual implementation of each intrinsic lives in `rust2js.rs` There's a number of TODO items to finish before this merges. This includes reimplementing the anyref pass and actually implementing import maps for other targets. Those will come soon in follow-up commits, but the entire `tests/wasm/main.rs` suite is currently passing and this seems like a good checkpoint.
2019-05-23 09:15:26 -07:00
#[derive(Default, Debug)]
Second large refactor for WebIDL bindings This commit is the second, and hopefully last massive, refactor for using WebIDL bindings internally in `wasm-bindgen`. This commit actually fully executes on the task at hand, moving `wasm-bindgen` to internally using WebIDL bindings throughout its code generation, anyref passes, etc. This actually fixes a number of issues that have existed in the anyref pass for some time now! The main changes here are to basically remove the usage of `Descriptor` from generating JS bindings. Instead two new types are introduced: `NonstandardIncoming` and `NonstandardOutgoing` which are bindings lists used for incoming/outgoing bindings. These mirror the standard terminology and literally have variants which are the standard values. All `Descriptor` types are now mapped into lists of incoming/outgoing bindings and used for process in wasm-bindgen. All JS generation has been refactored and updated to now process these lists of bindings instead of the previous `Descriptor`. In other words this commit takes `js2rust.rs` and `rust2js.rs` and first splits them in two. Interpretation of `Descriptor` and what to do for conversions is in the binding selection modules. The actual generation of JS from the binding selection is now performed by `incoming.rs` and `outgoing.rs`. To boot this also deduplicates all the code between the argument handling of `js2rust.rs` and return value handling of `rust2js.rs`. This means that to implement a new binding you only need to implement it one place and it's implemented for free in the other! This commit is not the end of the story though. I would like to add a mdoe to `wasm-bindgen` that literally emits a WebIDL bindings section. That's left for a third (and hopefully final) refactoring which is also intended to optimize generated JS for bindings. This commit currently loses the optimization where an imported is hooked up by value directly whenever a shim isn't needed. It's planned that the next refactoring to emit a webidl binding section that can be added back in. It shouldn't be too too hard hopefully since all the scaffolding is in place now. cc #1524
2019-06-10 07:09:51 -07:00
pub struct NonstandardWebidlSection {
/// Store of all WebIDL types. Used currently to store all function types
/// specified in `Bindings`. This is intended to be passed through verbatim
/// to a final WebIDL bindings section.
pub types: ast::WebidlTypes,
/// A mapping from all bound exported functions to the binding that we have
/// listed for them. This is the master list of every binding that will be
/// bound and have a shim generated for it in the wasm module.
pub exports: HashMap<ExportId, Binding>,
/// Similar to `exports` above, except for imports. This will describe all
/// imports from the wasm module to indicate what the JS shim is expected to
/// do.
pub imports: HashMap<ImportId, Binding>,
/// For closures and such we'll be calling entries in the function table
/// with rich arguments (just like we call exports) so to do that we
/// describe all the elem indices that we need to modify here as well.
First refactor for WebIDL bindings This commit starts the `wasm-bindgen` CLI tool down the road to being a true polyfill for WebIDL bindings. This refactor is probably the first of a few, but is hopefully the largest and most sprawling and everything will be a bit more targeted from here on out. The goal of this refactoring is to separate out the massive `crates/cli-support/src/js/mod.rs` into a number of separate pieces of functionality. It currently takes care of basically everything including: * Binding intrinsics * Handling anyref transformations * Generating all JS for imports/exports * All the logic for how to import and how to name imports * Execution and management of wasm-bindgen closures Many of these are separable concerns and most overlap with WebIDL bindings. The internal refactoring here is intended to make it more clear who's responsible for what as well as making some existing operations much more straightforward. At a high-level, the following changes are done: 1. A `src/webidl.rs` module is introduced. The purpose of this module is to take all of the raw wasm-bindgen custom sections from the module and transform them into a WebIDL bindings section. This module has a placeholder `WebidlCustomSection` which is nowhere near the actual custom section but if you squint is in theory very similar. It's hoped that this will eventually become the true WebIDL custom section, currently being developed in an external crate. Currently, however, the WebIDL bindings custom section only covers a subset of the functionality we export to wasm-bindgen users. To avoid leaving them high and dry this module also contains an auxiliary custom section named `WasmBindgenAux`. This custom section isn't intended to have a binary format, but is intended to represent a theoretical custom section necessary to couple with WebIDL bindings to achieve all our desired functionality in `wasm-bindgen`. It'll never be standardized, but it'll also never be serialized :) 2. The `src/webidl.rs` module now takes over quite a bit of functionality from `src/js/mod.rs`. Namely it handles synthesis of an `export_map` and an `import_map` mapping export/import IDs to exactly what's expected to be hooked up there. This does not include type information (as that's in the bindings section) but rather includes things like "this is the method of class A" or "this import is from module `foo`" and things like that. These could arguably be subsumed by future JS features as well, but that's for another time! 3. All handling of wasm-bindgen "descriptor functions" now happens in a dedicated `src/descriptors.rs` module. The output of this module is its own custom section (intended to be immediately consumed by the WebIDL module) which is in theory what we want to ourselves emit one day but rustc isn't capable of doing so right now. 4. Invocations and generations of imports are completely overhauled. Using the `import_map` generated in the WebIDL step all imports are now handled much more precisely in one location rather than haphazardly throughout the module. This means we have precise information about each import of the module and we only modify exactly what we're looking at. This also vastly simplifies intrinsic generation since it's all simply a codegen part of the `rust2js.rs` module now. 5. Handling of direct imports which don't have a JS shim generated is slightly different from before and is intended to be future-compatible with WebIDL bindings in its full glory, but we'll need to update it to handle cases for constructors and method calls eventually as well. 6. Intrinsic definitions now live in their own file (`src/intrinsic.rs`) and have a separated definition for their symbol name and signature. The actual implementation of each intrinsic lives in `rust2js.rs` There's a number of TODO items to finish before this merges. This includes reimplementing the anyref pass and actually implementing import maps for other targets. Those will come soon in follow-up commits, but the entire `tests/wasm/main.rs` suite is currently passing and this seems like a good checkpoint.
2019-05-23 09:15:26 -07:00
///
Second large refactor for WebIDL bindings This commit is the second, and hopefully last massive, refactor for using WebIDL bindings internally in `wasm-bindgen`. This commit actually fully executes on the task at hand, moving `wasm-bindgen` to internally using WebIDL bindings throughout its code generation, anyref passes, etc. This actually fixes a number of issues that have existed in the anyref pass for some time now! The main changes here are to basically remove the usage of `Descriptor` from generating JS bindings. Instead two new types are introduced: `NonstandardIncoming` and `NonstandardOutgoing` which are bindings lists used for incoming/outgoing bindings. These mirror the standard terminology and literally have variants which are the standard values. All `Descriptor` types are now mapped into lists of incoming/outgoing bindings and used for process in wasm-bindgen. All JS generation has been refactored and updated to now process these lists of bindings instead of the previous `Descriptor`. In other words this commit takes `js2rust.rs` and `rust2js.rs` and first splits them in two. Interpretation of `Descriptor` and what to do for conversions is in the binding selection modules. The actual generation of JS from the binding selection is now performed by `incoming.rs` and `outgoing.rs`. To boot this also deduplicates all the code between the argument handling of `js2rust.rs` and return value handling of `rust2js.rs`. This means that to implement a new binding you only need to implement it one place and it's implemented for free in the other! This commit is not the end of the story though. I would like to add a mdoe to `wasm-bindgen` that literally emits a WebIDL bindings section. That's left for a third (and hopefully final) refactoring which is also intended to optimize generated JS for bindings. This commit currently loses the optimization where an imported is hooked up by value directly whenever a shim isn't needed. It's planned that the next refactoring to emit a webidl binding section that can be added back in. It shouldn't be too too hard hopefully since all the scaffolding is in place now. cc #1524
2019-06-10 07:09:51 -07:00
/// This is a list of pairs where the first element in the list is the
/// element index in the function table being described and the `Binding`
/// describes the signature that it's supposed to have.
First refactor for WebIDL bindings This commit starts the `wasm-bindgen` CLI tool down the road to being a true polyfill for WebIDL bindings. This refactor is probably the first of a few, but is hopefully the largest and most sprawling and everything will be a bit more targeted from here on out. The goal of this refactoring is to separate out the massive `crates/cli-support/src/js/mod.rs` into a number of separate pieces of functionality. It currently takes care of basically everything including: * Binding intrinsics * Handling anyref transformations * Generating all JS for imports/exports * All the logic for how to import and how to name imports * Execution and management of wasm-bindgen closures Many of these are separable concerns and most overlap with WebIDL bindings. The internal refactoring here is intended to make it more clear who's responsible for what as well as making some existing operations much more straightforward. At a high-level, the following changes are done: 1. A `src/webidl.rs` module is introduced. The purpose of this module is to take all of the raw wasm-bindgen custom sections from the module and transform them into a WebIDL bindings section. This module has a placeholder `WebidlCustomSection` which is nowhere near the actual custom section but if you squint is in theory very similar. It's hoped that this will eventually become the true WebIDL custom section, currently being developed in an external crate. Currently, however, the WebIDL bindings custom section only covers a subset of the functionality we export to wasm-bindgen users. To avoid leaving them high and dry this module also contains an auxiliary custom section named `WasmBindgenAux`. This custom section isn't intended to have a binary format, but is intended to represent a theoretical custom section necessary to couple with WebIDL bindings to achieve all our desired functionality in `wasm-bindgen`. It'll never be standardized, but it'll also never be serialized :) 2. The `src/webidl.rs` module now takes over quite a bit of functionality from `src/js/mod.rs`. Namely it handles synthesis of an `export_map` and an `import_map` mapping export/import IDs to exactly what's expected to be hooked up there. This does not include type information (as that's in the bindings section) but rather includes things like "this is the method of class A" or "this import is from module `foo`" and things like that. These could arguably be subsumed by future JS features as well, but that's for another time! 3. All handling of wasm-bindgen "descriptor functions" now happens in a dedicated `src/descriptors.rs` module. The output of this module is its own custom section (intended to be immediately consumed by the WebIDL module) which is in theory what we want to ourselves emit one day but rustc isn't capable of doing so right now. 4. Invocations and generations of imports are completely overhauled. Using the `import_map` generated in the WebIDL step all imports are now handled much more precisely in one location rather than haphazardly throughout the module. This means we have precise information about each import of the module and we only modify exactly what we're looking at. This also vastly simplifies intrinsic generation since it's all simply a codegen part of the `rust2js.rs` module now. 5. Handling of direct imports which don't have a JS shim generated is slightly different from before and is intended to be future-compatible with WebIDL bindings in its full glory, but we'll need to update it to handle cases for constructors and method calls eventually as well. 6. Intrinsic definitions now live in their own file (`src/intrinsic.rs`) and have a separated definition for their symbol name and signature. The actual implementation of each intrinsic lives in `rust2js.rs` There's a number of TODO items to finish before this merges. This includes reimplementing the anyref pass and actually implementing import maps for other targets. Those will come soon in follow-up commits, but the entire `tests/wasm/main.rs` suite is currently passing and this seems like a good checkpoint.
2019-05-23 09:15:26 -07:00
///
Second large refactor for WebIDL bindings This commit is the second, and hopefully last massive, refactor for using WebIDL bindings internally in `wasm-bindgen`. This commit actually fully executes on the task at hand, moving `wasm-bindgen` to internally using WebIDL bindings throughout its code generation, anyref passes, etc. This actually fixes a number of issues that have existed in the anyref pass for some time now! The main changes here are to basically remove the usage of `Descriptor` from generating JS bindings. Instead two new types are introduced: `NonstandardIncoming` and `NonstandardOutgoing` which are bindings lists used for incoming/outgoing bindings. These mirror the standard terminology and literally have variants which are the standard values. All `Descriptor` types are now mapped into lists of incoming/outgoing bindings and used for process in wasm-bindgen. All JS generation has been refactored and updated to now process these lists of bindings instead of the previous `Descriptor`. In other words this commit takes `js2rust.rs` and `rust2js.rs` and first splits them in two. Interpretation of `Descriptor` and what to do for conversions is in the binding selection modules. The actual generation of JS from the binding selection is now performed by `incoming.rs` and `outgoing.rs`. To boot this also deduplicates all the code between the argument handling of `js2rust.rs` and return value handling of `rust2js.rs`. This means that to implement a new binding you only need to implement it one place and it's implemented for free in the other! This commit is not the end of the story though. I would like to add a mdoe to `wasm-bindgen` that literally emits a WebIDL bindings section. That's left for a third (and hopefully final) refactoring which is also intended to optimize generated JS for bindings. This commit currently loses the optimization where an imported is hooked up by value directly whenever a shim isn't needed. It's planned that the next refactoring to emit a webidl binding section that can be added back in. It shouldn't be too too hard hopefully since all the scaffolding is in place now. cc #1524
2019-06-10 07:09:51 -07:00
/// The index within this table itself is then used to call actually
/// transformed functions.
pub elems: Vec<(u32, Binding)>,
First refactor for WebIDL bindings This commit starts the `wasm-bindgen` CLI tool down the road to being a true polyfill for WebIDL bindings. This refactor is probably the first of a few, but is hopefully the largest and most sprawling and everything will be a bit more targeted from here on out. The goal of this refactoring is to separate out the massive `crates/cli-support/src/js/mod.rs` into a number of separate pieces of functionality. It currently takes care of basically everything including: * Binding intrinsics * Handling anyref transformations * Generating all JS for imports/exports * All the logic for how to import and how to name imports * Execution and management of wasm-bindgen closures Many of these are separable concerns and most overlap with WebIDL bindings. The internal refactoring here is intended to make it more clear who's responsible for what as well as making some existing operations much more straightforward. At a high-level, the following changes are done: 1. A `src/webidl.rs` module is introduced. The purpose of this module is to take all of the raw wasm-bindgen custom sections from the module and transform them into a WebIDL bindings section. This module has a placeholder `WebidlCustomSection` which is nowhere near the actual custom section but if you squint is in theory very similar. It's hoped that this will eventually become the true WebIDL custom section, currently being developed in an external crate. Currently, however, the WebIDL bindings custom section only covers a subset of the functionality we export to wasm-bindgen users. To avoid leaving them high and dry this module also contains an auxiliary custom section named `WasmBindgenAux`. This custom section isn't intended to have a binary format, but is intended to represent a theoretical custom section necessary to couple with WebIDL bindings to achieve all our desired functionality in `wasm-bindgen`. It'll never be standardized, but it'll also never be serialized :) 2. The `src/webidl.rs` module now takes over quite a bit of functionality from `src/js/mod.rs`. Namely it handles synthesis of an `export_map` and an `import_map` mapping export/import IDs to exactly what's expected to be hooked up there. This does not include type information (as that's in the bindings section) but rather includes things like "this is the method of class A" or "this import is from module `foo`" and things like that. These could arguably be subsumed by future JS features as well, but that's for another time! 3. All handling of wasm-bindgen "descriptor functions" now happens in a dedicated `src/descriptors.rs` module. The output of this module is its own custom section (intended to be immediately consumed by the WebIDL module) which is in theory what we want to ourselves emit one day but rustc isn't capable of doing so right now. 4. Invocations and generations of imports are completely overhauled. Using the `import_map` generated in the WebIDL step all imports are now handled much more precisely in one location rather than haphazardly throughout the module. This means we have precise information about each import of the module and we only modify exactly what we're looking at. This also vastly simplifies intrinsic generation since it's all simply a codegen part of the `rust2js.rs` module now. 5. Handling of direct imports which don't have a JS shim generated is slightly different from before and is intended to be future-compatible with WebIDL bindings in its full glory, but we'll need to update it to handle cases for constructors and method calls eventually as well. 6. Intrinsic definitions now live in their own file (`src/intrinsic.rs`) and have a separated definition for their symbol name and signature. The actual implementation of each intrinsic lives in `rust2js.rs` There's a number of TODO items to finish before this merges. This includes reimplementing the anyref pass and actually implementing import maps for other targets. Those will come soon in follow-up commits, but the entire `tests/wasm/main.rs` suite is currently passing and this seems like a good checkpoint.
2019-05-23 09:15:26 -07:00
}
Second large refactor for WebIDL bindings This commit is the second, and hopefully last massive, refactor for using WebIDL bindings internally in `wasm-bindgen`. This commit actually fully executes on the task at hand, moving `wasm-bindgen` to internally using WebIDL bindings throughout its code generation, anyref passes, etc. This actually fixes a number of issues that have existed in the anyref pass for some time now! The main changes here are to basically remove the usage of `Descriptor` from generating JS bindings. Instead two new types are introduced: `NonstandardIncoming` and `NonstandardOutgoing` which are bindings lists used for incoming/outgoing bindings. These mirror the standard terminology and literally have variants which are the standard values. All `Descriptor` types are now mapped into lists of incoming/outgoing bindings and used for process in wasm-bindgen. All JS generation has been refactored and updated to now process these lists of bindings instead of the previous `Descriptor`. In other words this commit takes `js2rust.rs` and `rust2js.rs` and first splits them in two. Interpretation of `Descriptor` and what to do for conversions is in the binding selection modules. The actual generation of JS from the binding selection is now performed by `incoming.rs` and `outgoing.rs`. To boot this also deduplicates all the code between the argument handling of `js2rust.rs` and return value handling of `rust2js.rs`. This means that to implement a new binding you only need to implement it one place and it's implemented for free in the other! This commit is not the end of the story though. I would like to add a mdoe to `wasm-bindgen` that literally emits a WebIDL bindings section. That's left for a third (and hopefully final) refactoring which is also intended to optimize generated JS for bindings. This commit currently loses the optimization where an imported is hooked up by value directly whenever a shim isn't needed. It's planned that the next refactoring to emit a webidl binding section that can be added back in. It shouldn't be too too hard hopefully since all the scaffolding is in place now. cc #1524
2019-06-10 07:09:51 -07:00
pub type NonstandardWebidlSectionId = TypedCustomSectionId<NonstandardWebidlSection>;
First refactor for WebIDL bindings This commit starts the `wasm-bindgen` CLI tool down the road to being a true polyfill for WebIDL bindings. This refactor is probably the first of a few, but is hopefully the largest and most sprawling and everything will be a bit more targeted from here on out. The goal of this refactoring is to separate out the massive `crates/cli-support/src/js/mod.rs` into a number of separate pieces of functionality. It currently takes care of basically everything including: * Binding intrinsics * Handling anyref transformations * Generating all JS for imports/exports * All the logic for how to import and how to name imports * Execution and management of wasm-bindgen closures Many of these are separable concerns and most overlap with WebIDL bindings. The internal refactoring here is intended to make it more clear who's responsible for what as well as making some existing operations much more straightforward. At a high-level, the following changes are done: 1. A `src/webidl.rs` module is introduced. The purpose of this module is to take all of the raw wasm-bindgen custom sections from the module and transform them into a WebIDL bindings section. This module has a placeholder `WebidlCustomSection` which is nowhere near the actual custom section but if you squint is in theory very similar. It's hoped that this will eventually become the true WebIDL custom section, currently being developed in an external crate. Currently, however, the WebIDL bindings custom section only covers a subset of the functionality we export to wasm-bindgen users. To avoid leaving them high and dry this module also contains an auxiliary custom section named `WasmBindgenAux`. This custom section isn't intended to have a binary format, but is intended to represent a theoretical custom section necessary to couple with WebIDL bindings to achieve all our desired functionality in `wasm-bindgen`. It'll never be standardized, but it'll also never be serialized :) 2. The `src/webidl.rs` module now takes over quite a bit of functionality from `src/js/mod.rs`. Namely it handles synthesis of an `export_map` and an `import_map` mapping export/import IDs to exactly what's expected to be hooked up there. This does not include type information (as that's in the bindings section) but rather includes things like "this is the method of class A" or "this import is from module `foo`" and things like that. These could arguably be subsumed by future JS features as well, but that's for another time! 3. All handling of wasm-bindgen "descriptor functions" now happens in a dedicated `src/descriptors.rs` module. The output of this module is its own custom section (intended to be immediately consumed by the WebIDL module) which is in theory what we want to ourselves emit one day but rustc isn't capable of doing so right now. 4. Invocations and generations of imports are completely overhauled. Using the `import_map` generated in the WebIDL step all imports are now handled much more precisely in one location rather than haphazardly throughout the module. This means we have precise information about each import of the module and we only modify exactly what we're looking at. This also vastly simplifies intrinsic generation since it's all simply a codegen part of the `rust2js.rs` module now. 5. Handling of direct imports which don't have a JS shim generated is slightly different from before and is intended to be future-compatible with WebIDL bindings in its full glory, but we'll need to update it to handle cases for constructors and method calls eventually as well. 6. Intrinsic definitions now live in their own file (`src/intrinsic.rs`) and have a separated definition for their symbol name and signature. The actual implementation of each intrinsic lives in `rust2js.rs` There's a number of TODO items to finish before this merges. This includes reimplementing the anyref pass and actually implementing import maps for other targets. Those will come soon in follow-up commits, but the entire `tests/wasm/main.rs` suite is currently passing and this seems like a good checkpoint.
2019-05-23 09:15:26 -07:00
Second large refactor for WebIDL bindings This commit is the second, and hopefully last massive, refactor for using WebIDL bindings internally in `wasm-bindgen`. This commit actually fully executes on the task at hand, moving `wasm-bindgen` to internally using WebIDL bindings throughout its code generation, anyref passes, etc. This actually fixes a number of issues that have existed in the anyref pass for some time now! The main changes here are to basically remove the usage of `Descriptor` from generating JS bindings. Instead two new types are introduced: `NonstandardIncoming` and `NonstandardOutgoing` which are bindings lists used for incoming/outgoing bindings. These mirror the standard terminology and literally have variants which are the standard values. All `Descriptor` types are now mapped into lists of incoming/outgoing bindings and used for process in wasm-bindgen. All JS generation has been refactored and updated to now process these lists of bindings instead of the previous `Descriptor`. In other words this commit takes `js2rust.rs` and `rust2js.rs` and first splits them in two. Interpretation of `Descriptor` and what to do for conversions is in the binding selection modules. The actual generation of JS from the binding selection is now performed by `incoming.rs` and `outgoing.rs`. To boot this also deduplicates all the code between the argument handling of `js2rust.rs` and return value handling of `rust2js.rs`. This means that to implement a new binding you only need to implement it one place and it's implemented for free in the other! This commit is not the end of the story though. I would like to add a mdoe to `wasm-bindgen` that literally emits a WebIDL bindings section. That's left for a third (and hopefully final) refactoring which is also intended to optimize generated JS for bindings. This commit currently loses the optimization where an imported is hooked up by value directly whenever a shim isn't needed. It's planned that the next refactoring to emit a webidl binding section that can be added back in. It shouldn't be too too hard hopefully since all the scaffolding is in place now. cc #1524
2019-06-10 07:09:51 -07:00
/// A non-standard wasm-bindgen-specifi WebIDL binding. This is meant to vaguely
/// resemble a `FuctionBinding` in the official WebIDL bindings proposal, or at
/// least make it very easy to manufacture an official value from this one.
First refactor for WebIDL bindings This commit starts the `wasm-bindgen` CLI tool down the road to being a true polyfill for WebIDL bindings. This refactor is probably the first of a few, but is hopefully the largest and most sprawling and everything will be a bit more targeted from here on out. The goal of this refactoring is to separate out the massive `crates/cli-support/src/js/mod.rs` into a number of separate pieces of functionality. It currently takes care of basically everything including: * Binding intrinsics * Handling anyref transformations * Generating all JS for imports/exports * All the logic for how to import and how to name imports * Execution and management of wasm-bindgen closures Many of these are separable concerns and most overlap with WebIDL bindings. The internal refactoring here is intended to make it more clear who's responsible for what as well as making some existing operations much more straightforward. At a high-level, the following changes are done: 1. A `src/webidl.rs` module is introduced. The purpose of this module is to take all of the raw wasm-bindgen custom sections from the module and transform them into a WebIDL bindings section. This module has a placeholder `WebidlCustomSection` which is nowhere near the actual custom section but if you squint is in theory very similar. It's hoped that this will eventually become the true WebIDL custom section, currently being developed in an external crate. Currently, however, the WebIDL bindings custom section only covers a subset of the functionality we export to wasm-bindgen users. To avoid leaving them high and dry this module also contains an auxiliary custom section named `WasmBindgenAux`. This custom section isn't intended to have a binary format, but is intended to represent a theoretical custom section necessary to couple with WebIDL bindings to achieve all our desired functionality in `wasm-bindgen`. It'll never be standardized, but it'll also never be serialized :) 2. The `src/webidl.rs` module now takes over quite a bit of functionality from `src/js/mod.rs`. Namely it handles synthesis of an `export_map` and an `import_map` mapping export/import IDs to exactly what's expected to be hooked up there. This does not include type information (as that's in the bindings section) but rather includes things like "this is the method of class A" or "this import is from module `foo`" and things like that. These could arguably be subsumed by future JS features as well, but that's for another time! 3. All handling of wasm-bindgen "descriptor functions" now happens in a dedicated `src/descriptors.rs` module. The output of this module is its own custom section (intended to be immediately consumed by the WebIDL module) which is in theory what we want to ourselves emit one day but rustc isn't capable of doing so right now. 4. Invocations and generations of imports are completely overhauled. Using the `import_map` generated in the WebIDL step all imports are now handled much more precisely in one location rather than haphazardly throughout the module. This means we have precise information about each import of the module and we only modify exactly what we're looking at. This also vastly simplifies intrinsic generation since it's all simply a codegen part of the `rust2js.rs` module now. 5. Handling of direct imports which don't have a JS shim generated is slightly different from before and is intended to be future-compatible with WebIDL bindings in its full glory, but we'll need to update it to handle cases for constructors and method calls eventually as well. 6. Intrinsic definitions now live in their own file (`src/intrinsic.rs`) and have a separated definition for their symbol name and signature. The actual implementation of each intrinsic lives in `rust2js.rs` There's a number of TODO items to finish before this merges. This includes reimplementing the anyref pass and actually implementing import maps for other targets. Those will come soon in follow-up commits, but the entire `tests/wasm/main.rs` suite is currently passing and this seems like a good checkpoint.
2019-05-23 09:15:26 -07:00
#[derive(Debug, Clone)]
Second large refactor for WebIDL bindings This commit is the second, and hopefully last massive, refactor for using WebIDL bindings internally in `wasm-bindgen`. This commit actually fully executes on the task at hand, moving `wasm-bindgen` to internally using WebIDL bindings throughout its code generation, anyref passes, etc. This actually fixes a number of issues that have existed in the anyref pass for some time now! The main changes here are to basically remove the usage of `Descriptor` from generating JS bindings. Instead two new types are introduced: `NonstandardIncoming` and `NonstandardOutgoing` which are bindings lists used for incoming/outgoing bindings. These mirror the standard terminology and literally have variants which are the standard values. All `Descriptor` types are now mapped into lists of incoming/outgoing bindings and used for process in wasm-bindgen. All JS generation has been refactored and updated to now process these lists of bindings instead of the previous `Descriptor`. In other words this commit takes `js2rust.rs` and `rust2js.rs` and first splits them in two. Interpretation of `Descriptor` and what to do for conversions is in the binding selection modules. The actual generation of JS from the binding selection is now performed by `incoming.rs` and `outgoing.rs`. To boot this also deduplicates all the code between the argument handling of `js2rust.rs` and return value handling of `rust2js.rs`. This means that to implement a new binding you only need to implement it one place and it's implemented for free in the other! This commit is not the end of the story though. I would like to add a mdoe to `wasm-bindgen` that literally emits a WebIDL bindings section. That's left for a third (and hopefully final) refactoring which is also intended to optimize generated JS for bindings. This commit currently loses the optimization where an imported is hooked up by value directly whenever a shim isn't needed. It's planned that the next refactoring to emit a webidl binding section that can be added back in. It shouldn't be too too hard hopefully since all the scaffolding is in place now. cc #1524
2019-06-10 07:09:51 -07:00
pub struct Binding {
/// The WebAssembly type that the function is expected to have. Note that
/// this may not match the actual bound function's type! That's because this
/// type includes `anyref` but the Rust compiler never emits anyref. This
/// is, however, used for the `anyref` pass to know what to transform to
/// `anyref`.
pub wasm_ty: walrus::TypeId,
/// The WebIDL type of this binding, which is an index into the webidl
/// binding section's `types` field.
pub webidl_ty: ast::WebidlFunctionId,
/// A list of incoming bindings. For exports this is the list of arguments,
/// and for imports this is the return value.
pub incoming: Vec<NonstandardIncoming>,
/// A list of outgoing bindings. For exports this is the return value and
/// for imports this is the list of arguments.
pub outgoing: Vec<NonstandardOutgoing>,
/// An unfortunate necessity of today's implementation. Ideally WebIDL
/// bindings are used with multi-value support in wasm everywhere, but today
/// few engines support multi-value and LLVM certainly doesn't. Aggregates
/// are then always returned through an out-ptr, so this indicates that if
/// an out-ptr is present what wasm types are being transmitted through it.
pub return_via_outptr: Option<Vec<walrus::ValType>>,
First refactor for WebIDL bindings This commit starts the `wasm-bindgen` CLI tool down the road to being a true polyfill for WebIDL bindings. This refactor is probably the first of a few, but is hopefully the largest and most sprawling and everything will be a bit more targeted from here on out. The goal of this refactoring is to separate out the massive `crates/cli-support/src/js/mod.rs` into a number of separate pieces of functionality. It currently takes care of basically everything including: * Binding intrinsics * Handling anyref transformations * Generating all JS for imports/exports * All the logic for how to import and how to name imports * Execution and management of wasm-bindgen closures Many of these are separable concerns and most overlap with WebIDL bindings. The internal refactoring here is intended to make it more clear who's responsible for what as well as making some existing operations much more straightforward. At a high-level, the following changes are done: 1. A `src/webidl.rs` module is introduced. The purpose of this module is to take all of the raw wasm-bindgen custom sections from the module and transform them into a WebIDL bindings section. This module has a placeholder `WebidlCustomSection` which is nowhere near the actual custom section but if you squint is in theory very similar. It's hoped that this will eventually become the true WebIDL custom section, currently being developed in an external crate. Currently, however, the WebIDL bindings custom section only covers a subset of the functionality we export to wasm-bindgen users. To avoid leaving them high and dry this module also contains an auxiliary custom section named `WasmBindgenAux`. This custom section isn't intended to have a binary format, but is intended to represent a theoretical custom section necessary to couple with WebIDL bindings to achieve all our desired functionality in `wasm-bindgen`. It'll never be standardized, but it'll also never be serialized :) 2. The `src/webidl.rs` module now takes over quite a bit of functionality from `src/js/mod.rs`. Namely it handles synthesis of an `export_map` and an `import_map` mapping export/import IDs to exactly what's expected to be hooked up there. This does not include type information (as that's in the bindings section) but rather includes things like "this is the method of class A" or "this import is from module `foo`" and things like that. These could arguably be subsumed by future JS features as well, but that's for another time! 3. All handling of wasm-bindgen "descriptor functions" now happens in a dedicated `src/descriptors.rs` module. The output of this module is its own custom section (intended to be immediately consumed by the WebIDL module) which is in theory what we want to ourselves emit one day but rustc isn't capable of doing so right now. 4. Invocations and generations of imports are completely overhauled. Using the `import_map` generated in the WebIDL step all imports are now handled much more precisely in one location rather than haphazardly throughout the module. This means we have precise information about each import of the module and we only modify exactly what we're looking at. This also vastly simplifies intrinsic generation since it's all simply a codegen part of the `rust2js.rs` module now. 5. Handling of direct imports which don't have a JS shim generated is slightly different from before and is intended to be future-compatible with WebIDL bindings in its full glory, but we'll need to update it to handle cases for constructors and method calls eventually as well. 6. Intrinsic definitions now live in their own file (`src/intrinsic.rs`) and have a separated definition for their symbol name and signature. The actual implementation of each intrinsic lives in `rust2js.rs` There's a number of TODO items to finish before this merges. This includes reimplementing the anyref pass and actually implementing import maps for other targets. Those will come soon in follow-up commits, but the entire `tests/wasm/main.rs` suite is currently passing and this seems like a good checkpoint.
2019-05-23 09:15:26 -07:00
}
impl Binding {
/// Does this binding's wasm function signature have any `anyref`s?
pub fn contains_anyref(&self, module: &walrus::Module) -> bool {
let ty = module.types.get(self.wasm_ty);
ty.params()
.iter()
.chain(ty.results())
.any(|ty| *ty == walrus::ValType::Anyref)
}
}
First refactor for WebIDL bindings This commit starts the `wasm-bindgen` CLI tool down the road to being a true polyfill for WebIDL bindings. This refactor is probably the first of a few, but is hopefully the largest and most sprawling and everything will be a bit more targeted from here on out. The goal of this refactoring is to separate out the massive `crates/cli-support/src/js/mod.rs` into a number of separate pieces of functionality. It currently takes care of basically everything including: * Binding intrinsics * Handling anyref transformations * Generating all JS for imports/exports * All the logic for how to import and how to name imports * Execution and management of wasm-bindgen closures Many of these are separable concerns and most overlap with WebIDL bindings. The internal refactoring here is intended to make it more clear who's responsible for what as well as making some existing operations much more straightforward. At a high-level, the following changes are done: 1. A `src/webidl.rs` module is introduced. The purpose of this module is to take all of the raw wasm-bindgen custom sections from the module and transform them into a WebIDL bindings section. This module has a placeholder `WebidlCustomSection` which is nowhere near the actual custom section but if you squint is in theory very similar. It's hoped that this will eventually become the true WebIDL custom section, currently being developed in an external crate. Currently, however, the WebIDL bindings custom section only covers a subset of the functionality we export to wasm-bindgen users. To avoid leaving them high and dry this module also contains an auxiliary custom section named `WasmBindgenAux`. This custom section isn't intended to have a binary format, but is intended to represent a theoretical custom section necessary to couple with WebIDL bindings to achieve all our desired functionality in `wasm-bindgen`. It'll never be standardized, but it'll also never be serialized :) 2. The `src/webidl.rs` module now takes over quite a bit of functionality from `src/js/mod.rs`. Namely it handles synthesis of an `export_map` and an `import_map` mapping export/import IDs to exactly what's expected to be hooked up there. This does not include type information (as that's in the bindings section) but rather includes things like "this is the method of class A" or "this import is from module `foo`" and things like that. These could arguably be subsumed by future JS features as well, but that's for another time! 3. All handling of wasm-bindgen "descriptor functions" now happens in a dedicated `src/descriptors.rs` module. The output of this module is its own custom section (intended to be immediately consumed by the WebIDL module) which is in theory what we want to ourselves emit one day but rustc isn't capable of doing so right now. 4. Invocations and generations of imports are completely overhauled. Using the `import_map` generated in the WebIDL step all imports are now handled much more precisely in one location rather than haphazardly throughout the module. This means we have precise information about each import of the module and we only modify exactly what we're looking at. This also vastly simplifies intrinsic generation since it's all simply a codegen part of the `rust2js.rs` module now. 5. Handling of direct imports which don't have a JS shim generated is slightly different from before and is intended to be future-compatible with WebIDL bindings in its full glory, but we'll need to update it to handle cases for constructors and method calls eventually as well. 6. Intrinsic definitions now live in their own file (`src/intrinsic.rs`) and have a separated definition for their symbol name and signature. The actual implementation of each intrinsic lives in `rust2js.rs` There's a number of TODO items to finish before this merges. This includes reimplementing the anyref pass and actually implementing import maps for other targets. Those will come soon in follow-up commits, but the entire `tests/wasm/main.rs` suite is currently passing and this seems like a good checkpoint.
2019-05-23 09:15:26 -07:00
/// A synthetic custom section which is not standardized, never will be, and
/// cannot be serialized or parsed. This is synthesized from all of the
/// compiler-emitted wasm-bindgen sections and then immediately removed to be
/// processed in the JS generation pass.
#[derive(Default, Debug)]
pub struct WasmBindgenAux {
/// Extra typescript annotations that should be appended to the generated
/// TypeScript file. This is provided via a custom attribute in Rust code.
pub extra_typescript: String,
/// A map from identifier to the contents of each local module defined via
/// the `#[wasm_bindgen(module = "/foo.js")]` import options.
pub local_modules: HashMap<String, String>,
/// A map from unique crate identifier to the list of inline JS snippets for
/// that crate identifier.
pub snippets: HashMap<String, Vec<String>>,
/// A list of all `package.json` files that are intended to be included in
/// the final build.
pub package_jsons: HashSet<PathBuf>,
/// A map from exported function id to where it's expected to be exported
/// to.
pub export_map: HashMap<ExportId, AuxExport>,
/// A map from imported function id to what it's expected to import.
pub import_map: HashMap<ImportId, AuxImport>,
/// Small bits of metadata about imports.
pub imports_with_catch: HashSet<ImportId>,
pub imports_with_variadic: HashSet<ImportId>,
pub imports_with_assert_no_shim: HashSet<ImportId>,
First refactor for WebIDL bindings This commit starts the `wasm-bindgen` CLI tool down the road to being a true polyfill for WebIDL bindings. This refactor is probably the first of a few, but is hopefully the largest and most sprawling and everything will be a bit more targeted from here on out. The goal of this refactoring is to separate out the massive `crates/cli-support/src/js/mod.rs` into a number of separate pieces of functionality. It currently takes care of basically everything including: * Binding intrinsics * Handling anyref transformations * Generating all JS for imports/exports * All the logic for how to import and how to name imports * Execution and management of wasm-bindgen closures Many of these are separable concerns and most overlap with WebIDL bindings. The internal refactoring here is intended to make it more clear who's responsible for what as well as making some existing operations much more straightforward. At a high-level, the following changes are done: 1. A `src/webidl.rs` module is introduced. The purpose of this module is to take all of the raw wasm-bindgen custom sections from the module and transform them into a WebIDL bindings section. This module has a placeholder `WebidlCustomSection` which is nowhere near the actual custom section but if you squint is in theory very similar. It's hoped that this will eventually become the true WebIDL custom section, currently being developed in an external crate. Currently, however, the WebIDL bindings custom section only covers a subset of the functionality we export to wasm-bindgen users. To avoid leaving them high and dry this module also contains an auxiliary custom section named `WasmBindgenAux`. This custom section isn't intended to have a binary format, but is intended to represent a theoretical custom section necessary to couple with WebIDL bindings to achieve all our desired functionality in `wasm-bindgen`. It'll never be standardized, but it'll also never be serialized :) 2. The `src/webidl.rs` module now takes over quite a bit of functionality from `src/js/mod.rs`. Namely it handles synthesis of an `export_map` and an `import_map` mapping export/import IDs to exactly what's expected to be hooked up there. This does not include type information (as that's in the bindings section) but rather includes things like "this is the method of class A" or "this import is from module `foo`" and things like that. These could arguably be subsumed by future JS features as well, but that's for another time! 3. All handling of wasm-bindgen "descriptor functions" now happens in a dedicated `src/descriptors.rs` module. The output of this module is its own custom section (intended to be immediately consumed by the WebIDL module) which is in theory what we want to ourselves emit one day but rustc isn't capable of doing so right now. 4. Invocations and generations of imports are completely overhauled. Using the `import_map` generated in the WebIDL step all imports are now handled much more precisely in one location rather than haphazardly throughout the module. This means we have precise information about each import of the module and we only modify exactly what we're looking at. This also vastly simplifies intrinsic generation since it's all simply a codegen part of the `rust2js.rs` module now. 5. Handling of direct imports which don't have a JS shim generated is slightly different from before and is intended to be future-compatible with WebIDL bindings in its full glory, but we'll need to update it to handle cases for constructors and method calls eventually as well. 6. Intrinsic definitions now live in their own file (`src/intrinsic.rs`) and have a separated definition for their symbol name and signature. The actual implementation of each intrinsic lives in `rust2js.rs` There's a number of TODO items to finish before this merges. This includes reimplementing the anyref pass and actually implementing import maps for other targets. Those will come soon in follow-up commits, but the entire `tests/wasm/main.rs` suite is currently passing and this seems like a good checkpoint.
2019-05-23 09:15:26 -07:00
/// Auxiliary information to go into JS/TypeScript bindings describing the
/// exported enums from Rust.
pub enums: Vec<AuxEnum>,
/// Auxiliary information to go into JS/TypeScript bindings describing the
/// exported structs from Rust and their fields they've got exported.
pub structs: Vec<AuxStruct>,
}
pub type WasmBindgenAuxId = TypedCustomSectionId<WasmBindgenAux>;
#[derive(Debug)]
pub struct AuxExport {
/// When generating errors about this export, a helpful name to remember it
/// by.
pub debug_name: String,
/// Comments parsed in Rust and forwarded here to show up in JS bindings.
pub comments: String,
/// Argument names in Rust forwarded here to configure the names that show
/// up in TypeScript bindings.
pub arg_names: Option<Vec<String>>,
/// What kind of function this is and where it shows up
pub kind: AuxExportKind,
}
/// All possible kinds of exports from a wasm module.
///
/// This `enum` says where to place an exported wasm function. For example it
/// may want to get hooked up to a JS class, or it may want to be exported as a
/// free function (etc).
///
/// TODO: it feels like this should not really be here per se. We probably want
/// to either construct the JS object itself from within wasm or somehow move
/// more of this information into some other section. Really what this is is
/// sort of an "export map" saying how to wire up all the free functions from
/// the wasm module into the output expected JS module. All our functions here
/// currently take integer parameters and require a JS wrapper, but ideally
/// we'd change them one day to taking/receiving `anyref` which then use some
/// sort of webidl import to customize behavior or something like that. In any
/// case this doesn't feel quite right in terms of priviledge separation, so
/// we'll want to work on this. For now though it works.
#[derive(Debug)]
pub enum AuxExportKind {
/// A free function that's just listed on the exported module
Function(String),
/// A function that's used to create an instane of a class. The function
/// actually return just an integer which is put on an JS object currently.
Constructor(String),
/// This function is intended to be a getter for a field on a class. The
/// first argument is the internal pointer and the returned value is
/// expected to be the field.
Getter { class: String, field: String },
/// This function is intended to be a setter for a field on a class. The
/// first argument is the internal pointer and the second argument is
/// expected to be the field's new value.
Setter { class: String, field: String },
/// This is a free function (ish) but scoped inside of a class name.
StaticFunction { class: String, name: String },
/// This is a member function of a class where the first parameter is the
/// implicit integer stored in the class instance.
Method {
class: String,
name: String,
/// Whether or not this is calling a by-value method in Rust and should
/// clear the internal pointer in JS automatically.
consumed: bool,
},
}
#[derive(Debug)]
pub struct AuxEnum {
/// The name of this enum
pub name: String,
/// The copied Rust comments to forward to JS
pub comments: String,
/// A list of variants with their name and value
pub variants: Vec<(String, u32)>,
}
#[derive(Debug)]
pub struct AuxStruct {
/// The name of this struct
pub name: String,
/// The copied Rust comments to forward to JS
pub comments: String,
}
/// All possible types of imports that can be imported by a wasm module.
///
/// This `enum` is intended to map out what an imported value is. For example
/// this contains a ton of shims and various ways you can call a function. The
/// base variant here is `Value` which simply means "hook this up to the import"
/// and the signatures will match up.
///
/// Note that this is *not* the same as the webidl bindings section. This is
/// intended to be coupled with that to map out what actually gets hooked up to
/// an import in the wasm module. The two work in tandem.
///
/// Some of these items here are native to JS (like `Value`, indexing
/// operations, etc). Others are shims generated by wasm-bindgen (like `Closure`
/// or `Instanceof`).
#[derive(Debug)]
pub enum AuxImport {
/// This import is expected to simply be whatever is the value that's
/// imported
Value(AuxValue),
/// This import is expected to be a function that takes an `anyref` and
/// returns a `bool`. It's expected that it tests if the argument is an
/// instance of (using `instanceof`) the name specified.
///
/// TODO: can we use `Reflect` or something like that to avoid an extra kind
/// of import here?
Instanceof(JsImport),
/// This import is expected to be a shim that returns the JS value named by
/// `JsImport`.
Static(JsImport),
/// This import is intended to manufacture a JS closure with the given
/// signature and then return that back to Rust.
Second large refactor for WebIDL bindings This commit is the second, and hopefully last massive, refactor for using WebIDL bindings internally in `wasm-bindgen`. This commit actually fully executes on the task at hand, moving `wasm-bindgen` to internally using WebIDL bindings throughout its code generation, anyref passes, etc. This actually fixes a number of issues that have existed in the anyref pass for some time now! The main changes here are to basically remove the usage of `Descriptor` from generating JS bindings. Instead two new types are introduced: `NonstandardIncoming` and `NonstandardOutgoing` which are bindings lists used for incoming/outgoing bindings. These mirror the standard terminology and literally have variants which are the standard values. All `Descriptor` types are now mapped into lists of incoming/outgoing bindings and used for process in wasm-bindgen. All JS generation has been refactored and updated to now process these lists of bindings instead of the previous `Descriptor`. In other words this commit takes `js2rust.rs` and `rust2js.rs` and first splits them in two. Interpretation of `Descriptor` and what to do for conversions is in the binding selection modules. The actual generation of JS from the binding selection is now performed by `incoming.rs` and `outgoing.rs`. To boot this also deduplicates all the code between the argument handling of `js2rust.rs` and return value handling of `rust2js.rs`. This means that to implement a new binding you only need to implement it one place and it's implemented for free in the other! This commit is not the end of the story though. I would like to add a mdoe to `wasm-bindgen` that literally emits a WebIDL bindings section. That's left for a third (and hopefully final) refactoring which is also intended to optimize generated JS for bindings. This commit currently loses the optimization where an imported is hooked up by value directly whenever a shim isn't needed. It's planned that the next refactoring to emit a webidl binding section that can be added back in. It shouldn't be too too hard hopefully since all the scaffolding is in place now. cc #1524
2019-06-10 07:09:51 -07:00
Closure {
mutable: bool, // whether or not this was a `FnMut` closure
dtor: u32, // table element index of the destructor function
binding_idx: u32,
nargs: usize,
},
First refactor for WebIDL bindings This commit starts the `wasm-bindgen` CLI tool down the road to being a true polyfill for WebIDL bindings. This refactor is probably the first of a few, but is hopefully the largest and most sprawling and everything will be a bit more targeted from here on out. The goal of this refactoring is to separate out the massive `crates/cli-support/src/js/mod.rs` into a number of separate pieces of functionality. It currently takes care of basically everything including: * Binding intrinsics * Handling anyref transformations * Generating all JS for imports/exports * All the logic for how to import and how to name imports * Execution and management of wasm-bindgen closures Many of these are separable concerns and most overlap with WebIDL bindings. The internal refactoring here is intended to make it more clear who's responsible for what as well as making some existing operations much more straightforward. At a high-level, the following changes are done: 1. A `src/webidl.rs` module is introduced. The purpose of this module is to take all of the raw wasm-bindgen custom sections from the module and transform them into a WebIDL bindings section. This module has a placeholder `WebidlCustomSection` which is nowhere near the actual custom section but if you squint is in theory very similar. It's hoped that this will eventually become the true WebIDL custom section, currently being developed in an external crate. Currently, however, the WebIDL bindings custom section only covers a subset of the functionality we export to wasm-bindgen users. To avoid leaving them high and dry this module also contains an auxiliary custom section named `WasmBindgenAux`. This custom section isn't intended to have a binary format, but is intended to represent a theoretical custom section necessary to couple with WebIDL bindings to achieve all our desired functionality in `wasm-bindgen`. It'll never be standardized, but it'll also never be serialized :) 2. The `src/webidl.rs` module now takes over quite a bit of functionality from `src/js/mod.rs`. Namely it handles synthesis of an `export_map` and an `import_map` mapping export/import IDs to exactly what's expected to be hooked up there. This does not include type information (as that's in the bindings section) but rather includes things like "this is the method of class A" or "this import is from module `foo`" and things like that. These could arguably be subsumed by future JS features as well, but that's for another time! 3. All handling of wasm-bindgen "descriptor functions" now happens in a dedicated `src/descriptors.rs` module. The output of this module is its own custom section (intended to be immediately consumed by the WebIDL module) which is in theory what we want to ourselves emit one day but rustc isn't capable of doing so right now. 4. Invocations and generations of imports are completely overhauled. Using the `import_map` generated in the WebIDL step all imports are now handled much more precisely in one location rather than haphazardly throughout the module. This means we have precise information about each import of the module and we only modify exactly what we're looking at. This also vastly simplifies intrinsic generation since it's all simply a codegen part of the `rust2js.rs` module now. 5. Handling of direct imports which don't have a JS shim generated is slightly different from before and is intended to be future-compatible with WebIDL bindings in its full glory, but we'll need to update it to handle cases for constructors and method calls eventually as well. 6. Intrinsic definitions now live in their own file (`src/intrinsic.rs`) and have a separated definition for their symbol name and signature. The actual implementation of each intrinsic lives in `rust2js.rs` There's a number of TODO items to finish before this merges. This includes reimplementing the anyref pass and actually implementing import maps for other targets. Those will come soon in follow-up commits, but the entire `tests/wasm/main.rs` suite is currently passing and this seems like a good checkpoint.
2019-05-23 09:15:26 -07:00
/// This import is expected to be a shim that simply calls the `foo` method
/// on the first object, passing along all other parameters and returning
/// the resulting value.
StructuralMethod(String),
/// This import is a "structural getter" which simply returns the `.field`
/// value of the first argument as an object.
///
/// e.g. `function(x) { return x.foo; }`
StructuralGetter(String),
/// This import is a "structural getter" which simply returns the `.field`
/// value of the specified class
///
/// e.g. `function() { return TheClass.foo; }`
StructuralClassGetter(JsImport, String),
/// This import is a "structural setter" which simply sets the `.field`
/// value of the first argument to the second argument.
///
/// e.g. `function(x, y) { x.foo = y; }`
StructuralSetter(String),
/// This import is a "structural setter" which simply sets the `.field`
/// value of the specified class to the first argument of the function.
///
/// e.g. `function(x) { TheClass.foo = x; }`
StructuralClassSetter(JsImport, String),
/// This import is expected to be a shim that is an indexing getter of the
/// JS class here, where the first argument of the function is the field to
/// look up. The return value is the value of the field.
///
/// e.g. `function(x) { return TheClass[x]; }`
///
/// TODO: can we use `Reflect` or something like that to avoid an extra kind
/// of import here?
IndexingGetterOfClass(JsImport),
/// This import is expected to be a shim that is an indexing getter of the
/// first argument interpreted as an object where the field to look up is
/// the second argument.
///
/// e.g. `function(x, y) { return x[y]; }`
///
/// TODO: can we use `Reflect` or something like that to avoid an extra kind
/// of import here?
IndexingGetterOfObject,
/// This import is expected to be a shim that is an indexing setter of the
/// JS class here, where the first argument of the function is the field to
/// set and the second is the value to set it to.
///
/// e.g. `function(x, y) { TheClass[x] = y; }`
///
/// TODO: can we use `Reflect` or something like that to avoid an extra kind
/// of import here?
IndexingSetterOfClass(JsImport),
/// This import is expected to be a shim that is an indexing setter of the
/// first argument interpreted as an object where the next two fields are
/// the field to set and the value to set it to.
///
/// e.g. `function(x, y, z) { x[y] = z; }`
///
/// TODO: can we use `Reflect` or something like that to avoid an extra kind
/// of import here?
IndexingSetterOfObject,
/// This import is expected to be a shim that is an indexing deleter of the
/// JS class here, where the first argument of the function is the field to
/// delete.
///
/// e.g. `function(x) { delete TheClass[x]; }`
///
/// TODO: can we use `Reflect` or something like that to avoid an extra kind
/// of import here?
IndexingDeleterOfClass(JsImport),
/// This import is expected to be a shim that is an indexing deleter of the
/// first argument interpreted as an object where the second argument is
/// the field to delete.
///
/// e.g. `function(x, y) { delete x[y]; }`
///
/// TODO: can we use `Reflect` or something like that to avoid an extra kind
/// of import here?
IndexingDeleterOfObject,
/// This import is a generated shim which will wrap the provided pointer in
/// a JS object corresponding to the Class name given here. The class name
/// is one that is exported from the Rust/wasm.
///
/// TODO: sort of like the export map below we should ideally create the
/// `anyref` from within Rust itself and then return it directly rather than
/// requiring an intrinsic here to do so.
WrapInExportedClass(String),
/// This is an intrinsic function expected to be implemented with a JS glue
/// shim. Each intrinsic has its own expected signature and implementation.
Intrinsic(Intrinsic),
}
/// Values that can be imported verbatim to hook up to an import.
#[derive(Debug)]
pub enum AuxValue {
/// A bare JS value, no transformations, just put it in the slot.
Bare(JsImport),
/// A getter function for the class listed for the field, acquired using
/// `getOwnPropertyDescriptor`.
Getter(JsImport, String),
/// Like `Getter`, but accesses a field of a class instead of an instance
/// of the class.
ClassGetter(JsImport, String),
/// Like `Getter`, except the `set` property.
Setter(JsImport, String),
/// Like `Setter`, but for class fields instead of instance fields.
ClassSetter(JsImport, String),
}
/// What can actually be imported and typically a value in each of the variants
/// above of `AuxImport`
///
/// A `JsImport` is intended to indicate what exactly is being imported for a
/// particular operation.
#[derive(Debug, Hash, Eq, PartialEq, Clone)]
pub struct JsImport {
/// The base of whatever is being imported, either from a module, the global
/// namespace, or similar.
pub name: JsImportName,
/// Various field accesses (like `.foo.bar.baz`) to hang off the `name`
/// above.
pub fields: Vec<String>,
}
/// Return value of `determine_import` which is where we look at an imported
/// function AST and figure out where it's actually being imported from
/// (performing some validation checks and whatnot).
#[derive(Debug, Hash, Eq, PartialEq, Clone)]
pub enum JsImportName {
/// An item is imported from the global scope. The `name` is what's
/// imported.
Global { name: String },
/// Same as `Global`, except the `name` is imported via an ESM import from
/// the specified `module` path.
Module { module: String, name: String },
/// Same as `Module`, except we're importing from a local module defined in
/// a local JS snippet.
LocalModule { module: String, name: String },
/// Same as `Module`, except we're importing from an `inline_js` attribute
InlineJs {
unique_crate_identifier: String,
snippet_idx_in_crate: usize,
name: String,
},
/// A global import which may have a number of vendor prefixes associated
/// with it, like `webkitAudioPrefix`. The `name` is the name to test
/// whether it's prefixed.
VendorPrefixed { name: String, prefixes: Vec<String> },
}
struct Context<'a> {
start_found: bool,
module: &'a mut Module,
Second large refactor for WebIDL bindings This commit is the second, and hopefully last massive, refactor for using WebIDL bindings internally in `wasm-bindgen`. This commit actually fully executes on the task at hand, moving `wasm-bindgen` to internally using WebIDL bindings throughout its code generation, anyref passes, etc. This actually fixes a number of issues that have existed in the anyref pass for some time now! The main changes here are to basically remove the usage of `Descriptor` from generating JS bindings. Instead two new types are introduced: `NonstandardIncoming` and `NonstandardOutgoing` which are bindings lists used for incoming/outgoing bindings. These mirror the standard terminology and literally have variants which are the standard values. All `Descriptor` types are now mapped into lists of incoming/outgoing bindings and used for process in wasm-bindgen. All JS generation has been refactored and updated to now process these lists of bindings instead of the previous `Descriptor`. In other words this commit takes `js2rust.rs` and `rust2js.rs` and first splits them in two. Interpretation of `Descriptor` and what to do for conversions is in the binding selection modules. The actual generation of JS from the binding selection is now performed by `incoming.rs` and `outgoing.rs`. To boot this also deduplicates all the code between the argument handling of `js2rust.rs` and return value handling of `rust2js.rs`. This means that to implement a new binding you only need to implement it one place and it's implemented for free in the other! This commit is not the end of the story though. I would like to add a mdoe to `wasm-bindgen` that literally emits a WebIDL bindings section. That's left for a third (and hopefully final) refactoring which is also intended to optimize generated JS for bindings. This commit currently loses the optimization where an imported is hooked up by value directly whenever a shim isn't needed. It's planned that the next refactoring to emit a webidl binding section that can be added back in. It shouldn't be too too hard hopefully since all the scaffolding is in place now. cc #1524
2019-06-10 07:09:51 -07:00
bindings: NonstandardWebidlSection,
First refactor for WebIDL bindings This commit starts the `wasm-bindgen` CLI tool down the road to being a true polyfill for WebIDL bindings. This refactor is probably the first of a few, but is hopefully the largest and most sprawling and everything will be a bit more targeted from here on out. The goal of this refactoring is to separate out the massive `crates/cli-support/src/js/mod.rs` into a number of separate pieces of functionality. It currently takes care of basically everything including: * Binding intrinsics * Handling anyref transformations * Generating all JS for imports/exports * All the logic for how to import and how to name imports * Execution and management of wasm-bindgen closures Many of these are separable concerns and most overlap with WebIDL bindings. The internal refactoring here is intended to make it more clear who's responsible for what as well as making some existing operations much more straightforward. At a high-level, the following changes are done: 1. A `src/webidl.rs` module is introduced. The purpose of this module is to take all of the raw wasm-bindgen custom sections from the module and transform them into a WebIDL bindings section. This module has a placeholder `WebidlCustomSection` which is nowhere near the actual custom section but if you squint is in theory very similar. It's hoped that this will eventually become the true WebIDL custom section, currently being developed in an external crate. Currently, however, the WebIDL bindings custom section only covers a subset of the functionality we export to wasm-bindgen users. To avoid leaving them high and dry this module also contains an auxiliary custom section named `WasmBindgenAux`. This custom section isn't intended to have a binary format, but is intended to represent a theoretical custom section necessary to couple with WebIDL bindings to achieve all our desired functionality in `wasm-bindgen`. It'll never be standardized, but it'll also never be serialized :) 2. The `src/webidl.rs` module now takes over quite a bit of functionality from `src/js/mod.rs`. Namely it handles synthesis of an `export_map` and an `import_map` mapping export/import IDs to exactly what's expected to be hooked up there. This does not include type information (as that's in the bindings section) but rather includes things like "this is the method of class A" or "this import is from module `foo`" and things like that. These could arguably be subsumed by future JS features as well, but that's for another time! 3. All handling of wasm-bindgen "descriptor functions" now happens in a dedicated `src/descriptors.rs` module. The output of this module is its own custom section (intended to be immediately consumed by the WebIDL module) which is in theory what we want to ourselves emit one day but rustc isn't capable of doing so right now. 4. Invocations and generations of imports are completely overhauled. Using the `import_map` generated in the WebIDL step all imports are now handled much more precisely in one location rather than haphazardly throughout the module. This means we have precise information about each import of the module and we only modify exactly what we're looking at. This also vastly simplifies intrinsic generation since it's all simply a codegen part of the `rust2js.rs` module now. 5. Handling of direct imports which don't have a JS shim generated is slightly different from before and is intended to be future-compatible with WebIDL bindings in its full glory, but we'll need to update it to handle cases for constructors and method calls eventually as well. 6. Intrinsic definitions now live in their own file (`src/intrinsic.rs`) and have a separated definition for their symbol name and signature. The actual implementation of each intrinsic lives in `rust2js.rs` There's a number of TODO items to finish before this merges. This includes reimplementing the anyref pass and actually implementing import maps for other targets. Those will come soon in follow-up commits, but the entire `tests/wasm/main.rs` suite is currently passing and this seems like a good checkpoint.
2019-05-23 09:15:26 -07:00
aux: WasmBindgenAux,
function_exports: HashMap<String, (ExportId, FunctionId)>,
function_imports: HashMap<String, (ImportId, FunctionId)>,
vendor_prefixes: HashMap<String, Vec<String>>,
unique_crate_identifier: &'a str,
descriptors: HashMap<String, Descriptor>,
}
Second large refactor for WebIDL bindings This commit is the second, and hopefully last massive, refactor for using WebIDL bindings internally in `wasm-bindgen`. This commit actually fully executes on the task at hand, moving `wasm-bindgen` to internally using WebIDL bindings throughout its code generation, anyref passes, etc. This actually fixes a number of issues that have existed in the anyref pass for some time now! The main changes here are to basically remove the usage of `Descriptor` from generating JS bindings. Instead two new types are introduced: `NonstandardIncoming` and `NonstandardOutgoing` which are bindings lists used for incoming/outgoing bindings. These mirror the standard terminology and literally have variants which are the standard values. All `Descriptor` types are now mapped into lists of incoming/outgoing bindings and used for process in wasm-bindgen. All JS generation has been refactored and updated to now process these lists of bindings instead of the previous `Descriptor`. In other words this commit takes `js2rust.rs` and `rust2js.rs` and first splits them in two. Interpretation of `Descriptor` and what to do for conversions is in the binding selection modules. The actual generation of JS from the binding selection is now performed by `incoming.rs` and `outgoing.rs`. To boot this also deduplicates all the code between the argument handling of `js2rust.rs` and return value handling of `rust2js.rs`. This means that to implement a new binding you only need to implement it one place and it's implemented for free in the other! This commit is not the end of the story though. I would like to add a mdoe to `wasm-bindgen` that literally emits a WebIDL bindings section. That's left for a third (and hopefully final) refactoring which is also intended to optimize generated JS for bindings. This commit currently loses the optimization where an imported is hooked up by value directly whenever a shim isn't needed. It's planned that the next refactoring to emit a webidl binding section that can be added back in. It shouldn't be too too hard hopefully since all the scaffolding is in place now. cc #1524
2019-06-10 07:09:51 -07:00
pub fn process(
module: &mut Module,
) -> Result<(NonstandardWebidlSectionId, WasmBindgenAuxId), Error> {
First refactor for WebIDL bindings This commit starts the `wasm-bindgen` CLI tool down the road to being a true polyfill for WebIDL bindings. This refactor is probably the first of a few, but is hopefully the largest and most sprawling and everything will be a bit more targeted from here on out. The goal of this refactoring is to separate out the massive `crates/cli-support/src/js/mod.rs` into a number of separate pieces of functionality. It currently takes care of basically everything including: * Binding intrinsics * Handling anyref transformations * Generating all JS for imports/exports * All the logic for how to import and how to name imports * Execution and management of wasm-bindgen closures Many of these are separable concerns and most overlap with WebIDL bindings. The internal refactoring here is intended to make it more clear who's responsible for what as well as making some existing operations much more straightforward. At a high-level, the following changes are done: 1. A `src/webidl.rs` module is introduced. The purpose of this module is to take all of the raw wasm-bindgen custom sections from the module and transform them into a WebIDL bindings section. This module has a placeholder `WebidlCustomSection` which is nowhere near the actual custom section but if you squint is in theory very similar. It's hoped that this will eventually become the true WebIDL custom section, currently being developed in an external crate. Currently, however, the WebIDL bindings custom section only covers a subset of the functionality we export to wasm-bindgen users. To avoid leaving them high and dry this module also contains an auxiliary custom section named `WasmBindgenAux`. This custom section isn't intended to have a binary format, but is intended to represent a theoretical custom section necessary to couple with WebIDL bindings to achieve all our desired functionality in `wasm-bindgen`. It'll never be standardized, but it'll also never be serialized :) 2. The `src/webidl.rs` module now takes over quite a bit of functionality from `src/js/mod.rs`. Namely it handles synthesis of an `export_map` and an `import_map` mapping export/import IDs to exactly what's expected to be hooked up there. This does not include type information (as that's in the bindings section) but rather includes things like "this is the method of class A" or "this import is from module `foo`" and things like that. These could arguably be subsumed by future JS features as well, but that's for another time! 3. All handling of wasm-bindgen "descriptor functions" now happens in a dedicated `src/descriptors.rs` module. The output of this module is its own custom section (intended to be immediately consumed by the WebIDL module) which is in theory what we want to ourselves emit one day but rustc isn't capable of doing so right now. 4. Invocations and generations of imports are completely overhauled. Using the `import_map` generated in the WebIDL step all imports are now handled much more precisely in one location rather than haphazardly throughout the module. This means we have precise information about each import of the module and we only modify exactly what we're looking at. This also vastly simplifies intrinsic generation since it's all simply a codegen part of the `rust2js.rs` module now. 5. Handling of direct imports which don't have a JS shim generated is slightly different from before and is intended to be future-compatible with WebIDL bindings in its full glory, but we'll need to update it to handle cases for constructors and method calls eventually as well. 6. Intrinsic definitions now live in their own file (`src/intrinsic.rs`) and have a separated definition for their symbol name and signature. The actual implementation of each intrinsic lives in `rust2js.rs` There's a number of TODO items to finish before this merges. This includes reimplementing the anyref pass and actually implementing import maps for other targets. Those will come soon in follow-up commits, but the entire `tests/wasm/main.rs` suite is currently passing and this seems like a good checkpoint.
2019-05-23 09:15:26 -07:00
let mut storage = Vec::new();
let programs = extract_programs(module, &mut storage)?;
let mut cx = Context {
bindings: Default::default(),
aux: Default::default(),
function_exports: Default::default(),
function_imports: Default::default(),
vendor_prefixes: Default::default(),
descriptors: Default::default(),
unique_crate_identifier: "",
module,
start_found: false,
};
Second large refactor for WebIDL bindings This commit is the second, and hopefully last massive, refactor for using WebIDL bindings internally in `wasm-bindgen`. This commit actually fully executes on the task at hand, moving `wasm-bindgen` to internally using WebIDL bindings throughout its code generation, anyref passes, etc. This actually fixes a number of issues that have existed in the anyref pass for some time now! The main changes here are to basically remove the usage of `Descriptor` from generating JS bindings. Instead two new types are introduced: `NonstandardIncoming` and `NonstandardOutgoing` which are bindings lists used for incoming/outgoing bindings. These mirror the standard terminology and literally have variants which are the standard values. All `Descriptor` types are now mapped into lists of incoming/outgoing bindings and used for process in wasm-bindgen. All JS generation has been refactored and updated to now process these lists of bindings instead of the previous `Descriptor`. In other words this commit takes `js2rust.rs` and `rust2js.rs` and first splits them in two. Interpretation of `Descriptor` and what to do for conversions is in the binding selection modules. The actual generation of JS from the binding selection is now performed by `incoming.rs` and `outgoing.rs`. To boot this also deduplicates all the code between the argument handling of `js2rust.rs` and return value handling of `rust2js.rs`. This means that to implement a new binding you only need to implement it one place and it's implemented for free in the other! This commit is not the end of the story though. I would like to add a mdoe to `wasm-bindgen` that literally emits a WebIDL bindings section. That's left for a third (and hopefully final) refactoring which is also intended to optimize generated JS for bindings. This commit currently loses the optimization where an imported is hooked up by value directly whenever a shim isn't needed. It's planned that the next refactoring to emit a webidl binding section that can be added back in. It shouldn't be too too hard hopefully since all the scaffolding is in place now. cc #1524
2019-06-10 07:09:51 -07:00
cx.init()?;
First refactor for WebIDL bindings This commit starts the `wasm-bindgen` CLI tool down the road to being a true polyfill for WebIDL bindings. This refactor is probably the first of a few, but is hopefully the largest and most sprawling and everything will be a bit more targeted from here on out. The goal of this refactoring is to separate out the massive `crates/cli-support/src/js/mod.rs` into a number of separate pieces of functionality. It currently takes care of basically everything including: * Binding intrinsics * Handling anyref transformations * Generating all JS for imports/exports * All the logic for how to import and how to name imports * Execution and management of wasm-bindgen closures Many of these are separable concerns and most overlap with WebIDL bindings. The internal refactoring here is intended to make it more clear who's responsible for what as well as making some existing operations much more straightforward. At a high-level, the following changes are done: 1. A `src/webidl.rs` module is introduced. The purpose of this module is to take all of the raw wasm-bindgen custom sections from the module and transform them into a WebIDL bindings section. This module has a placeholder `WebidlCustomSection` which is nowhere near the actual custom section but if you squint is in theory very similar. It's hoped that this will eventually become the true WebIDL custom section, currently being developed in an external crate. Currently, however, the WebIDL bindings custom section only covers a subset of the functionality we export to wasm-bindgen users. To avoid leaving them high and dry this module also contains an auxiliary custom section named `WasmBindgenAux`. This custom section isn't intended to have a binary format, but is intended to represent a theoretical custom section necessary to couple with WebIDL bindings to achieve all our desired functionality in `wasm-bindgen`. It'll never be standardized, but it'll also never be serialized :) 2. The `src/webidl.rs` module now takes over quite a bit of functionality from `src/js/mod.rs`. Namely it handles synthesis of an `export_map` and an `import_map` mapping export/import IDs to exactly what's expected to be hooked up there. This does not include type information (as that's in the bindings section) but rather includes things like "this is the method of class A" or "this import is from module `foo`" and things like that. These could arguably be subsumed by future JS features as well, but that's for another time! 3. All handling of wasm-bindgen "descriptor functions" now happens in a dedicated `src/descriptors.rs` module. The output of this module is its own custom section (intended to be immediately consumed by the WebIDL module) which is in theory what we want to ourselves emit one day but rustc isn't capable of doing so right now. 4. Invocations and generations of imports are completely overhauled. Using the `import_map` generated in the WebIDL step all imports are now handled much more precisely in one location rather than haphazardly throughout the module. This means we have precise information about each import of the module and we only modify exactly what we're looking at. This also vastly simplifies intrinsic generation since it's all simply a codegen part of the `rust2js.rs` module now. 5. Handling of direct imports which don't have a JS shim generated is slightly different from before and is intended to be future-compatible with WebIDL bindings in its full glory, but we'll need to update it to handle cases for constructors and method calls eventually as well. 6. Intrinsic definitions now live in their own file (`src/intrinsic.rs`) and have a separated definition for their symbol name and signature. The actual implementation of each intrinsic lives in `rust2js.rs` There's a number of TODO items to finish before this merges. This includes reimplementing the anyref pass and actually implementing import maps for other targets. Those will come soon in follow-up commits, but the entire `tests/wasm/main.rs` suite is currently passing and this seems like a good checkpoint.
2019-05-23 09:15:26 -07:00
for program in programs {
cx.program(program)?;
}
cx.verify()?;
let bindings = cx.module.customs.add(cx.bindings);
let aux = cx.module.customs.add(cx.aux);
Ok((bindings, aux))
}
impl<'a> Context<'a> {
Second large refactor for WebIDL bindings This commit is the second, and hopefully last massive, refactor for using WebIDL bindings internally in `wasm-bindgen`. This commit actually fully executes on the task at hand, moving `wasm-bindgen` to internally using WebIDL bindings throughout its code generation, anyref passes, etc. This actually fixes a number of issues that have existed in the anyref pass for some time now! The main changes here are to basically remove the usage of `Descriptor` from generating JS bindings. Instead two new types are introduced: `NonstandardIncoming` and `NonstandardOutgoing` which are bindings lists used for incoming/outgoing bindings. These mirror the standard terminology and literally have variants which are the standard values. All `Descriptor` types are now mapped into lists of incoming/outgoing bindings and used for process in wasm-bindgen. All JS generation has been refactored and updated to now process these lists of bindings instead of the previous `Descriptor`. In other words this commit takes `js2rust.rs` and `rust2js.rs` and first splits them in two. Interpretation of `Descriptor` and what to do for conversions is in the binding selection modules. The actual generation of JS from the binding selection is now performed by `incoming.rs` and `outgoing.rs`. To boot this also deduplicates all the code between the argument handling of `js2rust.rs` and return value handling of `rust2js.rs`. This means that to implement a new binding you only need to implement it one place and it's implemented for free in the other! This commit is not the end of the story though. I would like to add a mdoe to `wasm-bindgen` that literally emits a WebIDL bindings section. That's left for a third (and hopefully final) refactoring which is also intended to optimize generated JS for bindings. This commit currently loses the optimization where an imported is hooked up by value directly whenever a shim isn't needed. It's planned that the next refactoring to emit a webidl binding section that can be added back in. It shouldn't be too too hard hopefully since all the scaffolding is in place now. cc #1524
2019-06-10 07:09:51 -07:00
fn init(&mut self) -> Result<(), Error> {
First refactor for WebIDL bindings This commit starts the `wasm-bindgen` CLI tool down the road to being a true polyfill for WebIDL bindings. This refactor is probably the first of a few, but is hopefully the largest and most sprawling and everything will be a bit more targeted from here on out. The goal of this refactoring is to separate out the massive `crates/cli-support/src/js/mod.rs` into a number of separate pieces of functionality. It currently takes care of basically everything including: * Binding intrinsics * Handling anyref transformations * Generating all JS for imports/exports * All the logic for how to import and how to name imports * Execution and management of wasm-bindgen closures Many of these are separable concerns and most overlap with WebIDL bindings. The internal refactoring here is intended to make it more clear who's responsible for what as well as making some existing operations much more straightforward. At a high-level, the following changes are done: 1. A `src/webidl.rs` module is introduced. The purpose of this module is to take all of the raw wasm-bindgen custom sections from the module and transform them into a WebIDL bindings section. This module has a placeholder `WebidlCustomSection` which is nowhere near the actual custom section but if you squint is in theory very similar. It's hoped that this will eventually become the true WebIDL custom section, currently being developed in an external crate. Currently, however, the WebIDL bindings custom section only covers a subset of the functionality we export to wasm-bindgen users. To avoid leaving them high and dry this module also contains an auxiliary custom section named `WasmBindgenAux`. This custom section isn't intended to have a binary format, but is intended to represent a theoretical custom section necessary to couple with WebIDL bindings to achieve all our desired functionality in `wasm-bindgen`. It'll never be standardized, but it'll also never be serialized :) 2. The `src/webidl.rs` module now takes over quite a bit of functionality from `src/js/mod.rs`. Namely it handles synthesis of an `export_map` and an `import_map` mapping export/import IDs to exactly what's expected to be hooked up there. This does not include type information (as that's in the bindings section) but rather includes things like "this is the method of class A" or "this import is from module `foo`" and things like that. These could arguably be subsumed by future JS features as well, but that's for another time! 3. All handling of wasm-bindgen "descriptor functions" now happens in a dedicated `src/descriptors.rs` module. The output of this module is its own custom section (intended to be immediately consumed by the WebIDL module) which is in theory what we want to ourselves emit one day but rustc isn't capable of doing so right now. 4. Invocations and generations of imports are completely overhauled. Using the `import_map` generated in the WebIDL step all imports are now handled much more precisely in one location rather than haphazardly throughout the module. This means we have precise information about each import of the module and we only modify exactly what we're looking at. This also vastly simplifies intrinsic generation since it's all simply a codegen part of the `rust2js.rs` module now. 5. Handling of direct imports which don't have a JS shim generated is slightly different from before and is intended to be future-compatible with WebIDL bindings in its full glory, but we'll need to update it to handle cases for constructors and method calls eventually as well. 6. Intrinsic definitions now live in their own file (`src/intrinsic.rs`) and have a separated definition for their symbol name and signature. The actual implementation of each intrinsic lives in `rust2js.rs` There's a number of TODO items to finish before this merges. This includes reimplementing the anyref pass and actually implementing import maps for other targets. Those will come soon in follow-up commits, but the entire `tests/wasm/main.rs` suite is currently passing and this seems like a good checkpoint.
2019-05-23 09:15:26 -07:00
// Make a map from string name to ids of all exports
for export in self.module.exports.iter() {
if let walrus::ExportItem::Function(f) = export.item {
self.function_exports
.insert(export.name.clone(), (export.id(), f));
}
}
// Make a map from string name to ids of all imports from our
// placeholder module name which we'll want to be sure that we've got a
// location listed of what to import there for each item.
Second large refactor for WebIDL bindings This commit is the second, and hopefully last massive, refactor for using WebIDL bindings internally in `wasm-bindgen`. This commit actually fully executes on the task at hand, moving `wasm-bindgen` to internally using WebIDL bindings throughout its code generation, anyref passes, etc. This actually fixes a number of issues that have existed in the anyref pass for some time now! The main changes here are to basically remove the usage of `Descriptor` from generating JS bindings. Instead two new types are introduced: `NonstandardIncoming` and `NonstandardOutgoing` which are bindings lists used for incoming/outgoing bindings. These mirror the standard terminology and literally have variants which are the standard values. All `Descriptor` types are now mapped into lists of incoming/outgoing bindings and used for process in wasm-bindgen. All JS generation has been refactored and updated to now process these lists of bindings instead of the previous `Descriptor`. In other words this commit takes `js2rust.rs` and `rust2js.rs` and first splits them in two. Interpretation of `Descriptor` and what to do for conversions is in the binding selection modules. The actual generation of JS from the binding selection is now performed by `incoming.rs` and `outgoing.rs`. To boot this also deduplicates all the code between the argument handling of `js2rust.rs` and return value handling of `rust2js.rs`. This means that to implement a new binding you only need to implement it one place and it's implemented for free in the other! This commit is not the end of the story though. I would like to add a mdoe to `wasm-bindgen` that literally emits a WebIDL bindings section. That's left for a third (and hopefully final) refactoring which is also intended to optimize generated JS for bindings. This commit currently loses the optimization where an imported is hooked up by value directly whenever a shim isn't needed. It's planned that the next refactoring to emit a webidl binding section that can be added back in. It shouldn't be too too hard hopefully since all the scaffolding is in place now. cc #1524
2019-06-10 07:09:51 -07:00
let mut intrinsics = Vec::new();
First refactor for WebIDL bindings This commit starts the `wasm-bindgen` CLI tool down the road to being a true polyfill for WebIDL bindings. This refactor is probably the first of a few, but is hopefully the largest and most sprawling and everything will be a bit more targeted from here on out. The goal of this refactoring is to separate out the massive `crates/cli-support/src/js/mod.rs` into a number of separate pieces of functionality. It currently takes care of basically everything including: * Binding intrinsics * Handling anyref transformations * Generating all JS for imports/exports * All the logic for how to import and how to name imports * Execution and management of wasm-bindgen closures Many of these are separable concerns and most overlap with WebIDL bindings. The internal refactoring here is intended to make it more clear who's responsible for what as well as making some existing operations much more straightforward. At a high-level, the following changes are done: 1. A `src/webidl.rs` module is introduced. The purpose of this module is to take all of the raw wasm-bindgen custom sections from the module and transform them into a WebIDL bindings section. This module has a placeholder `WebidlCustomSection` which is nowhere near the actual custom section but if you squint is in theory very similar. It's hoped that this will eventually become the true WebIDL custom section, currently being developed in an external crate. Currently, however, the WebIDL bindings custom section only covers a subset of the functionality we export to wasm-bindgen users. To avoid leaving them high and dry this module also contains an auxiliary custom section named `WasmBindgenAux`. This custom section isn't intended to have a binary format, but is intended to represent a theoretical custom section necessary to couple with WebIDL bindings to achieve all our desired functionality in `wasm-bindgen`. It'll never be standardized, but it'll also never be serialized :) 2. The `src/webidl.rs` module now takes over quite a bit of functionality from `src/js/mod.rs`. Namely it handles synthesis of an `export_map` and an `import_map` mapping export/import IDs to exactly what's expected to be hooked up there. This does not include type information (as that's in the bindings section) but rather includes things like "this is the method of class A" or "this import is from module `foo`" and things like that. These could arguably be subsumed by future JS features as well, but that's for another time! 3. All handling of wasm-bindgen "descriptor functions" now happens in a dedicated `src/descriptors.rs` module. The output of this module is its own custom section (intended to be immediately consumed by the WebIDL module) which is in theory what we want to ourselves emit one day but rustc isn't capable of doing so right now. 4. Invocations and generations of imports are completely overhauled. Using the `import_map` generated in the WebIDL step all imports are now handled much more precisely in one location rather than haphazardly throughout the module. This means we have precise information about each import of the module and we only modify exactly what we're looking at. This also vastly simplifies intrinsic generation since it's all simply a codegen part of the `rust2js.rs` module now. 5. Handling of direct imports which don't have a JS shim generated is slightly different from before and is intended to be future-compatible with WebIDL bindings in its full glory, but we'll need to update it to handle cases for constructors and method calls eventually as well. 6. Intrinsic definitions now live in their own file (`src/intrinsic.rs`) and have a separated definition for their symbol name and signature. The actual implementation of each intrinsic lives in `rust2js.rs` There's a number of TODO items to finish before this merges. This includes reimplementing the anyref pass and actually implementing import maps for other targets. Those will come soon in follow-up commits, but the entire `tests/wasm/main.rs` suite is currently passing and this seems like a good checkpoint.
2019-05-23 09:15:26 -07:00
for import in self.module.imports.iter() {
if import.module != PLACEHOLDER_MODULE {
continue;
}
if let walrus::ImportKind::Function(f) = import.kind {
self.function_imports
.insert(import.name.clone(), (import.id(), f));
if let Some(intrinsic) = Intrinsic::from_symbol(&import.name) {
Second large refactor for WebIDL bindings This commit is the second, and hopefully last massive, refactor for using WebIDL bindings internally in `wasm-bindgen`. This commit actually fully executes on the task at hand, moving `wasm-bindgen` to internally using WebIDL bindings throughout its code generation, anyref passes, etc. This actually fixes a number of issues that have existed in the anyref pass for some time now! The main changes here are to basically remove the usage of `Descriptor` from generating JS bindings. Instead two new types are introduced: `NonstandardIncoming` and `NonstandardOutgoing` which are bindings lists used for incoming/outgoing bindings. These mirror the standard terminology and literally have variants which are the standard values. All `Descriptor` types are now mapped into lists of incoming/outgoing bindings and used for process in wasm-bindgen. All JS generation has been refactored and updated to now process these lists of bindings instead of the previous `Descriptor`. In other words this commit takes `js2rust.rs` and `rust2js.rs` and first splits them in two. Interpretation of `Descriptor` and what to do for conversions is in the binding selection modules. The actual generation of JS from the binding selection is now performed by `incoming.rs` and `outgoing.rs`. To boot this also deduplicates all the code between the argument handling of `js2rust.rs` and return value handling of `rust2js.rs`. This means that to implement a new binding you only need to implement it one place and it's implemented for free in the other! This commit is not the end of the story though. I would like to add a mdoe to `wasm-bindgen` that literally emits a WebIDL bindings section. That's left for a third (and hopefully final) refactoring which is also intended to optimize generated JS for bindings. This commit currently loses the optimization where an imported is hooked up by value directly whenever a shim isn't needed. It's planned that the next refactoring to emit a webidl binding section that can be added back in. It shouldn't be too too hard hopefully since all the scaffolding is in place now. cc #1524
2019-06-10 07:09:51 -07:00
intrinsics.push((import.id(), intrinsic));
First refactor for WebIDL bindings This commit starts the `wasm-bindgen` CLI tool down the road to being a true polyfill for WebIDL bindings. This refactor is probably the first of a few, but is hopefully the largest and most sprawling and everything will be a bit more targeted from here on out. The goal of this refactoring is to separate out the massive `crates/cli-support/src/js/mod.rs` into a number of separate pieces of functionality. It currently takes care of basically everything including: * Binding intrinsics * Handling anyref transformations * Generating all JS for imports/exports * All the logic for how to import and how to name imports * Execution and management of wasm-bindgen closures Many of these are separable concerns and most overlap with WebIDL bindings. The internal refactoring here is intended to make it more clear who's responsible for what as well as making some existing operations much more straightforward. At a high-level, the following changes are done: 1. A `src/webidl.rs` module is introduced. The purpose of this module is to take all of the raw wasm-bindgen custom sections from the module and transform them into a WebIDL bindings section. This module has a placeholder `WebidlCustomSection` which is nowhere near the actual custom section but if you squint is in theory very similar. It's hoped that this will eventually become the true WebIDL custom section, currently being developed in an external crate. Currently, however, the WebIDL bindings custom section only covers a subset of the functionality we export to wasm-bindgen users. To avoid leaving them high and dry this module also contains an auxiliary custom section named `WasmBindgenAux`. This custom section isn't intended to have a binary format, but is intended to represent a theoretical custom section necessary to couple with WebIDL bindings to achieve all our desired functionality in `wasm-bindgen`. It'll never be standardized, but it'll also never be serialized :) 2. The `src/webidl.rs` module now takes over quite a bit of functionality from `src/js/mod.rs`. Namely it handles synthesis of an `export_map` and an `import_map` mapping export/import IDs to exactly what's expected to be hooked up there. This does not include type information (as that's in the bindings section) but rather includes things like "this is the method of class A" or "this import is from module `foo`" and things like that. These could arguably be subsumed by future JS features as well, but that's for another time! 3. All handling of wasm-bindgen "descriptor functions" now happens in a dedicated `src/descriptors.rs` module. The output of this module is its own custom section (intended to be immediately consumed by the WebIDL module) which is in theory what we want to ourselves emit one day but rustc isn't capable of doing so right now. 4. Invocations and generations of imports are completely overhauled. Using the `import_map` generated in the WebIDL step all imports are now handled much more precisely in one location rather than haphazardly throughout the module. This means we have precise information about each import of the module and we only modify exactly what we're looking at. This also vastly simplifies intrinsic generation since it's all simply a codegen part of the `rust2js.rs` module now. 5. Handling of direct imports which don't have a JS shim generated is slightly different from before and is intended to be future-compatible with WebIDL bindings in its full glory, but we'll need to update it to handle cases for constructors and method calls eventually as well. 6. Intrinsic definitions now live in their own file (`src/intrinsic.rs`) and have a separated definition for their symbol name and signature. The actual implementation of each intrinsic lives in `rust2js.rs` There's a number of TODO items to finish before this merges. This includes reimplementing the anyref pass and actually implementing import maps for other targets. Those will come soon in follow-up commits, but the entire `tests/wasm/main.rs` suite is currently passing and this seems like a good checkpoint.
2019-05-23 09:15:26 -07:00
}
}
}
Second large refactor for WebIDL bindings This commit is the second, and hopefully last massive, refactor for using WebIDL bindings internally in `wasm-bindgen`. This commit actually fully executes on the task at hand, moving `wasm-bindgen` to internally using WebIDL bindings throughout its code generation, anyref passes, etc. This actually fixes a number of issues that have existed in the anyref pass for some time now! The main changes here are to basically remove the usage of `Descriptor` from generating JS bindings. Instead two new types are introduced: `NonstandardIncoming` and `NonstandardOutgoing` which are bindings lists used for incoming/outgoing bindings. These mirror the standard terminology and literally have variants which are the standard values. All `Descriptor` types are now mapped into lists of incoming/outgoing bindings and used for process in wasm-bindgen. All JS generation has been refactored and updated to now process these lists of bindings instead of the previous `Descriptor`. In other words this commit takes `js2rust.rs` and `rust2js.rs` and first splits them in two. Interpretation of `Descriptor` and what to do for conversions is in the binding selection modules. The actual generation of JS from the binding selection is now performed by `incoming.rs` and `outgoing.rs`. To boot this also deduplicates all the code between the argument handling of `js2rust.rs` and return value handling of `rust2js.rs`. This means that to implement a new binding you only need to implement it one place and it's implemented for free in the other! This commit is not the end of the story though. I would like to add a mdoe to `wasm-bindgen` that literally emits a WebIDL bindings section. That's left for a third (and hopefully final) refactoring which is also intended to optimize generated JS for bindings. This commit currently loses the optimization where an imported is hooked up by value directly whenever a shim isn't needed. It's planned that the next refactoring to emit a webidl binding section that can be added back in. It shouldn't be too too hard hopefully since all the scaffolding is in place now. cc #1524
2019-06-10 07:09:51 -07:00
for (id, intrinsic) in intrinsics {
bindings::register_import(
self.module,
&mut self.bindings,
id,
intrinsic.binding(),
ast::WebidlFunctionKind::Static,
)?;
self.aux
.import_map
.insert(id, AuxImport::Intrinsic(intrinsic));
}
First refactor for WebIDL bindings This commit starts the `wasm-bindgen` CLI tool down the road to being a true polyfill for WebIDL bindings. This refactor is probably the first of a few, but is hopefully the largest and most sprawling and everything will be a bit more targeted from here on out. The goal of this refactoring is to separate out the massive `crates/cli-support/src/js/mod.rs` into a number of separate pieces of functionality. It currently takes care of basically everything including: * Binding intrinsics * Handling anyref transformations * Generating all JS for imports/exports * All the logic for how to import and how to name imports * Execution and management of wasm-bindgen closures Many of these are separable concerns and most overlap with WebIDL bindings. The internal refactoring here is intended to make it more clear who's responsible for what as well as making some existing operations much more straightforward. At a high-level, the following changes are done: 1. A `src/webidl.rs` module is introduced. The purpose of this module is to take all of the raw wasm-bindgen custom sections from the module and transform them into a WebIDL bindings section. This module has a placeholder `WebidlCustomSection` which is nowhere near the actual custom section but if you squint is in theory very similar. It's hoped that this will eventually become the true WebIDL custom section, currently being developed in an external crate. Currently, however, the WebIDL bindings custom section only covers a subset of the functionality we export to wasm-bindgen users. To avoid leaving them high and dry this module also contains an auxiliary custom section named `WasmBindgenAux`. This custom section isn't intended to have a binary format, but is intended to represent a theoretical custom section necessary to couple with WebIDL bindings to achieve all our desired functionality in `wasm-bindgen`. It'll never be standardized, but it'll also never be serialized :) 2. The `src/webidl.rs` module now takes over quite a bit of functionality from `src/js/mod.rs`. Namely it handles synthesis of an `export_map` and an `import_map` mapping export/import IDs to exactly what's expected to be hooked up there. This does not include type information (as that's in the bindings section) but rather includes things like "this is the method of class A" or "this import is from module `foo`" and things like that. These could arguably be subsumed by future JS features as well, but that's for another time! 3. All handling of wasm-bindgen "descriptor functions" now happens in a dedicated `src/descriptors.rs` module. The output of this module is its own custom section (intended to be immediately consumed by the WebIDL module) which is in theory what we want to ourselves emit one day but rustc isn't capable of doing so right now. 4. Invocations and generations of imports are completely overhauled. Using the `import_map` generated in the WebIDL step all imports are now handled much more precisely in one location rather than haphazardly throughout the module. This means we have precise information about each import of the module and we only modify exactly what we're looking at. This also vastly simplifies intrinsic generation since it's all simply a codegen part of the `rust2js.rs` module now. 5. Handling of direct imports which don't have a JS shim generated is slightly different from before and is intended to be future-compatible with WebIDL bindings in its full glory, but we'll need to update it to handle cases for constructors and method calls eventually as well. 6. Intrinsic definitions now live in their own file (`src/intrinsic.rs`) and have a separated definition for their symbol name and signature. The actual implementation of each intrinsic lives in `rust2js.rs` There's a number of TODO items to finish before this merges. This includes reimplementing the anyref pass and actually implementing import maps for other targets. Those will come soon in follow-up commits, but the entire `tests/wasm/main.rs` suite is currently passing and this seems like a good checkpoint.
2019-05-23 09:15:26 -07:00
if let Some(custom) = self
.module
.customs
.delete_typed::<WasmBindgenDescriptorsSection>()
{
let WasmBindgenDescriptorsSection {
descriptors,
closure_imports,
} = *custom;
First refactor for WebIDL bindings This commit starts the `wasm-bindgen` CLI tool down the road to being a true polyfill for WebIDL bindings. This refactor is probably the first of a few, but is hopefully the largest and most sprawling and everything will be a bit more targeted from here on out. The goal of this refactoring is to separate out the massive `crates/cli-support/src/js/mod.rs` into a number of separate pieces of functionality. It currently takes care of basically everything including: * Binding intrinsics * Handling anyref transformations * Generating all JS for imports/exports * All the logic for how to import and how to name imports * Execution and management of wasm-bindgen closures Many of these are separable concerns and most overlap with WebIDL bindings. The internal refactoring here is intended to make it more clear who's responsible for what as well as making some existing operations much more straightforward. At a high-level, the following changes are done: 1. A `src/webidl.rs` module is introduced. The purpose of this module is to take all of the raw wasm-bindgen custom sections from the module and transform them into a WebIDL bindings section. This module has a placeholder `WebidlCustomSection` which is nowhere near the actual custom section but if you squint is in theory very similar. It's hoped that this will eventually become the true WebIDL custom section, currently being developed in an external crate. Currently, however, the WebIDL bindings custom section only covers a subset of the functionality we export to wasm-bindgen users. To avoid leaving them high and dry this module also contains an auxiliary custom section named `WasmBindgenAux`. This custom section isn't intended to have a binary format, but is intended to represent a theoretical custom section necessary to couple with WebIDL bindings to achieve all our desired functionality in `wasm-bindgen`. It'll never be standardized, but it'll also never be serialized :) 2. The `src/webidl.rs` module now takes over quite a bit of functionality from `src/js/mod.rs`. Namely it handles synthesis of an `export_map` and an `import_map` mapping export/import IDs to exactly what's expected to be hooked up there. This does not include type information (as that's in the bindings section) but rather includes things like "this is the method of class A" or "this import is from module `foo`" and things like that. These could arguably be subsumed by future JS features as well, but that's for another time! 3. All handling of wasm-bindgen "descriptor functions" now happens in a dedicated `src/descriptors.rs` module. The output of this module is its own custom section (intended to be immediately consumed by the WebIDL module) which is in theory what we want to ourselves emit one day but rustc isn't capable of doing so right now. 4. Invocations and generations of imports are completely overhauled. Using the `import_map` generated in the WebIDL step all imports are now handled much more precisely in one location rather than haphazardly throughout the module. This means we have precise information about each import of the module and we only modify exactly what we're looking at. This also vastly simplifies intrinsic generation since it's all simply a codegen part of the `rust2js.rs` module now. 5. Handling of direct imports which don't have a JS shim generated is slightly different from before and is intended to be future-compatible with WebIDL bindings in its full glory, but we'll need to update it to handle cases for constructors and method calls eventually as well. 6. Intrinsic definitions now live in their own file (`src/intrinsic.rs`) and have a separated definition for their symbol name and signature. The actual implementation of each intrinsic lives in `rust2js.rs` There's a number of TODO items to finish before this merges. This includes reimplementing the anyref pass and actually implementing import maps for other targets. Those will come soon in follow-up commits, but the entire `tests/wasm/main.rs` suite is currently passing and this seems like a good checkpoint.
2019-05-23 09:15:26 -07:00
// Store all the executed descriptors in our own field so we have
// access to them while processing programs.
self.descriptors.extend(descriptors);
First refactor for WebIDL bindings This commit starts the `wasm-bindgen` CLI tool down the road to being a true polyfill for WebIDL bindings. This refactor is probably the first of a few, but is hopefully the largest and most sprawling and everything will be a bit more targeted from here on out. The goal of this refactoring is to separate out the massive `crates/cli-support/src/js/mod.rs` into a number of separate pieces of functionality. It currently takes care of basically everything including: * Binding intrinsics * Handling anyref transformations * Generating all JS for imports/exports * All the logic for how to import and how to name imports * Execution and management of wasm-bindgen closures Many of these are separable concerns and most overlap with WebIDL bindings. The internal refactoring here is intended to make it more clear who's responsible for what as well as making some existing operations much more straightforward. At a high-level, the following changes are done: 1. A `src/webidl.rs` module is introduced. The purpose of this module is to take all of the raw wasm-bindgen custom sections from the module and transform them into a WebIDL bindings section. This module has a placeholder `WebidlCustomSection` which is nowhere near the actual custom section but if you squint is in theory very similar. It's hoped that this will eventually become the true WebIDL custom section, currently being developed in an external crate. Currently, however, the WebIDL bindings custom section only covers a subset of the functionality we export to wasm-bindgen users. To avoid leaving them high and dry this module also contains an auxiliary custom section named `WasmBindgenAux`. This custom section isn't intended to have a binary format, but is intended to represent a theoretical custom section necessary to couple with WebIDL bindings to achieve all our desired functionality in `wasm-bindgen`. It'll never be standardized, but it'll also never be serialized :) 2. The `src/webidl.rs` module now takes over quite a bit of functionality from `src/js/mod.rs`. Namely it handles synthesis of an `export_map` and an `import_map` mapping export/import IDs to exactly what's expected to be hooked up there. This does not include type information (as that's in the bindings section) but rather includes things like "this is the method of class A" or "this import is from module `foo`" and things like that. These could arguably be subsumed by future JS features as well, but that's for another time! 3. All handling of wasm-bindgen "descriptor functions" now happens in a dedicated `src/descriptors.rs` module. The output of this module is its own custom section (intended to be immediately consumed by the WebIDL module) which is in theory what we want to ourselves emit one day but rustc isn't capable of doing so right now. 4. Invocations and generations of imports are completely overhauled. Using the `import_map` generated in the WebIDL step all imports are now handled much more precisely in one location rather than haphazardly throughout the module. This means we have precise information about each import of the module and we only modify exactly what we're looking at. This also vastly simplifies intrinsic generation since it's all simply a codegen part of the `rust2js.rs` module now. 5. Handling of direct imports which don't have a JS shim generated is slightly different from before and is intended to be future-compatible with WebIDL bindings in its full glory, but we'll need to update it to handle cases for constructors and method calls eventually as well. 6. Intrinsic definitions now live in their own file (`src/intrinsic.rs`) and have a separated definition for their symbol name and signature. The actual implementation of each intrinsic lives in `rust2js.rs` There's a number of TODO items to finish before this merges. This includes reimplementing the anyref pass and actually implementing import maps for other targets. Those will come soon in follow-up commits, but the entire `tests/wasm/main.rs` suite is currently passing and this seems like a good checkpoint.
2019-05-23 09:15:26 -07:00
// Register all the injected closure imports as that they're expected
// to manufacture a particular type of closure.
Second large refactor for WebIDL bindings This commit is the second, and hopefully last massive, refactor for using WebIDL bindings internally in `wasm-bindgen`. This commit actually fully executes on the task at hand, moving `wasm-bindgen` to internally using WebIDL bindings throughout its code generation, anyref passes, etc. This actually fixes a number of issues that have existed in the anyref pass for some time now! The main changes here are to basically remove the usage of `Descriptor` from generating JS bindings. Instead two new types are introduced: `NonstandardIncoming` and `NonstandardOutgoing` which are bindings lists used for incoming/outgoing bindings. These mirror the standard terminology and literally have variants which are the standard values. All `Descriptor` types are now mapped into lists of incoming/outgoing bindings and used for process in wasm-bindgen. All JS generation has been refactored and updated to now process these lists of bindings instead of the previous `Descriptor`. In other words this commit takes `js2rust.rs` and `rust2js.rs` and first splits them in two. Interpretation of `Descriptor` and what to do for conversions is in the binding selection modules. The actual generation of JS from the binding selection is now performed by `incoming.rs` and `outgoing.rs`. To boot this also deduplicates all the code between the argument handling of `js2rust.rs` and return value handling of `rust2js.rs`. This means that to implement a new binding you only need to implement it one place and it's implemented for free in the other! This commit is not the end of the story though. I would like to add a mdoe to `wasm-bindgen` that literally emits a WebIDL bindings section. That's left for a third (and hopefully final) refactoring which is also intended to optimize generated JS for bindings. This commit currently loses the optimization where an imported is hooked up by value directly whenever a shim isn't needed. It's planned that the next refactoring to emit a webidl binding section that can be added back in. It shouldn't be too too hard hopefully since all the scaffolding is in place now. cc #1524
2019-06-10 07:09:51 -07:00
//
// First we register the imported function shim which returns a
// `JsValue` for the closure. We manufacture this signature's
// binding since it's not listed anywhere.
//
// Next we register the corresponding table element's binding in
// the webidl bindings section. This binding will later be used to
// generate a shim (if necessary) for the table element.
//
// Finally we store all this metadata in the import map which we've
// learned so when a binding for the import is generated we can
// generate all the appropriate shims.
for (id, descriptor) in closure_imports {
First refactor for WebIDL bindings This commit starts the `wasm-bindgen` CLI tool down the road to being a true polyfill for WebIDL bindings. This refactor is probably the first of a few, but is hopefully the largest and most sprawling and everything will be a bit more targeted from here on out. The goal of this refactoring is to separate out the massive `crates/cli-support/src/js/mod.rs` into a number of separate pieces of functionality. It currently takes care of basically everything including: * Binding intrinsics * Handling anyref transformations * Generating all JS for imports/exports * All the logic for how to import and how to name imports * Execution and management of wasm-bindgen closures Many of these are separable concerns and most overlap with WebIDL bindings. The internal refactoring here is intended to make it more clear who's responsible for what as well as making some existing operations much more straightforward. At a high-level, the following changes are done: 1. A `src/webidl.rs` module is introduced. The purpose of this module is to take all of the raw wasm-bindgen custom sections from the module and transform them into a WebIDL bindings section. This module has a placeholder `WebidlCustomSection` which is nowhere near the actual custom section but if you squint is in theory very similar. It's hoped that this will eventually become the true WebIDL custom section, currently being developed in an external crate. Currently, however, the WebIDL bindings custom section only covers a subset of the functionality we export to wasm-bindgen users. To avoid leaving them high and dry this module also contains an auxiliary custom section named `WasmBindgenAux`. This custom section isn't intended to have a binary format, but is intended to represent a theoretical custom section necessary to couple with WebIDL bindings to achieve all our desired functionality in `wasm-bindgen`. It'll never be standardized, but it'll also never be serialized :) 2. The `src/webidl.rs` module now takes over quite a bit of functionality from `src/js/mod.rs`. Namely it handles synthesis of an `export_map` and an `import_map` mapping export/import IDs to exactly what's expected to be hooked up there. This does not include type information (as that's in the bindings section) but rather includes things like "this is the method of class A" or "this import is from module `foo`" and things like that. These could arguably be subsumed by future JS features as well, but that's for another time! 3. All handling of wasm-bindgen "descriptor functions" now happens in a dedicated `src/descriptors.rs` module. The output of this module is its own custom section (intended to be immediately consumed by the WebIDL module) which is in theory what we want to ourselves emit one day but rustc isn't capable of doing so right now. 4. Invocations and generations of imports are completely overhauled. Using the `import_map` generated in the WebIDL step all imports are now handled much more precisely in one location rather than haphazardly throughout the module. This means we have precise information about each import of the module and we only modify exactly what we're looking at. This also vastly simplifies intrinsic generation since it's all simply a codegen part of the `rust2js.rs` module now. 5. Handling of direct imports which don't have a JS shim generated is slightly different from before and is intended to be future-compatible with WebIDL bindings in its full glory, but we'll need to update it to handle cases for constructors and method calls eventually as well. 6. Intrinsic definitions now live in their own file (`src/intrinsic.rs`) and have a separated definition for their symbol name and signature. The actual implementation of each intrinsic lives in `rust2js.rs` There's a number of TODO items to finish before this merges. This includes reimplementing the anyref pass and actually implementing import maps for other targets. Those will come soon in follow-up commits, but the entire `tests/wasm/main.rs` suite is currently passing and this seems like a good checkpoint.
2019-05-23 09:15:26 -07:00
let binding = Function {
shim_idx: 0,
arguments: vec![Descriptor::I32; 3],
ret: Descriptor::Anyref,
};
Second large refactor for WebIDL bindings This commit is the second, and hopefully last massive, refactor for using WebIDL bindings internally in `wasm-bindgen`. This commit actually fully executes on the task at hand, moving `wasm-bindgen` to internally using WebIDL bindings throughout its code generation, anyref passes, etc. This actually fixes a number of issues that have existed in the anyref pass for some time now! The main changes here are to basically remove the usage of `Descriptor` from generating JS bindings. Instead two new types are introduced: `NonstandardIncoming` and `NonstandardOutgoing` which are bindings lists used for incoming/outgoing bindings. These mirror the standard terminology and literally have variants which are the standard values. All `Descriptor` types are now mapped into lists of incoming/outgoing bindings and used for process in wasm-bindgen. All JS generation has been refactored and updated to now process these lists of bindings instead of the previous `Descriptor`. In other words this commit takes `js2rust.rs` and `rust2js.rs` and first splits them in two. Interpretation of `Descriptor` and what to do for conversions is in the binding selection modules. The actual generation of JS from the binding selection is now performed by `incoming.rs` and `outgoing.rs`. To boot this also deduplicates all the code between the argument handling of `js2rust.rs` and return value handling of `rust2js.rs`. This means that to implement a new binding you only need to implement it one place and it's implemented for free in the other! This commit is not the end of the story though. I would like to add a mdoe to `wasm-bindgen` that literally emits a WebIDL bindings section. That's left for a third (and hopefully final) refactoring which is also intended to optimize generated JS for bindings. This commit currently loses the optimization where an imported is hooked up by value directly whenever a shim isn't needed. It's planned that the next refactoring to emit a webidl binding section that can be added back in. It shouldn't be too too hard hopefully since all the scaffolding is in place now. cc #1524
2019-06-10 07:09:51 -07:00
bindings::register_import(
self.module,
&mut self.bindings,
id,
binding,
ast::WebidlFunctionKind::Static,
)?;
// Synthesize the two integer pointers we pass through which
// aren't present in the signature but are present in the wasm
// signature.
let mut function = descriptor.function.clone();
let nargs = function.arguments.len();
function.arguments.insert(0, Descriptor::I32);
function.arguments.insert(0, Descriptor::I32);
let binding_idx = bindings::register_table_element(
self.module,
&mut self.bindings,
descriptor.shim_idx,
function,
)?;
self.aux.import_map.insert(
id,
AuxImport::Closure {
dtor: descriptor.dtor_idx,
mutable: descriptor.mutable,
binding_idx,
nargs,
},
);
First refactor for WebIDL bindings This commit starts the `wasm-bindgen` CLI tool down the road to being a true polyfill for WebIDL bindings. This refactor is probably the first of a few, but is hopefully the largest and most sprawling and everything will be a bit more targeted from here on out. The goal of this refactoring is to separate out the massive `crates/cli-support/src/js/mod.rs` into a number of separate pieces of functionality. It currently takes care of basically everything including: * Binding intrinsics * Handling anyref transformations * Generating all JS for imports/exports * All the logic for how to import and how to name imports * Execution and management of wasm-bindgen closures Many of these are separable concerns and most overlap with WebIDL bindings. The internal refactoring here is intended to make it more clear who's responsible for what as well as making some existing operations much more straightforward. At a high-level, the following changes are done: 1. A `src/webidl.rs` module is introduced. The purpose of this module is to take all of the raw wasm-bindgen custom sections from the module and transform them into a WebIDL bindings section. This module has a placeholder `WebidlCustomSection` which is nowhere near the actual custom section but if you squint is in theory very similar. It's hoped that this will eventually become the true WebIDL custom section, currently being developed in an external crate. Currently, however, the WebIDL bindings custom section only covers a subset of the functionality we export to wasm-bindgen users. To avoid leaving them high and dry this module also contains an auxiliary custom section named `WasmBindgenAux`. This custom section isn't intended to have a binary format, but is intended to represent a theoretical custom section necessary to couple with WebIDL bindings to achieve all our desired functionality in `wasm-bindgen`. It'll never be standardized, but it'll also never be serialized :) 2. The `src/webidl.rs` module now takes over quite a bit of functionality from `src/js/mod.rs`. Namely it handles synthesis of an `export_map` and an `import_map` mapping export/import IDs to exactly what's expected to be hooked up there. This does not include type information (as that's in the bindings section) but rather includes things like "this is the method of class A" or "this import is from module `foo`" and things like that. These could arguably be subsumed by future JS features as well, but that's for another time! 3. All handling of wasm-bindgen "descriptor functions" now happens in a dedicated `src/descriptors.rs` module. The output of this module is its own custom section (intended to be immediately consumed by the WebIDL module) which is in theory what we want to ourselves emit one day but rustc isn't capable of doing so right now. 4. Invocations and generations of imports are completely overhauled. Using the `import_map` generated in the WebIDL step all imports are now handled much more precisely in one location rather than haphazardly throughout the module. This means we have precise information about each import of the module and we only modify exactly what we're looking at. This also vastly simplifies intrinsic generation since it's all simply a codegen part of the `rust2js.rs` module now. 5. Handling of direct imports which don't have a JS shim generated is slightly different from before and is intended to be future-compatible with WebIDL bindings in its full glory, but we'll need to update it to handle cases for constructors and method calls eventually as well. 6. Intrinsic definitions now live in their own file (`src/intrinsic.rs`) and have a separated definition for their symbol name and signature. The actual implementation of each intrinsic lives in `rust2js.rs` There's a number of TODO items to finish before this merges. This includes reimplementing the anyref pass and actually implementing import maps for other targets. Those will come soon in follow-up commits, but the entire `tests/wasm/main.rs` suite is currently passing and this seems like a good checkpoint.
2019-05-23 09:15:26 -07:00
}
}
Second large refactor for WebIDL bindings This commit is the second, and hopefully last massive, refactor for using WebIDL bindings internally in `wasm-bindgen`. This commit actually fully executes on the task at hand, moving `wasm-bindgen` to internally using WebIDL bindings throughout its code generation, anyref passes, etc. This actually fixes a number of issues that have existed in the anyref pass for some time now! The main changes here are to basically remove the usage of `Descriptor` from generating JS bindings. Instead two new types are introduced: `NonstandardIncoming` and `NonstandardOutgoing` which are bindings lists used for incoming/outgoing bindings. These mirror the standard terminology and literally have variants which are the standard values. All `Descriptor` types are now mapped into lists of incoming/outgoing bindings and used for process in wasm-bindgen. All JS generation has been refactored and updated to now process these lists of bindings instead of the previous `Descriptor`. In other words this commit takes `js2rust.rs` and `rust2js.rs` and first splits them in two. Interpretation of `Descriptor` and what to do for conversions is in the binding selection modules. The actual generation of JS from the binding selection is now performed by `incoming.rs` and `outgoing.rs`. To boot this also deduplicates all the code between the argument handling of `js2rust.rs` and return value handling of `rust2js.rs`. This means that to implement a new binding you only need to implement it one place and it's implemented for free in the other! This commit is not the end of the story though. I would like to add a mdoe to `wasm-bindgen` that literally emits a WebIDL bindings section. That's left for a third (and hopefully final) refactoring which is also intended to optimize generated JS for bindings. This commit currently loses the optimization where an imported is hooked up by value directly whenever a shim isn't needed. It's planned that the next refactoring to emit a webidl binding section that can be added back in. It shouldn't be too too hard hopefully since all the scaffolding is in place now. cc #1524
2019-06-10 07:09:51 -07:00
Ok(())
First refactor for WebIDL bindings This commit starts the `wasm-bindgen` CLI tool down the road to being a true polyfill for WebIDL bindings. This refactor is probably the first of a few, but is hopefully the largest and most sprawling and everything will be a bit more targeted from here on out. The goal of this refactoring is to separate out the massive `crates/cli-support/src/js/mod.rs` into a number of separate pieces of functionality. It currently takes care of basically everything including: * Binding intrinsics * Handling anyref transformations * Generating all JS for imports/exports * All the logic for how to import and how to name imports * Execution and management of wasm-bindgen closures Many of these are separable concerns and most overlap with WebIDL bindings. The internal refactoring here is intended to make it more clear who's responsible for what as well as making some existing operations much more straightforward. At a high-level, the following changes are done: 1. A `src/webidl.rs` module is introduced. The purpose of this module is to take all of the raw wasm-bindgen custom sections from the module and transform them into a WebIDL bindings section. This module has a placeholder `WebidlCustomSection` which is nowhere near the actual custom section but if you squint is in theory very similar. It's hoped that this will eventually become the true WebIDL custom section, currently being developed in an external crate. Currently, however, the WebIDL bindings custom section only covers a subset of the functionality we export to wasm-bindgen users. To avoid leaving them high and dry this module also contains an auxiliary custom section named `WasmBindgenAux`. This custom section isn't intended to have a binary format, but is intended to represent a theoretical custom section necessary to couple with WebIDL bindings to achieve all our desired functionality in `wasm-bindgen`. It'll never be standardized, but it'll also never be serialized :) 2. The `src/webidl.rs` module now takes over quite a bit of functionality from `src/js/mod.rs`. Namely it handles synthesis of an `export_map` and an `import_map` mapping export/import IDs to exactly what's expected to be hooked up there. This does not include type information (as that's in the bindings section) but rather includes things like "this is the method of class A" or "this import is from module `foo`" and things like that. These could arguably be subsumed by future JS features as well, but that's for another time! 3. All handling of wasm-bindgen "descriptor functions" now happens in a dedicated `src/descriptors.rs` module. The output of this module is its own custom section (intended to be immediately consumed by the WebIDL module) which is in theory what we want to ourselves emit one day but rustc isn't capable of doing so right now. 4. Invocations and generations of imports are completely overhauled. Using the `import_map` generated in the WebIDL step all imports are now handled much more precisely in one location rather than haphazardly throughout the module. This means we have precise information about each import of the module and we only modify exactly what we're looking at. This also vastly simplifies intrinsic generation since it's all simply a codegen part of the `rust2js.rs` module now. 5. Handling of direct imports which don't have a JS shim generated is slightly different from before and is intended to be future-compatible with WebIDL bindings in its full glory, but we'll need to update it to handle cases for constructors and method calls eventually as well. 6. Intrinsic definitions now live in their own file (`src/intrinsic.rs`) and have a separated definition for their symbol name and signature. The actual implementation of each intrinsic lives in `rust2js.rs` There's a number of TODO items to finish before this merges. This includes reimplementing the anyref pass and actually implementing import maps for other targets. Those will come soon in follow-up commits, but the entire `tests/wasm/main.rs` suite is currently passing and this seems like a good checkpoint.
2019-05-23 09:15:26 -07:00
}
fn program(&mut self, program: decode::Program<'a>) -> Result<(), Error> {
self.unique_crate_identifier = program.unique_crate_identifier;
let decode::Program {
exports,
enums,
imports,
structs,
typescript_custom_sections,
local_modules,
inline_js,
unique_crate_identifier,
package_json,
} = program;
for module in local_modules {
// All local modules we find should be unique, but the same module
// may have showed up in a few different blocks. If that's the case
// all the same identifiers should have the same contents.
if let Some(prev) = self
.aux
.local_modules
.insert(module.identifier.to_string(), module.contents.to_string())
{
assert_eq!(prev, module.contents);
}
}
if let Some(s) = package_json {
self.aux.package_jsons.insert(s.into());
}
for export in exports {
self.export(export)?;
}
// Register vendor prefixes for all types before we walk over all the
// imports to ensure that if a vendor prefix is listed somewhere it'll
// apply to all the imports.
for import in imports.iter() {
if let decode::ImportKind::Type(ty) = &import.kind {
if ty.vendor_prefixes.len() == 0 {
continue;
}
self.vendor_prefixes
.entry(ty.name.to_string())
.or_insert(Vec::new())
.extend(ty.vendor_prefixes.iter().map(|s| s.to_string()));
}
}
for import in imports {
self.import(import)?;
}
for enum_ in enums {
self.enum_(enum_)?;
}
for struct_ in structs {
self.struct_(struct_)?;
}
for section in typescript_custom_sections {
self.aux.extra_typescript.push_str(section);
self.aux.extra_typescript.push_str("\n\n");
}
self.aux
.snippets
.entry(unique_crate_identifier.to_string())
.or_insert(Vec::new())
.extend(inline_js.iter().map(|s| s.to_string()));
Ok(())
}
fn export(&mut self, export: decode::Export<'_>) -> Result<(), Error> {
let wasm_name = match &export.class {
Some(class) => struct_function_export_name(class, export.function.name),
None => export.function.name.to_string(),
};
let mut descriptor = match self.descriptors.remove(&wasm_name) {
First refactor for WebIDL bindings This commit starts the `wasm-bindgen` CLI tool down the road to being a true polyfill for WebIDL bindings. This refactor is probably the first of a few, but is hopefully the largest and most sprawling and everything will be a bit more targeted from here on out. The goal of this refactoring is to separate out the massive `crates/cli-support/src/js/mod.rs` into a number of separate pieces of functionality. It currently takes care of basically everything including: * Binding intrinsics * Handling anyref transformations * Generating all JS for imports/exports * All the logic for how to import and how to name imports * Execution and management of wasm-bindgen closures Many of these are separable concerns and most overlap with WebIDL bindings. The internal refactoring here is intended to make it more clear who's responsible for what as well as making some existing operations much more straightforward. At a high-level, the following changes are done: 1. A `src/webidl.rs` module is introduced. The purpose of this module is to take all of the raw wasm-bindgen custom sections from the module and transform them into a WebIDL bindings section. This module has a placeholder `WebidlCustomSection` which is nowhere near the actual custom section but if you squint is in theory very similar. It's hoped that this will eventually become the true WebIDL custom section, currently being developed in an external crate. Currently, however, the WebIDL bindings custom section only covers a subset of the functionality we export to wasm-bindgen users. To avoid leaving them high and dry this module also contains an auxiliary custom section named `WasmBindgenAux`. This custom section isn't intended to have a binary format, but is intended to represent a theoretical custom section necessary to couple with WebIDL bindings to achieve all our desired functionality in `wasm-bindgen`. It'll never be standardized, but it'll also never be serialized :) 2. The `src/webidl.rs` module now takes over quite a bit of functionality from `src/js/mod.rs`. Namely it handles synthesis of an `export_map` and an `import_map` mapping export/import IDs to exactly what's expected to be hooked up there. This does not include type information (as that's in the bindings section) but rather includes things like "this is the method of class A" or "this import is from module `foo`" and things like that. These could arguably be subsumed by future JS features as well, but that's for another time! 3. All handling of wasm-bindgen "descriptor functions" now happens in a dedicated `src/descriptors.rs` module. The output of this module is its own custom section (intended to be immediately consumed by the WebIDL module) which is in theory what we want to ourselves emit one day but rustc isn't capable of doing so right now. 4. Invocations and generations of imports are completely overhauled. Using the `import_map` generated in the WebIDL step all imports are now handled much more precisely in one location rather than haphazardly throughout the module. This means we have precise information about each import of the module and we only modify exactly what we're looking at. This also vastly simplifies intrinsic generation since it's all simply a codegen part of the `rust2js.rs` module now. 5. Handling of direct imports which don't have a JS shim generated is slightly different from before and is intended to be future-compatible with WebIDL bindings in its full glory, but we'll need to update it to handle cases for constructors and method calls eventually as well. 6. Intrinsic definitions now live in their own file (`src/intrinsic.rs`) and have a separated definition for their symbol name and signature. The actual implementation of each intrinsic lives in `rust2js.rs` There's a number of TODO items to finish before this merges. This includes reimplementing the anyref pass and actually implementing import maps for other targets. Those will come soon in follow-up commits, but the entire `tests/wasm/main.rs` suite is currently passing and this seems like a good checkpoint.
2019-05-23 09:15:26 -07:00
None => return Ok(()),
Some(d) => d.unwrap_function(),
};
let (export_id, id) = self.function_exports[&wasm_name];
if export.start {
self.add_start_function(id)?;
}
let kind = match export.class {
Some(class) => {
let class = class.to_string();
match export.method_kind {
decode::MethodKind::Constructor => AuxExportKind::Constructor(class),
decode::MethodKind::Operation(op) => match op.kind {
decode::OperationKind::Getter(f) => {
descriptor.arguments.insert(0, Descriptor::I32);
AuxExportKind::Getter {
class,
field: f.to_string(),
}
}
decode::OperationKind::Setter(f) => {
descriptor.arguments.insert(0, Descriptor::I32);
AuxExportKind::Setter {
class,
field: f.to_string(),
}
}
First refactor for WebIDL bindings This commit starts the `wasm-bindgen` CLI tool down the road to being a true polyfill for WebIDL bindings. This refactor is probably the first of a few, but is hopefully the largest and most sprawling and everything will be a bit more targeted from here on out. The goal of this refactoring is to separate out the massive `crates/cli-support/src/js/mod.rs` into a number of separate pieces of functionality. It currently takes care of basically everything including: * Binding intrinsics * Handling anyref transformations * Generating all JS for imports/exports * All the logic for how to import and how to name imports * Execution and management of wasm-bindgen closures Many of these are separable concerns and most overlap with WebIDL bindings. The internal refactoring here is intended to make it more clear who's responsible for what as well as making some existing operations much more straightforward. At a high-level, the following changes are done: 1. A `src/webidl.rs` module is introduced. The purpose of this module is to take all of the raw wasm-bindgen custom sections from the module and transform them into a WebIDL bindings section. This module has a placeholder `WebidlCustomSection` which is nowhere near the actual custom section but if you squint is in theory very similar. It's hoped that this will eventually become the true WebIDL custom section, currently being developed in an external crate. Currently, however, the WebIDL bindings custom section only covers a subset of the functionality we export to wasm-bindgen users. To avoid leaving them high and dry this module also contains an auxiliary custom section named `WasmBindgenAux`. This custom section isn't intended to have a binary format, but is intended to represent a theoretical custom section necessary to couple with WebIDL bindings to achieve all our desired functionality in `wasm-bindgen`. It'll never be standardized, but it'll also never be serialized :) 2. The `src/webidl.rs` module now takes over quite a bit of functionality from `src/js/mod.rs`. Namely it handles synthesis of an `export_map` and an `import_map` mapping export/import IDs to exactly what's expected to be hooked up there. This does not include type information (as that's in the bindings section) but rather includes things like "this is the method of class A" or "this import is from module `foo`" and things like that. These could arguably be subsumed by future JS features as well, but that's for another time! 3. All handling of wasm-bindgen "descriptor functions" now happens in a dedicated `src/descriptors.rs` module. The output of this module is its own custom section (intended to be immediately consumed by the WebIDL module) which is in theory what we want to ourselves emit one day but rustc isn't capable of doing so right now. 4. Invocations and generations of imports are completely overhauled. Using the `import_map` generated in the WebIDL step all imports are now handled much more precisely in one location rather than haphazardly throughout the module. This means we have precise information about each import of the module and we only modify exactly what we're looking at. This also vastly simplifies intrinsic generation since it's all simply a codegen part of the `rust2js.rs` module now. 5. Handling of direct imports which don't have a JS shim generated is slightly different from before and is intended to be future-compatible with WebIDL bindings in its full glory, but we'll need to update it to handle cases for constructors and method calls eventually as well. 6. Intrinsic definitions now live in their own file (`src/intrinsic.rs`) and have a separated definition for their symbol name and signature. The actual implementation of each intrinsic lives in `rust2js.rs` There's a number of TODO items to finish before this merges. This includes reimplementing the anyref pass and actually implementing import maps for other targets. Those will come soon in follow-up commits, but the entire `tests/wasm/main.rs` suite is currently passing and this seems like a good checkpoint.
2019-05-23 09:15:26 -07:00
_ if op.is_static => AuxExportKind::StaticFunction {
class,
name: export.function.name.to_string(),
},
_ => {
descriptor.arguments.insert(0, Descriptor::I32);
AuxExportKind::Method {
class,
name: export.function.name.to_string(),
consumed: export.consumed,
}
}
First refactor for WebIDL bindings This commit starts the `wasm-bindgen` CLI tool down the road to being a true polyfill for WebIDL bindings. This refactor is probably the first of a few, but is hopefully the largest and most sprawling and everything will be a bit more targeted from here on out. The goal of this refactoring is to separate out the massive `crates/cli-support/src/js/mod.rs` into a number of separate pieces of functionality. It currently takes care of basically everything including: * Binding intrinsics * Handling anyref transformations * Generating all JS for imports/exports * All the logic for how to import and how to name imports * Execution and management of wasm-bindgen closures Many of these are separable concerns and most overlap with WebIDL bindings. The internal refactoring here is intended to make it more clear who's responsible for what as well as making some existing operations much more straightforward. At a high-level, the following changes are done: 1. A `src/webidl.rs` module is introduced. The purpose of this module is to take all of the raw wasm-bindgen custom sections from the module and transform them into a WebIDL bindings section. This module has a placeholder `WebidlCustomSection` which is nowhere near the actual custom section but if you squint is in theory very similar. It's hoped that this will eventually become the true WebIDL custom section, currently being developed in an external crate. Currently, however, the WebIDL bindings custom section only covers a subset of the functionality we export to wasm-bindgen users. To avoid leaving them high and dry this module also contains an auxiliary custom section named `WasmBindgenAux`. This custom section isn't intended to have a binary format, but is intended to represent a theoretical custom section necessary to couple with WebIDL bindings to achieve all our desired functionality in `wasm-bindgen`. It'll never be standardized, but it'll also never be serialized :) 2. The `src/webidl.rs` module now takes over quite a bit of functionality from `src/js/mod.rs`. Namely it handles synthesis of an `export_map` and an `import_map` mapping export/import IDs to exactly what's expected to be hooked up there. This does not include type information (as that's in the bindings section) but rather includes things like "this is the method of class A" or "this import is from module `foo`" and things like that. These could arguably be subsumed by future JS features as well, but that's for another time! 3. All handling of wasm-bindgen "descriptor functions" now happens in a dedicated `src/descriptors.rs` module. The output of this module is its own custom section (intended to be immediately consumed by the WebIDL module) which is in theory what we want to ourselves emit one day but rustc isn't capable of doing so right now. 4. Invocations and generations of imports are completely overhauled. Using the `import_map` generated in the WebIDL step all imports are now handled much more precisely in one location rather than haphazardly throughout the module. This means we have precise information about each import of the module and we only modify exactly what we're looking at. This also vastly simplifies intrinsic generation since it's all simply a codegen part of the `rust2js.rs` module now. 5. Handling of direct imports which don't have a JS shim generated is slightly different from before and is intended to be future-compatible with WebIDL bindings in its full glory, but we'll need to update it to handle cases for constructors and method calls eventually as well. 6. Intrinsic definitions now live in their own file (`src/intrinsic.rs`) and have a separated definition for their symbol name and signature. The actual implementation of each intrinsic lives in `rust2js.rs` There's a number of TODO items to finish before this merges. This includes reimplementing the anyref pass and actually implementing import maps for other targets. Those will come soon in follow-up commits, but the entire `tests/wasm/main.rs` suite is currently passing and this seems like a good checkpoint.
2019-05-23 09:15:26 -07:00
},
}
}
None => AuxExportKind::Function(export.function.name.to_string()),
};
self.aux.export_map.insert(
export_id,
AuxExport {
debug_name: wasm_name,
comments: concatenate_comments(&export.comments),
arg_names: Some(export.function.arg_names),
kind,
},
);
Second large refactor for WebIDL bindings This commit is the second, and hopefully last massive, refactor for using WebIDL bindings internally in `wasm-bindgen`. This commit actually fully executes on the task at hand, moving `wasm-bindgen` to internally using WebIDL bindings throughout its code generation, anyref passes, etc. This actually fixes a number of issues that have existed in the anyref pass for some time now! The main changes here are to basically remove the usage of `Descriptor` from generating JS bindings. Instead two new types are introduced: `NonstandardIncoming` and `NonstandardOutgoing` which are bindings lists used for incoming/outgoing bindings. These mirror the standard terminology and literally have variants which are the standard values. All `Descriptor` types are now mapped into lists of incoming/outgoing bindings and used for process in wasm-bindgen. All JS generation has been refactored and updated to now process these lists of bindings instead of the previous `Descriptor`. In other words this commit takes `js2rust.rs` and `rust2js.rs` and first splits them in two. Interpretation of `Descriptor` and what to do for conversions is in the binding selection modules. The actual generation of JS from the binding selection is now performed by `incoming.rs` and `outgoing.rs`. To boot this also deduplicates all the code between the argument handling of `js2rust.rs` and return value handling of `rust2js.rs`. This means that to implement a new binding you only need to implement it one place and it's implemented for free in the other! This commit is not the end of the story though. I would like to add a mdoe to `wasm-bindgen` that literally emits a WebIDL bindings section. That's left for a third (and hopefully final) refactoring which is also intended to optimize generated JS for bindings. This commit currently loses the optimization where an imported is hooked up by value directly whenever a shim isn't needed. It's planned that the next refactoring to emit a webidl binding section that can be added back in. It shouldn't be too too hard hopefully since all the scaffolding is in place now. cc #1524
2019-06-10 07:09:51 -07:00
bindings::register_export(self.module, &mut self.bindings, export_id, descriptor)?;
First refactor for WebIDL bindings This commit starts the `wasm-bindgen` CLI tool down the road to being a true polyfill for WebIDL bindings. This refactor is probably the first of a few, but is hopefully the largest and most sprawling and everything will be a bit more targeted from here on out. The goal of this refactoring is to separate out the massive `crates/cli-support/src/js/mod.rs` into a number of separate pieces of functionality. It currently takes care of basically everything including: * Binding intrinsics * Handling anyref transformations * Generating all JS for imports/exports * All the logic for how to import and how to name imports * Execution and management of wasm-bindgen closures Many of these are separable concerns and most overlap with WebIDL bindings. The internal refactoring here is intended to make it more clear who's responsible for what as well as making some existing operations much more straightforward. At a high-level, the following changes are done: 1. A `src/webidl.rs` module is introduced. The purpose of this module is to take all of the raw wasm-bindgen custom sections from the module and transform them into a WebIDL bindings section. This module has a placeholder `WebidlCustomSection` which is nowhere near the actual custom section but if you squint is in theory very similar. It's hoped that this will eventually become the true WebIDL custom section, currently being developed in an external crate. Currently, however, the WebIDL bindings custom section only covers a subset of the functionality we export to wasm-bindgen users. To avoid leaving them high and dry this module also contains an auxiliary custom section named `WasmBindgenAux`. This custom section isn't intended to have a binary format, but is intended to represent a theoretical custom section necessary to couple with WebIDL bindings to achieve all our desired functionality in `wasm-bindgen`. It'll never be standardized, but it'll also never be serialized :) 2. The `src/webidl.rs` module now takes over quite a bit of functionality from `src/js/mod.rs`. Namely it handles synthesis of an `export_map` and an `import_map` mapping export/import IDs to exactly what's expected to be hooked up there. This does not include type information (as that's in the bindings section) but rather includes things like "this is the method of class A" or "this import is from module `foo`" and things like that. These could arguably be subsumed by future JS features as well, but that's for another time! 3. All handling of wasm-bindgen "descriptor functions" now happens in a dedicated `src/descriptors.rs` module. The output of this module is its own custom section (intended to be immediately consumed by the WebIDL module) which is in theory what we want to ourselves emit one day but rustc isn't capable of doing so right now. 4. Invocations and generations of imports are completely overhauled. Using the `import_map` generated in the WebIDL step all imports are now handled much more precisely in one location rather than haphazardly throughout the module. This means we have precise information about each import of the module and we only modify exactly what we're looking at. This also vastly simplifies intrinsic generation since it's all simply a codegen part of the `rust2js.rs` module now. 5. Handling of direct imports which don't have a JS shim generated is slightly different from before and is intended to be future-compatible with WebIDL bindings in its full glory, but we'll need to update it to handle cases for constructors and method calls eventually as well. 6. Intrinsic definitions now live in their own file (`src/intrinsic.rs`) and have a separated definition for their symbol name and signature. The actual implementation of each intrinsic lives in `rust2js.rs` There's a number of TODO items to finish before this merges. This includes reimplementing the anyref pass and actually implementing import maps for other targets. Those will come soon in follow-up commits, but the entire `tests/wasm/main.rs` suite is currently passing and this seems like a good checkpoint.
2019-05-23 09:15:26 -07:00
Ok(())
}
fn add_start_function(&mut self, id: FunctionId) -> Result<(), Error> {
if self.start_found {
bail!("cannot specify two `start` functions");
}
self.start_found = true;
let prev_start = match self.module.start {
Some(f) => f,
None => {
self.module.start = Some(id);
return Ok(());
}
};
// Note that we call the previous start function, if any, first. This is
// because the start function currently only shows up when it's injected
// through thread/anyref transforms. These injected start functions need
// to happen before user code, so we always schedule them first.
let mut builder = walrus::FunctionBuilder::new();
let call1 = builder.call(prev_start, Box::new([]));
let call2 = builder.call(id, Box::new([]));
let ty = self.module.funcs.get(id).ty();
let new_start = builder.finish(ty, Vec::new(), vec![call1, call2], self.module);
self.module.start = Some(new_start);
Ok(())
}
fn import(&mut self, import: decode::Import<'_>) -> Result<(), Error> {
match &import.kind {
decode::ImportKind::Function(f) => self.import_function(&import, f),
decode::ImportKind::Static(s) => self.import_static(&import, s),
decode::ImportKind::Type(t) => self.import_type(&import, t),
decode::ImportKind::Enum(_) => Ok(()),
}
}
fn import_function(
&mut self,
import: &decode::Import<'_>,
function: &decode::ImportFunction<'_>,
) -> Result<(), Error> {
let decode::ImportFunction {
shim,
catch,
variadic,
method,
structural,
function,
assert_no_shim,
First refactor for WebIDL bindings This commit starts the `wasm-bindgen` CLI tool down the road to being a true polyfill for WebIDL bindings. This refactor is probably the first of a few, but is hopefully the largest and most sprawling and everything will be a bit more targeted from here on out. The goal of this refactoring is to separate out the massive `crates/cli-support/src/js/mod.rs` into a number of separate pieces of functionality. It currently takes care of basically everything including: * Binding intrinsics * Handling anyref transformations * Generating all JS for imports/exports * All the logic for how to import and how to name imports * Execution and management of wasm-bindgen closures Many of these are separable concerns and most overlap with WebIDL bindings. The internal refactoring here is intended to make it more clear who's responsible for what as well as making some existing operations much more straightforward. At a high-level, the following changes are done: 1. A `src/webidl.rs` module is introduced. The purpose of this module is to take all of the raw wasm-bindgen custom sections from the module and transform them into a WebIDL bindings section. This module has a placeholder `WebidlCustomSection` which is nowhere near the actual custom section but if you squint is in theory very similar. It's hoped that this will eventually become the true WebIDL custom section, currently being developed in an external crate. Currently, however, the WebIDL bindings custom section only covers a subset of the functionality we export to wasm-bindgen users. To avoid leaving them high and dry this module also contains an auxiliary custom section named `WasmBindgenAux`. This custom section isn't intended to have a binary format, but is intended to represent a theoretical custom section necessary to couple with WebIDL bindings to achieve all our desired functionality in `wasm-bindgen`. It'll never be standardized, but it'll also never be serialized :) 2. The `src/webidl.rs` module now takes over quite a bit of functionality from `src/js/mod.rs`. Namely it handles synthesis of an `export_map` and an `import_map` mapping export/import IDs to exactly what's expected to be hooked up there. This does not include type information (as that's in the bindings section) but rather includes things like "this is the method of class A" or "this import is from module `foo`" and things like that. These could arguably be subsumed by future JS features as well, but that's for another time! 3. All handling of wasm-bindgen "descriptor functions" now happens in a dedicated `src/descriptors.rs` module. The output of this module is its own custom section (intended to be immediately consumed by the WebIDL module) which is in theory what we want to ourselves emit one day but rustc isn't capable of doing so right now. 4. Invocations and generations of imports are completely overhauled. Using the `import_map` generated in the WebIDL step all imports are now handled much more precisely in one location rather than haphazardly throughout the module. This means we have precise information about each import of the module and we only modify exactly what we're looking at. This also vastly simplifies intrinsic generation since it's all simply a codegen part of the `rust2js.rs` module now. 5. Handling of direct imports which don't have a JS shim generated is slightly different from before and is intended to be future-compatible with WebIDL bindings in its full glory, but we'll need to update it to handle cases for constructors and method calls eventually as well. 6. Intrinsic definitions now live in their own file (`src/intrinsic.rs`) and have a separated definition for their symbol name and signature. The actual implementation of each intrinsic lives in `rust2js.rs` There's a number of TODO items to finish before this merges. This includes reimplementing the anyref pass and actually implementing import maps for other targets. Those will come soon in follow-up commits, but the entire `tests/wasm/main.rs` suite is currently passing and this seems like a good checkpoint.
2019-05-23 09:15:26 -07:00
} = function;
let (import_id, _id) = match self.function_imports.get(*shim) {
Some(pair) => *pair,
None => return Ok(()),
};
let descriptor = match self.descriptors.remove(*shim) {
None => return Ok(()),
Some(d) => d.unwrap_function(),
};
// Record this for later as it affects JS binding generation, but note
// that this doesn't affect the WebIDL interface at all.
if *variadic {
self.aux.imports_with_variadic.insert(import_id);
}
if *catch {
self.aux.imports_with_catch.insert(import_id);
}
if *assert_no_shim {
self.aux.imports_with_assert_no_shim.insert(import_id);
}
First refactor for WebIDL bindings This commit starts the `wasm-bindgen` CLI tool down the road to being a true polyfill for WebIDL bindings. This refactor is probably the first of a few, but is hopefully the largest and most sprawling and everything will be a bit more targeted from here on out. The goal of this refactoring is to separate out the massive `crates/cli-support/src/js/mod.rs` into a number of separate pieces of functionality. It currently takes care of basically everything including: * Binding intrinsics * Handling anyref transformations * Generating all JS for imports/exports * All the logic for how to import and how to name imports * Execution and management of wasm-bindgen closures Many of these are separable concerns and most overlap with WebIDL bindings. The internal refactoring here is intended to make it more clear who's responsible for what as well as making some existing operations much more straightforward. At a high-level, the following changes are done: 1. A `src/webidl.rs` module is introduced. The purpose of this module is to take all of the raw wasm-bindgen custom sections from the module and transform them into a WebIDL bindings section. This module has a placeholder `WebidlCustomSection` which is nowhere near the actual custom section but if you squint is in theory very similar. It's hoped that this will eventually become the true WebIDL custom section, currently being developed in an external crate. Currently, however, the WebIDL bindings custom section only covers a subset of the functionality we export to wasm-bindgen users. To avoid leaving them high and dry this module also contains an auxiliary custom section named `WasmBindgenAux`. This custom section isn't intended to have a binary format, but is intended to represent a theoretical custom section necessary to couple with WebIDL bindings to achieve all our desired functionality in `wasm-bindgen`. It'll never be standardized, but it'll also never be serialized :) 2. The `src/webidl.rs` module now takes over quite a bit of functionality from `src/js/mod.rs`. Namely it handles synthesis of an `export_map` and an `import_map` mapping export/import IDs to exactly what's expected to be hooked up there. This does not include type information (as that's in the bindings section) but rather includes things like "this is the method of class A" or "this import is from module `foo`" and things like that. These could arguably be subsumed by future JS features as well, but that's for another time! 3. All handling of wasm-bindgen "descriptor functions" now happens in a dedicated `src/descriptors.rs` module. The output of this module is its own custom section (intended to be immediately consumed by the WebIDL module) which is in theory what we want to ourselves emit one day but rustc isn't capable of doing so right now. 4. Invocations and generations of imports are completely overhauled. Using the `import_map` generated in the WebIDL step all imports are now handled much more precisely in one location rather than haphazardly throughout the module. This means we have precise information about each import of the module and we only modify exactly what we're looking at. This also vastly simplifies intrinsic generation since it's all simply a codegen part of the `rust2js.rs` module now. 5. Handling of direct imports which don't have a JS shim generated is slightly different from before and is intended to be future-compatible with WebIDL bindings in its full glory, but we'll need to update it to handle cases for constructors and method calls eventually as well. 6. Intrinsic definitions now live in their own file (`src/intrinsic.rs`) and have a separated definition for their symbol name and signature. The actual implementation of each intrinsic lives in `rust2js.rs` There's a number of TODO items to finish before this merges. This includes reimplementing the anyref pass and actually implementing import maps for other targets. Those will come soon in follow-up commits, but the entire `tests/wasm/main.rs` suite is currently passing and this seems like a good checkpoint.
2019-05-23 09:15:26 -07:00
// Perform two functions here. First we're saving off our WebIDL
// bindings signature, indicating what we think our import is going to
// be. Next we're saving off other metadata indicating where this item
// is going to be imported from. The `import_map` table will record, for
// each import, what is getting hooked up to that slot of the import
// table to the WebAssembly instance.
let import = match method {
Some(data) => {
let class = self.determine_import(import, &data.class)?;
match &data.kind {
// NB: `structural` is ignored for constructors since the
// js type isn't expected to change anyway.
decode::MethodKind::Constructor => {
Second large refactor for WebIDL bindings This commit is the second, and hopefully last massive, refactor for using WebIDL bindings internally in `wasm-bindgen`. This commit actually fully executes on the task at hand, moving `wasm-bindgen` to internally using WebIDL bindings throughout its code generation, anyref passes, etc. This actually fixes a number of issues that have existed in the anyref pass for some time now! The main changes here are to basically remove the usage of `Descriptor` from generating JS bindings. Instead two new types are introduced: `NonstandardIncoming` and `NonstandardOutgoing` which are bindings lists used for incoming/outgoing bindings. These mirror the standard terminology and literally have variants which are the standard values. All `Descriptor` types are now mapped into lists of incoming/outgoing bindings and used for process in wasm-bindgen. All JS generation has been refactored and updated to now process these lists of bindings instead of the previous `Descriptor`. In other words this commit takes `js2rust.rs` and `rust2js.rs` and first splits them in two. Interpretation of `Descriptor` and what to do for conversions is in the binding selection modules. The actual generation of JS from the binding selection is now performed by `incoming.rs` and `outgoing.rs`. To boot this also deduplicates all the code between the argument handling of `js2rust.rs` and return value handling of `rust2js.rs`. This means that to implement a new binding you only need to implement it one place and it's implemented for free in the other! This commit is not the end of the story though. I would like to add a mdoe to `wasm-bindgen` that literally emits a WebIDL bindings section. That's left for a third (and hopefully final) refactoring which is also intended to optimize generated JS for bindings. This commit currently loses the optimization where an imported is hooked up by value directly whenever a shim isn't needed. It's planned that the next refactoring to emit a webidl binding section that can be added back in. It shouldn't be too too hard hopefully since all the scaffolding is in place now. cc #1524
2019-06-10 07:09:51 -07:00
bindings::register_import(
self.module,
&mut self.bindings,
import_id,
descriptor,
ast::WebidlFunctionKind::Constructor,
)?;
First refactor for WebIDL bindings This commit starts the `wasm-bindgen` CLI tool down the road to being a true polyfill for WebIDL bindings. This refactor is probably the first of a few, but is hopefully the largest and most sprawling and everything will be a bit more targeted from here on out. The goal of this refactoring is to separate out the massive `crates/cli-support/src/js/mod.rs` into a number of separate pieces of functionality. It currently takes care of basically everything including: * Binding intrinsics * Handling anyref transformations * Generating all JS for imports/exports * All the logic for how to import and how to name imports * Execution and management of wasm-bindgen closures Many of these are separable concerns and most overlap with WebIDL bindings. The internal refactoring here is intended to make it more clear who's responsible for what as well as making some existing operations much more straightforward. At a high-level, the following changes are done: 1. A `src/webidl.rs` module is introduced. The purpose of this module is to take all of the raw wasm-bindgen custom sections from the module and transform them into a WebIDL bindings section. This module has a placeholder `WebidlCustomSection` which is nowhere near the actual custom section but if you squint is in theory very similar. It's hoped that this will eventually become the true WebIDL custom section, currently being developed in an external crate. Currently, however, the WebIDL bindings custom section only covers a subset of the functionality we export to wasm-bindgen users. To avoid leaving them high and dry this module also contains an auxiliary custom section named `WasmBindgenAux`. This custom section isn't intended to have a binary format, but is intended to represent a theoretical custom section necessary to couple with WebIDL bindings to achieve all our desired functionality in `wasm-bindgen`. It'll never be standardized, but it'll also never be serialized :) 2. The `src/webidl.rs` module now takes over quite a bit of functionality from `src/js/mod.rs`. Namely it handles synthesis of an `export_map` and an `import_map` mapping export/import IDs to exactly what's expected to be hooked up there. This does not include type information (as that's in the bindings section) but rather includes things like "this is the method of class A" or "this import is from module `foo`" and things like that. These could arguably be subsumed by future JS features as well, but that's for another time! 3. All handling of wasm-bindgen "descriptor functions" now happens in a dedicated `src/descriptors.rs` module. The output of this module is its own custom section (intended to be immediately consumed by the WebIDL module) which is in theory what we want to ourselves emit one day but rustc isn't capable of doing so right now. 4. Invocations and generations of imports are completely overhauled. Using the `import_map` generated in the WebIDL step all imports are now handled much more precisely in one location rather than haphazardly throughout the module. This means we have precise information about each import of the module and we only modify exactly what we're looking at. This also vastly simplifies intrinsic generation since it's all simply a codegen part of the `rust2js.rs` module now. 5. Handling of direct imports which don't have a JS shim generated is slightly different from before and is intended to be future-compatible with WebIDL bindings in its full glory, but we'll need to update it to handle cases for constructors and method calls eventually as well. 6. Intrinsic definitions now live in their own file (`src/intrinsic.rs`) and have a separated definition for their symbol name and signature. The actual implementation of each intrinsic lives in `rust2js.rs` There's a number of TODO items to finish before this merges. This includes reimplementing the anyref pass and actually implementing import maps for other targets. Those will come soon in follow-up commits, but the entire `tests/wasm/main.rs` suite is currently passing and this seems like a good checkpoint.
2019-05-23 09:15:26 -07:00
AuxImport::Value(AuxValue::Bare(class))
}
decode::MethodKind::Operation(op) => {
let (import, method) =
self.determine_import_op(class, function, *structural, op)?;
Second large refactor for WebIDL bindings This commit is the second, and hopefully last massive, refactor for using WebIDL bindings internally in `wasm-bindgen`. This commit actually fully executes on the task at hand, moving `wasm-bindgen` to internally using WebIDL bindings throughout its code generation, anyref passes, etc. This actually fixes a number of issues that have existed in the anyref pass for some time now! The main changes here are to basically remove the usage of `Descriptor` from generating JS bindings. Instead two new types are introduced: `NonstandardIncoming` and `NonstandardOutgoing` which are bindings lists used for incoming/outgoing bindings. These mirror the standard terminology and literally have variants which are the standard values. All `Descriptor` types are now mapped into lists of incoming/outgoing bindings and used for process in wasm-bindgen. All JS generation has been refactored and updated to now process these lists of bindings instead of the previous `Descriptor`. In other words this commit takes `js2rust.rs` and `rust2js.rs` and first splits them in two. Interpretation of `Descriptor` and what to do for conversions is in the binding selection modules. The actual generation of JS from the binding selection is now performed by `incoming.rs` and `outgoing.rs`. To boot this also deduplicates all the code between the argument handling of `js2rust.rs` and return value handling of `rust2js.rs`. This means that to implement a new binding you only need to implement it one place and it's implemented for free in the other! This commit is not the end of the story though. I would like to add a mdoe to `wasm-bindgen` that literally emits a WebIDL bindings section. That's left for a third (and hopefully final) refactoring which is also intended to optimize generated JS for bindings. This commit currently loses the optimization where an imported is hooked up by value directly whenever a shim isn't needed. It's planned that the next refactoring to emit a webidl binding section that can be added back in. It shouldn't be too too hard hopefully since all the scaffolding is in place now. cc #1524
2019-06-10 07:09:51 -07:00
let kind = if method {
let kind = ast::WebidlFunctionKindMethod {
// TODO: what should this actually be?
ty: ast::WebidlScalarType::Any.into(),
};
ast::WebidlFunctionKind::Method(kind)
First refactor for WebIDL bindings This commit starts the `wasm-bindgen` CLI tool down the road to being a true polyfill for WebIDL bindings. This refactor is probably the first of a few, but is hopefully the largest and most sprawling and everything will be a bit more targeted from here on out. The goal of this refactoring is to separate out the massive `crates/cli-support/src/js/mod.rs` into a number of separate pieces of functionality. It currently takes care of basically everything including: * Binding intrinsics * Handling anyref transformations * Generating all JS for imports/exports * All the logic for how to import and how to name imports * Execution and management of wasm-bindgen closures Many of these are separable concerns and most overlap with WebIDL bindings. The internal refactoring here is intended to make it more clear who's responsible for what as well as making some existing operations much more straightforward. At a high-level, the following changes are done: 1. A `src/webidl.rs` module is introduced. The purpose of this module is to take all of the raw wasm-bindgen custom sections from the module and transform them into a WebIDL bindings section. This module has a placeholder `WebidlCustomSection` which is nowhere near the actual custom section but if you squint is in theory very similar. It's hoped that this will eventually become the true WebIDL custom section, currently being developed in an external crate. Currently, however, the WebIDL bindings custom section only covers a subset of the functionality we export to wasm-bindgen users. To avoid leaving them high and dry this module also contains an auxiliary custom section named `WasmBindgenAux`. This custom section isn't intended to have a binary format, but is intended to represent a theoretical custom section necessary to couple with WebIDL bindings to achieve all our desired functionality in `wasm-bindgen`. It'll never be standardized, but it'll also never be serialized :) 2. The `src/webidl.rs` module now takes over quite a bit of functionality from `src/js/mod.rs`. Namely it handles synthesis of an `export_map` and an `import_map` mapping export/import IDs to exactly what's expected to be hooked up there. This does not include type information (as that's in the bindings section) but rather includes things like "this is the method of class A" or "this import is from module `foo`" and things like that. These could arguably be subsumed by future JS features as well, but that's for another time! 3. All handling of wasm-bindgen "descriptor functions" now happens in a dedicated `src/descriptors.rs` module. The output of this module is its own custom section (intended to be immediately consumed by the WebIDL module) which is in theory what we want to ourselves emit one day but rustc isn't capable of doing so right now. 4. Invocations and generations of imports are completely overhauled. Using the `import_map` generated in the WebIDL step all imports are now handled much more precisely in one location rather than haphazardly throughout the module. This means we have precise information about each import of the module and we only modify exactly what we're looking at. This also vastly simplifies intrinsic generation since it's all simply a codegen part of the `rust2js.rs` module now. 5. Handling of direct imports which don't have a JS shim generated is slightly different from before and is intended to be future-compatible with WebIDL bindings in its full glory, but we'll need to update it to handle cases for constructors and method calls eventually as well. 6. Intrinsic definitions now live in their own file (`src/intrinsic.rs`) and have a separated definition for their symbol name and signature. The actual implementation of each intrinsic lives in `rust2js.rs` There's a number of TODO items to finish before this merges. This includes reimplementing the anyref pass and actually implementing import maps for other targets. Those will come soon in follow-up commits, but the entire `tests/wasm/main.rs` suite is currently passing and this seems like a good checkpoint.
2019-05-23 09:15:26 -07:00
} else {
Second large refactor for WebIDL bindings This commit is the second, and hopefully last massive, refactor for using WebIDL bindings internally in `wasm-bindgen`. This commit actually fully executes on the task at hand, moving `wasm-bindgen` to internally using WebIDL bindings throughout its code generation, anyref passes, etc. This actually fixes a number of issues that have existed in the anyref pass for some time now! The main changes here are to basically remove the usage of `Descriptor` from generating JS bindings. Instead two new types are introduced: `NonstandardIncoming` and `NonstandardOutgoing` which are bindings lists used for incoming/outgoing bindings. These mirror the standard terminology and literally have variants which are the standard values. All `Descriptor` types are now mapped into lists of incoming/outgoing bindings and used for process in wasm-bindgen. All JS generation has been refactored and updated to now process these lists of bindings instead of the previous `Descriptor`. In other words this commit takes `js2rust.rs` and `rust2js.rs` and first splits them in two. Interpretation of `Descriptor` and what to do for conversions is in the binding selection modules. The actual generation of JS from the binding selection is now performed by `incoming.rs` and `outgoing.rs`. To boot this also deduplicates all the code between the argument handling of `js2rust.rs` and return value handling of `rust2js.rs`. This means that to implement a new binding you only need to implement it one place and it's implemented for free in the other! This commit is not the end of the story though. I would like to add a mdoe to `wasm-bindgen` that literally emits a WebIDL bindings section. That's left for a third (and hopefully final) refactoring which is also intended to optimize generated JS for bindings. This commit currently loses the optimization where an imported is hooked up by value directly whenever a shim isn't needed. It's planned that the next refactoring to emit a webidl binding section that can be added back in. It shouldn't be too too hard hopefully since all the scaffolding is in place now. cc #1524
2019-06-10 07:09:51 -07:00
ast::WebidlFunctionKind::Static
First refactor for WebIDL bindings This commit starts the `wasm-bindgen` CLI tool down the road to being a true polyfill for WebIDL bindings. This refactor is probably the first of a few, but is hopefully the largest and most sprawling and everything will be a bit more targeted from here on out. The goal of this refactoring is to separate out the massive `crates/cli-support/src/js/mod.rs` into a number of separate pieces of functionality. It currently takes care of basically everything including: * Binding intrinsics * Handling anyref transformations * Generating all JS for imports/exports * All the logic for how to import and how to name imports * Execution and management of wasm-bindgen closures Many of these are separable concerns and most overlap with WebIDL bindings. The internal refactoring here is intended to make it more clear who's responsible for what as well as making some existing operations much more straightforward. At a high-level, the following changes are done: 1. A `src/webidl.rs` module is introduced. The purpose of this module is to take all of the raw wasm-bindgen custom sections from the module and transform them into a WebIDL bindings section. This module has a placeholder `WebidlCustomSection` which is nowhere near the actual custom section but if you squint is in theory very similar. It's hoped that this will eventually become the true WebIDL custom section, currently being developed in an external crate. Currently, however, the WebIDL bindings custom section only covers a subset of the functionality we export to wasm-bindgen users. To avoid leaving them high and dry this module also contains an auxiliary custom section named `WasmBindgenAux`. This custom section isn't intended to have a binary format, but is intended to represent a theoretical custom section necessary to couple with WebIDL bindings to achieve all our desired functionality in `wasm-bindgen`. It'll never be standardized, but it'll also never be serialized :) 2. The `src/webidl.rs` module now takes over quite a bit of functionality from `src/js/mod.rs`. Namely it handles synthesis of an `export_map` and an `import_map` mapping export/import IDs to exactly what's expected to be hooked up there. This does not include type information (as that's in the bindings section) but rather includes things like "this is the method of class A" or "this import is from module `foo`" and things like that. These could arguably be subsumed by future JS features as well, but that's for another time! 3. All handling of wasm-bindgen "descriptor functions" now happens in a dedicated `src/descriptors.rs` module. The output of this module is its own custom section (intended to be immediately consumed by the WebIDL module) which is in theory what we want to ourselves emit one day but rustc isn't capable of doing so right now. 4. Invocations and generations of imports are completely overhauled. Using the `import_map` generated in the WebIDL step all imports are now handled much more precisely in one location rather than haphazardly throughout the module. This means we have precise information about each import of the module and we only modify exactly what we're looking at. This also vastly simplifies intrinsic generation since it's all simply a codegen part of the `rust2js.rs` module now. 5. Handling of direct imports which don't have a JS shim generated is slightly different from before and is intended to be future-compatible with WebIDL bindings in its full glory, but we'll need to update it to handle cases for constructors and method calls eventually as well. 6. Intrinsic definitions now live in their own file (`src/intrinsic.rs`) and have a separated definition for their symbol name and signature. The actual implementation of each intrinsic lives in `rust2js.rs` There's a number of TODO items to finish before this merges. This includes reimplementing the anyref pass and actually implementing import maps for other targets. Those will come soon in follow-up commits, but the entire `tests/wasm/main.rs` suite is currently passing and this seems like a good checkpoint.
2019-05-23 09:15:26 -07:00
};
Second large refactor for WebIDL bindings This commit is the second, and hopefully last massive, refactor for using WebIDL bindings internally in `wasm-bindgen`. This commit actually fully executes on the task at hand, moving `wasm-bindgen` to internally using WebIDL bindings throughout its code generation, anyref passes, etc. This actually fixes a number of issues that have existed in the anyref pass for some time now! The main changes here are to basically remove the usage of `Descriptor` from generating JS bindings. Instead two new types are introduced: `NonstandardIncoming` and `NonstandardOutgoing` which are bindings lists used for incoming/outgoing bindings. These mirror the standard terminology and literally have variants which are the standard values. All `Descriptor` types are now mapped into lists of incoming/outgoing bindings and used for process in wasm-bindgen. All JS generation has been refactored and updated to now process these lists of bindings instead of the previous `Descriptor`. In other words this commit takes `js2rust.rs` and `rust2js.rs` and first splits them in two. Interpretation of `Descriptor` and what to do for conversions is in the binding selection modules. The actual generation of JS from the binding selection is now performed by `incoming.rs` and `outgoing.rs`. To boot this also deduplicates all the code between the argument handling of `js2rust.rs` and return value handling of `rust2js.rs`. This means that to implement a new binding you only need to implement it one place and it's implemented for free in the other! This commit is not the end of the story though. I would like to add a mdoe to `wasm-bindgen` that literally emits a WebIDL bindings section. That's left for a third (and hopefully final) refactoring which is also intended to optimize generated JS for bindings. This commit currently loses the optimization where an imported is hooked up by value directly whenever a shim isn't needed. It's planned that the next refactoring to emit a webidl binding section that can be added back in. It shouldn't be too too hard hopefully since all the scaffolding is in place now. cc #1524
2019-06-10 07:09:51 -07:00
bindings::register_import(
self.module,
&mut self.bindings,
import_id,
descriptor,
kind,
)?;
First refactor for WebIDL bindings This commit starts the `wasm-bindgen` CLI tool down the road to being a true polyfill for WebIDL bindings. This refactor is probably the first of a few, but is hopefully the largest and most sprawling and everything will be a bit more targeted from here on out. The goal of this refactoring is to separate out the massive `crates/cli-support/src/js/mod.rs` into a number of separate pieces of functionality. It currently takes care of basically everything including: * Binding intrinsics * Handling anyref transformations * Generating all JS for imports/exports * All the logic for how to import and how to name imports * Execution and management of wasm-bindgen closures Many of these are separable concerns and most overlap with WebIDL bindings. The internal refactoring here is intended to make it more clear who's responsible for what as well as making some existing operations much more straightforward. At a high-level, the following changes are done: 1. A `src/webidl.rs` module is introduced. The purpose of this module is to take all of the raw wasm-bindgen custom sections from the module and transform them into a WebIDL bindings section. This module has a placeholder `WebidlCustomSection` which is nowhere near the actual custom section but if you squint is in theory very similar. It's hoped that this will eventually become the true WebIDL custom section, currently being developed in an external crate. Currently, however, the WebIDL bindings custom section only covers a subset of the functionality we export to wasm-bindgen users. To avoid leaving them high and dry this module also contains an auxiliary custom section named `WasmBindgenAux`. This custom section isn't intended to have a binary format, but is intended to represent a theoretical custom section necessary to couple with WebIDL bindings to achieve all our desired functionality in `wasm-bindgen`. It'll never be standardized, but it'll also never be serialized :) 2. The `src/webidl.rs` module now takes over quite a bit of functionality from `src/js/mod.rs`. Namely it handles synthesis of an `export_map` and an `import_map` mapping export/import IDs to exactly what's expected to be hooked up there. This does not include type information (as that's in the bindings section) but rather includes things like "this is the method of class A" or "this import is from module `foo`" and things like that. These could arguably be subsumed by future JS features as well, but that's for another time! 3. All handling of wasm-bindgen "descriptor functions" now happens in a dedicated `src/descriptors.rs` module. The output of this module is its own custom section (intended to be immediately consumed by the WebIDL module) which is in theory what we want to ourselves emit one day but rustc isn't capable of doing so right now. 4. Invocations and generations of imports are completely overhauled. Using the `import_map` generated in the WebIDL step all imports are now handled much more precisely in one location rather than haphazardly throughout the module. This means we have precise information about each import of the module and we only modify exactly what we're looking at. This also vastly simplifies intrinsic generation since it's all simply a codegen part of the `rust2js.rs` module now. 5. Handling of direct imports which don't have a JS shim generated is slightly different from before and is intended to be future-compatible with WebIDL bindings in its full glory, but we'll need to update it to handle cases for constructors and method calls eventually as well. 6. Intrinsic definitions now live in their own file (`src/intrinsic.rs`) and have a separated definition for their symbol name and signature. The actual implementation of each intrinsic lives in `rust2js.rs` There's a number of TODO items to finish before this merges. This includes reimplementing the anyref pass and actually implementing import maps for other targets. Those will come soon in follow-up commits, but the entire `tests/wasm/main.rs` suite is currently passing and this seems like a good checkpoint.
2019-05-23 09:15:26 -07:00
import
}
}
}
// NB: `structural` is ignored for free functions since it's
// expected that the binding isn't changing anyway.
None => {
Second large refactor for WebIDL bindings This commit is the second, and hopefully last massive, refactor for using WebIDL bindings internally in `wasm-bindgen`. This commit actually fully executes on the task at hand, moving `wasm-bindgen` to internally using WebIDL bindings throughout its code generation, anyref passes, etc. This actually fixes a number of issues that have existed in the anyref pass for some time now! The main changes here are to basically remove the usage of `Descriptor` from generating JS bindings. Instead two new types are introduced: `NonstandardIncoming` and `NonstandardOutgoing` which are bindings lists used for incoming/outgoing bindings. These mirror the standard terminology and literally have variants which are the standard values. All `Descriptor` types are now mapped into lists of incoming/outgoing bindings and used for process in wasm-bindgen. All JS generation has been refactored and updated to now process these lists of bindings instead of the previous `Descriptor`. In other words this commit takes `js2rust.rs` and `rust2js.rs` and first splits them in two. Interpretation of `Descriptor` and what to do for conversions is in the binding selection modules. The actual generation of JS from the binding selection is now performed by `incoming.rs` and `outgoing.rs`. To boot this also deduplicates all the code between the argument handling of `js2rust.rs` and return value handling of `rust2js.rs`. This means that to implement a new binding you only need to implement it one place and it's implemented for free in the other! This commit is not the end of the story though. I would like to add a mdoe to `wasm-bindgen` that literally emits a WebIDL bindings section. That's left for a third (and hopefully final) refactoring which is also intended to optimize generated JS for bindings. This commit currently loses the optimization where an imported is hooked up by value directly whenever a shim isn't needed. It's planned that the next refactoring to emit a webidl binding section that can be added back in. It shouldn't be too too hard hopefully since all the scaffolding is in place now. cc #1524
2019-06-10 07:09:51 -07:00
bindings::register_import(
self.module,
&mut self.bindings,
import_id,
descriptor,
ast::WebidlFunctionKind::Static,
)?;
First refactor for WebIDL bindings This commit starts the `wasm-bindgen` CLI tool down the road to being a true polyfill for WebIDL bindings. This refactor is probably the first of a few, but is hopefully the largest and most sprawling and everything will be a bit more targeted from here on out. The goal of this refactoring is to separate out the massive `crates/cli-support/src/js/mod.rs` into a number of separate pieces of functionality. It currently takes care of basically everything including: * Binding intrinsics * Handling anyref transformations * Generating all JS for imports/exports * All the logic for how to import and how to name imports * Execution and management of wasm-bindgen closures Many of these are separable concerns and most overlap with WebIDL bindings. The internal refactoring here is intended to make it more clear who's responsible for what as well as making some existing operations much more straightforward. At a high-level, the following changes are done: 1. A `src/webidl.rs` module is introduced. The purpose of this module is to take all of the raw wasm-bindgen custom sections from the module and transform them into a WebIDL bindings section. This module has a placeholder `WebidlCustomSection` which is nowhere near the actual custom section but if you squint is in theory very similar. It's hoped that this will eventually become the true WebIDL custom section, currently being developed in an external crate. Currently, however, the WebIDL bindings custom section only covers a subset of the functionality we export to wasm-bindgen users. To avoid leaving them high and dry this module also contains an auxiliary custom section named `WasmBindgenAux`. This custom section isn't intended to have a binary format, but is intended to represent a theoretical custom section necessary to couple with WebIDL bindings to achieve all our desired functionality in `wasm-bindgen`. It'll never be standardized, but it'll also never be serialized :) 2. The `src/webidl.rs` module now takes over quite a bit of functionality from `src/js/mod.rs`. Namely it handles synthesis of an `export_map` and an `import_map` mapping export/import IDs to exactly what's expected to be hooked up there. This does not include type information (as that's in the bindings section) but rather includes things like "this is the method of class A" or "this import is from module `foo`" and things like that. These could arguably be subsumed by future JS features as well, but that's for another time! 3. All handling of wasm-bindgen "descriptor functions" now happens in a dedicated `src/descriptors.rs` module. The output of this module is its own custom section (intended to be immediately consumed by the WebIDL module) which is in theory what we want to ourselves emit one day but rustc isn't capable of doing so right now. 4. Invocations and generations of imports are completely overhauled. Using the `import_map` generated in the WebIDL step all imports are now handled much more precisely in one location rather than haphazardly throughout the module. This means we have precise information about each import of the module and we only modify exactly what we're looking at. This also vastly simplifies intrinsic generation since it's all simply a codegen part of the `rust2js.rs` module now. 5. Handling of direct imports which don't have a JS shim generated is slightly different from before and is intended to be future-compatible with WebIDL bindings in its full glory, but we'll need to update it to handle cases for constructors and method calls eventually as well. 6. Intrinsic definitions now live in their own file (`src/intrinsic.rs`) and have a separated definition for their symbol name and signature. The actual implementation of each intrinsic lives in `rust2js.rs` There's a number of TODO items to finish before this merges. This includes reimplementing the anyref pass and actually implementing import maps for other targets. Those will come soon in follow-up commits, but the entire `tests/wasm/main.rs` suite is currently passing and this seems like a good checkpoint.
2019-05-23 09:15:26 -07:00
let name = self.determine_import(import, function.name)?;
AuxImport::Value(AuxValue::Bare(name))
}
};
self.aux.import_map.insert(import_id, import);
Ok(())
}
/// The `bool` returned indicates whether the imported value should be
/// invoked as a method (first arg is implicitly `this`) or if the imported
/// value is a simple function-like shim
fn determine_import_op(
&mut self,
mut class: JsImport,
function: &decode::Function<'_>,
structural: bool,
op: &decode::Operation<'_>,
) -> Result<(AuxImport, bool), Error> {
match op.kind {
decode::OperationKind::Regular => {
if op.is_static {
class.fields.push(function.name.to_string());
Ok((AuxImport::Value(AuxValue::Bare(class)), false))
} else if structural {
Ok((
AuxImport::StructuralMethod(function.name.to_string()),
false,
))
} else {
class.fields.push("prototype".to_string());
class.fields.push(function.name.to_string());
Ok((AuxImport::Value(AuxValue::Bare(class)), true))
}
}
decode::OperationKind::Getter(field) => {
if structural {
if op.is_static {
Ok((
AuxImport::StructuralClassGetter(class, field.to_string()),
false,
))
} else {
Ok((AuxImport::StructuralGetter(field.to_string()), false))
}
} else {
let val = if op.is_static {
AuxValue::ClassGetter(class, field.to_string())
} else {
AuxValue::Getter(class, field.to_string())
};
Ok((AuxImport::Value(val), true))
}
}
decode::OperationKind::Setter(field) => {
if structural {
if op.is_static {
Ok((
AuxImport::StructuralClassSetter(class, field.to_string()),
false,
))
} else {
Ok((AuxImport::StructuralSetter(field.to_string()), false))
}
} else {
let val = if op.is_static {
AuxValue::ClassSetter(class, field.to_string())
} else {
AuxValue::Setter(class, field.to_string())
};
Ok((AuxImport::Value(val), true))
}
}
decode::OperationKind::IndexingGetter => {
if !structural {
bail!("indexing getters must always be structural");
}
if op.is_static {
Ok((AuxImport::IndexingGetterOfClass(class), false))
} else {
Ok((AuxImport::IndexingGetterOfObject, false))
}
}
decode::OperationKind::IndexingSetter => {
if !structural {
bail!("indexing setters must always be structural");
}
if op.is_static {
Ok((AuxImport::IndexingSetterOfClass(class), false))
} else {
Ok((AuxImport::IndexingSetterOfObject, false))
}
}
decode::OperationKind::IndexingDeleter => {
if !structural {
bail!("indexing deleters must always be structural");
}
if op.is_static {
Ok((AuxImport::IndexingDeleterOfClass(class), false))
} else {
Ok((AuxImport::IndexingDeleterOfObject, false))
}
}
}
}
fn import_static(
&mut self,
import: &decode::Import<'_>,
static_: &decode::ImportStatic<'_>,
) -> Result<(), Error> {
let (import_id, _id) = match self.function_imports.get(static_.shim) {
Some(pair) => *pair,
None => return Ok(()),
};
// Register the signature of this imported shim
Second large refactor for WebIDL bindings This commit is the second, and hopefully last massive, refactor for using WebIDL bindings internally in `wasm-bindgen`. This commit actually fully executes on the task at hand, moving `wasm-bindgen` to internally using WebIDL bindings throughout its code generation, anyref passes, etc. This actually fixes a number of issues that have existed in the anyref pass for some time now! The main changes here are to basically remove the usage of `Descriptor` from generating JS bindings. Instead two new types are introduced: `NonstandardIncoming` and `NonstandardOutgoing` which are bindings lists used for incoming/outgoing bindings. These mirror the standard terminology and literally have variants which are the standard values. All `Descriptor` types are now mapped into lists of incoming/outgoing bindings and used for process in wasm-bindgen. All JS generation has been refactored and updated to now process these lists of bindings instead of the previous `Descriptor`. In other words this commit takes `js2rust.rs` and `rust2js.rs` and first splits them in two. Interpretation of `Descriptor` and what to do for conversions is in the binding selection modules. The actual generation of JS from the binding selection is now performed by `incoming.rs` and `outgoing.rs`. To boot this also deduplicates all the code between the argument handling of `js2rust.rs` and return value handling of `rust2js.rs`. This means that to implement a new binding you only need to implement it one place and it's implemented for free in the other! This commit is not the end of the story though. I would like to add a mdoe to `wasm-bindgen` that literally emits a WebIDL bindings section. That's left for a third (and hopefully final) refactoring which is also intended to optimize generated JS for bindings. This commit currently loses the optimization where an imported is hooked up by value directly whenever a shim isn't needed. It's planned that the next refactoring to emit a webidl binding section that can be added back in. It shouldn't be too too hard hopefully since all the scaffolding is in place now. cc #1524
2019-06-10 07:09:51 -07:00
bindings::register_import(
self.module,
&mut self.bindings,
First refactor for WebIDL bindings This commit starts the `wasm-bindgen` CLI tool down the road to being a true polyfill for WebIDL bindings. This refactor is probably the first of a few, but is hopefully the largest and most sprawling and everything will be a bit more targeted from here on out. The goal of this refactoring is to separate out the massive `crates/cli-support/src/js/mod.rs` into a number of separate pieces of functionality. It currently takes care of basically everything including: * Binding intrinsics * Handling anyref transformations * Generating all JS for imports/exports * All the logic for how to import and how to name imports * Execution and management of wasm-bindgen closures Many of these are separable concerns and most overlap with WebIDL bindings. The internal refactoring here is intended to make it more clear who's responsible for what as well as making some existing operations much more straightforward. At a high-level, the following changes are done: 1. A `src/webidl.rs` module is introduced. The purpose of this module is to take all of the raw wasm-bindgen custom sections from the module and transform them into a WebIDL bindings section. This module has a placeholder `WebidlCustomSection` which is nowhere near the actual custom section but if you squint is in theory very similar. It's hoped that this will eventually become the true WebIDL custom section, currently being developed in an external crate. Currently, however, the WebIDL bindings custom section only covers a subset of the functionality we export to wasm-bindgen users. To avoid leaving them high and dry this module also contains an auxiliary custom section named `WasmBindgenAux`. This custom section isn't intended to have a binary format, but is intended to represent a theoretical custom section necessary to couple with WebIDL bindings to achieve all our desired functionality in `wasm-bindgen`. It'll never be standardized, but it'll also never be serialized :) 2. The `src/webidl.rs` module now takes over quite a bit of functionality from `src/js/mod.rs`. Namely it handles synthesis of an `export_map` and an `import_map` mapping export/import IDs to exactly what's expected to be hooked up there. This does not include type information (as that's in the bindings section) but rather includes things like "this is the method of class A" or "this import is from module `foo`" and things like that. These could arguably be subsumed by future JS features as well, but that's for another time! 3. All handling of wasm-bindgen "descriptor functions" now happens in a dedicated `src/descriptors.rs` module. The output of this module is its own custom section (intended to be immediately consumed by the WebIDL module) which is in theory what we want to ourselves emit one day but rustc isn't capable of doing so right now. 4. Invocations and generations of imports are completely overhauled. Using the `import_map` generated in the WebIDL step all imports are now handled much more precisely in one location rather than haphazardly throughout the module. This means we have precise information about each import of the module and we only modify exactly what we're looking at. This also vastly simplifies intrinsic generation since it's all simply a codegen part of the `rust2js.rs` module now. 5. Handling of direct imports which don't have a JS shim generated is slightly different from before and is intended to be future-compatible with WebIDL bindings in its full glory, but we'll need to update it to handle cases for constructors and method calls eventually as well. 6. Intrinsic definitions now live in their own file (`src/intrinsic.rs`) and have a separated definition for their symbol name and signature. The actual implementation of each intrinsic lives in `rust2js.rs` There's a number of TODO items to finish before this merges. This includes reimplementing the anyref pass and actually implementing import maps for other targets. Those will come soon in follow-up commits, but the entire `tests/wasm/main.rs` suite is currently passing and this seems like a good checkpoint.
2019-05-23 09:15:26 -07:00
import_id,
Second large refactor for WebIDL bindings This commit is the second, and hopefully last massive, refactor for using WebIDL bindings internally in `wasm-bindgen`. This commit actually fully executes on the task at hand, moving `wasm-bindgen` to internally using WebIDL bindings throughout its code generation, anyref passes, etc. This actually fixes a number of issues that have existed in the anyref pass for some time now! The main changes here are to basically remove the usage of `Descriptor` from generating JS bindings. Instead two new types are introduced: `NonstandardIncoming` and `NonstandardOutgoing` which are bindings lists used for incoming/outgoing bindings. These mirror the standard terminology and literally have variants which are the standard values. All `Descriptor` types are now mapped into lists of incoming/outgoing bindings and used for process in wasm-bindgen. All JS generation has been refactored and updated to now process these lists of bindings instead of the previous `Descriptor`. In other words this commit takes `js2rust.rs` and `rust2js.rs` and first splits them in two. Interpretation of `Descriptor` and what to do for conversions is in the binding selection modules. The actual generation of JS from the binding selection is now performed by `incoming.rs` and `outgoing.rs`. To boot this also deduplicates all the code between the argument handling of `js2rust.rs` and return value handling of `rust2js.rs`. This means that to implement a new binding you only need to implement it one place and it's implemented for free in the other! This commit is not the end of the story though. I would like to add a mdoe to `wasm-bindgen` that literally emits a WebIDL bindings section. That's left for a third (and hopefully final) refactoring which is also intended to optimize generated JS for bindings. This commit currently loses the optimization where an imported is hooked up by value directly whenever a shim isn't needed. It's planned that the next refactoring to emit a webidl binding section that can be added back in. It shouldn't be too too hard hopefully since all the scaffolding is in place now. cc #1524
2019-06-10 07:09:51 -07:00
Function {
First refactor for WebIDL bindings This commit starts the `wasm-bindgen` CLI tool down the road to being a true polyfill for WebIDL bindings. This refactor is probably the first of a few, but is hopefully the largest and most sprawling and everything will be a bit more targeted from here on out. The goal of this refactoring is to separate out the massive `crates/cli-support/src/js/mod.rs` into a number of separate pieces of functionality. It currently takes care of basically everything including: * Binding intrinsics * Handling anyref transformations * Generating all JS for imports/exports * All the logic for how to import and how to name imports * Execution and management of wasm-bindgen closures Many of these are separable concerns and most overlap with WebIDL bindings. The internal refactoring here is intended to make it more clear who's responsible for what as well as making some existing operations much more straightforward. At a high-level, the following changes are done: 1. A `src/webidl.rs` module is introduced. The purpose of this module is to take all of the raw wasm-bindgen custom sections from the module and transform them into a WebIDL bindings section. This module has a placeholder `WebidlCustomSection` which is nowhere near the actual custom section but if you squint is in theory very similar. It's hoped that this will eventually become the true WebIDL custom section, currently being developed in an external crate. Currently, however, the WebIDL bindings custom section only covers a subset of the functionality we export to wasm-bindgen users. To avoid leaving them high and dry this module also contains an auxiliary custom section named `WasmBindgenAux`. This custom section isn't intended to have a binary format, but is intended to represent a theoretical custom section necessary to couple with WebIDL bindings to achieve all our desired functionality in `wasm-bindgen`. It'll never be standardized, but it'll also never be serialized :) 2. The `src/webidl.rs` module now takes over quite a bit of functionality from `src/js/mod.rs`. Namely it handles synthesis of an `export_map` and an `import_map` mapping export/import IDs to exactly what's expected to be hooked up there. This does not include type information (as that's in the bindings section) but rather includes things like "this is the method of class A" or "this import is from module `foo`" and things like that. These could arguably be subsumed by future JS features as well, but that's for another time! 3. All handling of wasm-bindgen "descriptor functions" now happens in a dedicated `src/descriptors.rs` module. The output of this module is its own custom section (intended to be immediately consumed by the WebIDL module) which is in theory what we want to ourselves emit one day but rustc isn't capable of doing so right now. 4. Invocations and generations of imports are completely overhauled. Using the `import_map` generated in the WebIDL step all imports are now handled much more precisely in one location rather than haphazardly throughout the module. This means we have precise information about each import of the module and we only modify exactly what we're looking at. This also vastly simplifies intrinsic generation since it's all simply a codegen part of the `rust2js.rs` module now. 5. Handling of direct imports which don't have a JS shim generated is slightly different from before and is intended to be future-compatible with WebIDL bindings in its full glory, but we'll need to update it to handle cases for constructors and method calls eventually as well. 6. Intrinsic definitions now live in their own file (`src/intrinsic.rs`) and have a separated definition for their symbol name and signature. The actual implementation of each intrinsic lives in `rust2js.rs` There's a number of TODO items to finish before this merges. This includes reimplementing the anyref pass and actually implementing import maps for other targets. Those will come soon in follow-up commits, but the entire `tests/wasm/main.rs` suite is currently passing and this seems like a good checkpoint.
2019-05-23 09:15:26 -07:00
arguments: Vec::new(),
shim_idx: 0,
ret: Descriptor::Anyref,
Second large refactor for WebIDL bindings This commit is the second, and hopefully last massive, refactor for using WebIDL bindings internally in `wasm-bindgen`. This commit actually fully executes on the task at hand, moving `wasm-bindgen` to internally using WebIDL bindings throughout its code generation, anyref passes, etc. This actually fixes a number of issues that have existed in the anyref pass for some time now! The main changes here are to basically remove the usage of `Descriptor` from generating JS bindings. Instead two new types are introduced: `NonstandardIncoming` and `NonstandardOutgoing` which are bindings lists used for incoming/outgoing bindings. These mirror the standard terminology and literally have variants which are the standard values. All `Descriptor` types are now mapped into lists of incoming/outgoing bindings and used for process in wasm-bindgen. All JS generation has been refactored and updated to now process these lists of bindings instead of the previous `Descriptor`. In other words this commit takes `js2rust.rs` and `rust2js.rs` and first splits them in two. Interpretation of `Descriptor` and what to do for conversions is in the binding selection modules. The actual generation of JS from the binding selection is now performed by `incoming.rs` and `outgoing.rs`. To boot this also deduplicates all the code between the argument handling of `js2rust.rs` and return value handling of `rust2js.rs`. This means that to implement a new binding you only need to implement it one place and it's implemented for free in the other! This commit is not the end of the story though. I would like to add a mdoe to `wasm-bindgen` that literally emits a WebIDL bindings section. That's left for a third (and hopefully final) refactoring which is also intended to optimize generated JS for bindings. This commit currently loses the optimization where an imported is hooked up by value directly whenever a shim isn't needed. It's planned that the next refactoring to emit a webidl binding section that can be added back in. It shouldn't be too too hard hopefully since all the scaffolding is in place now. cc #1524
2019-06-10 07:09:51 -07:00
},
ast::WebidlFunctionKind::Static,
)?;
First refactor for WebIDL bindings This commit starts the `wasm-bindgen` CLI tool down the road to being a true polyfill for WebIDL bindings. This refactor is probably the first of a few, but is hopefully the largest and most sprawling and everything will be a bit more targeted from here on out. The goal of this refactoring is to separate out the massive `crates/cli-support/src/js/mod.rs` into a number of separate pieces of functionality. It currently takes care of basically everything including: * Binding intrinsics * Handling anyref transformations * Generating all JS for imports/exports * All the logic for how to import and how to name imports * Execution and management of wasm-bindgen closures Many of these are separable concerns and most overlap with WebIDL bindings. The internal refactoring here is intended to make it more clear who's responsible for what as well as making some existing operations much more straightforward. At a high-level, the following changes are done: 1. A `src/webidl.rs` module is introduced. The purpose of this module is to take all of the raw wasm-bindgen custom sections from the module and transform them into a WebIDL bindings section. This module has a placeholder `WebidlCustomSection` which is nowhere near the actual custom section but if you squint is in theory very similar. It's hoped that this will eventually become the true WebIDL custom section, currently being developed in an external crate. Currently, however, the WebIDL bindings custom section only covers a subset of the functionality we export to wasm-bindgen users. To avoid leaving them high and dry this module also contains an auxiliary custom section named `WasmBindgenAux`. This custom section isn't intended to have a binary format, but is intended to represent a theoretical custom section necessary to couple with WebIDL bindings to achieve all our desired functionality in `wasm-bindgen`. It'll never be standardized, but it'll also never be serialized :) 2. The `src/webidl.rs` module now takes over quite a bit of functionality from `src/js/mod.rs`. Namely it handles synthesis of an `export_map` and an `import_map` mapping export/import IDs to exactly what's expected to be hooked up there. This does not include type information (as that's in the bindings section) but rather includes things like "this is the method of class A" or "this import is from module `foo`" and things like that. These could arguably be subsumed by future JS features as well, but that's for another time! 3. All handling of wasm-bindgen "descriptor functions" now happens in a dedicated `src/descriptors.rs` module. The output of this module is its own custom section (intended to be immediately consumed by the WebIDL module) which is in theory what we want to ourselves emit one day but rustc isn't capable of doing so right now. 4. Invocations and generations of imports are completely overhauled. Using the `import_map` generated in the WebIDL step all imports are now handled much more precisely in one location rather than haphazardly throughout the module. This means we have precise information about each import of the module and we only modify exactly what we're looking at. This also vastly simplifies intrinsic generation since it's all simply a codegen part of the `rust2js.rs` module now. 5. Handling of direct imports which don't have a JS shim generated is slightly different from before and is intended to be future-compatible with WebIDL bindings in its full glory, but we'll need to update it to handle cases for constructors and method calls eventually as well. 6. Intrinsic definitions now live in their own file (`src/intrinsic.rs`) and have a separated definition for their symbol name and signature. The actual implementation of each intrinsic lives in `rust2js.rs` There's a number of TODO items to finish before this merges. This includes reimplementing the anyref pass and actually implementing import maps for other targets. Those will come soon in follow-up commits, but the entire `tests/wasm/main.rs` suite is currently passing and this seems like a good checkpoint.
2019-05-23 09:15:26 -07:00
// And then save off that this function is is an instanceof shim for an
// imported item.
let import = self.determine_import(import, &static_.name)?;
self.aux
.import_map
.insert(import_id, AuxImport::Static(import));
Ok(())
}
fn import_type(
&mut self,
import: &decode::Import<'_>,
type_: &decode::ImportType<'_>,
) -> Result<(), Error> {
let (import_id, _id) = match self.function_imports.get(type_.instanceof_shim) {
Some(pair) => *pair,
None => return Ok(()),
};
// Register the signature of this imported shim
Second large refactor for WebIDL bindings This commit is the second, and hopefully last massive, refactor for using WebIDL bindings internally in `wasm-bindgen`. This commit actually fully executes on the task at hand, moving `wasm-bindgen` to internally using WebIDL bindings throughout its code generation, anyref passes, etc. This actually fixes a number of issues that have existed in the anyref pass for some time now! The main changes here are to basically remove the usage of `Descriptor` from generating JS bindings. Instead two new types are introduced: `NonstandardIncoming` and `NonstandardOutgoing` which are bindings lists used for incoming/outgoing bindings. These mirror the standard terminology and literally have variants which are the standard values. All `Descriptor` types are now mapped into lists of incoming/outgoing bindings and used for process in wasm-bindgen. All JS generation has been refactored and updated to now process these lists of bindings instead of the previous `Descriptor`. In other words this commit takes `js2rust.rs` and `rust2js.rs` and first splits them in two. Interpretation of `Descriptor` and what to do for conversions is in the binding selection modules. The actual generation of JS from the binding selection is now performed by `incoming.rs` and `outgoing.rs`. To boot this also deduplicates all the code between the argument handling of `js2rust.rs` and return value handling of `rust2js.rs`. This means that to implement a new binding you only need to implement it one place and it's implemented for free in the other! This commit is not the end of the story though. I would like to add a mdoe to `wasm-bindgen` that literally emits a WebIDL bindings section. That's left for a third (and hopefully final) refactoring which is also intended to optimize generated JS for bindings. This commit currently loses the optimization where an imported is hooked up by value directly whenever a shim isn't needed. It's planned that the next refactoring to emit a webidl binding section that can be added back in. It shouldn't be too too hard hopefully since all the scaffolding is in place now. cc #1524
2019-06-10 07:09:51 -07:00
bindings::register_import(
self.module,
&mut self.bindings,
First refactor for WebIDL bindings This commit starts the `wasm-bindgen` CLI tool down the road to being a true polyfill for WebIDL bindings. This refactor is probably the first of a few, but is hopefully the largest and most sprawling and everything will be a bit more targeted from here on out. The goal of this refactoring is to separate out the massive `crates/cli-support/src/js/mod.rs` into a number of separate pieces of functionality. It currently takes care of basically everything including: * Binding intrinsics * Handling anyref transformations * Generating all JS for imports/exports * All the logic for how to import and how to name imports * Execution and management of wasm-bindgen closures Many of these are separable concerns and most overlap with WebIDL bindings. The internal refactoring here is intended to make it more clear who's responsible for what as well as making some existing operations much more straightforward. At a high-level, the following changes are done: 1. A `src/webidl.rs` module is introduced. The purpose of this module is to take all of the raw wasm-bindgen custom sections from the module and transform them into a WebIDL bindings section. This module has a placeholder `WebidlCustomSection` which is nowhere near the actual custom section but if you squint is in theory very similar. It's hoped that this will eventually become the true WebIDL custom section, currently being developed in an external crate. Currently, however, the WebIDL bindings custom section only covers a subset of the functionality we export to wasm-bindgen users. To avoid leaving them high and dry this module also contains an auxiliary custom section named `WasmBindgenAux`. This custom section isn't intended to have a binary format, but is intended to represent a theoretical custom section necessary to couple with WebIDL bindings to achieve all our desired functionality in `wasm-bindgen`. It'll never be standardized, but it'll also never be serialized :) 2. The `src/webidl.rs` module now takes over quite a bit of functionality from `src/js/mod.rs`. Namely it handles synthesis of an `export_map` and an `import_map` mapping export/import IDs to exactly what's expected to be hooked up there. This does not include type information (as that's in the bindings section) but rather includes things like "this is the method of class A" or "this import is from module `foo`" and things like that. These could arguably be subsumed by future JS features as well, but that's for another time! 3. All handling of wasm-bindgen "descriptor functions" now happens in a dedicated `src/descriptors.rs` module. The output of this module is its own custom section (intended to be immediately consumed by the WebIDL module) which is in theory what we want to ourselves emit one day but rustc isn't capable of doing so right now. 4. Invocations and generations of imports are completely overhauled. Using the `import_map` generated in the WebIDL step all imports are now handled much more precisely in one location rather than haphazardly throughout the module. This means we have precise information about each import of the module and we only modify exactly what we're looking at. This also vastly simplifies intrinsic generation since it's all simply a codegen part of the `rust2js.rs` module now. 5. Handling of direct imports which don't have a JS shim generated is slightly different from before and is intended to be future-compatible with WebIDL bindings in its full glory, but we'll need to update it to handle cases for constructors and method calls eventually as well. 6. Intrinsic definitions now live in their own file (`src/intrinsic.rs`) and have a separated definition for their symbol name and signature. The actual implementation of each intrinsic lives in `rust2js.rs` There's a number of TODO items to finish before this merges. This includes reimplementing the anyref pass and actually implementing import maps for other targets. Those will come soon in follow-up commits, but the entire `tests/wasm/main.rs` suite is currently passing and this seems like a good checkpoint.
2019-05-23 09:15:26 -07:00
import_id,
Second large refactor for WebIDL bindings This commit is the second, and hopefully last massive, refactor for using WebIDL bindings internally in `wasm-bindgen`. This commit actually fully executes on the task at hand, moving `wasm-bindgen` to internally using WebIDL bindings throughout its code generation, anyref passes, etc. This actually fixes a number of issues that have existed in the anyref pass for some time now! The main changes here are to basically remove the usage of `Descriptor` from generating JS bindings. Instead two new types are introduced: `NonstandardIncoming` and `NonstandardOutgoing` which are bindings lists used for incoming/outgoing bindings. These mirror the standard terminology and literally have variants which are the standard values. All `Descriptor` types are now mapped into lists of incoming/outgoing bindings and used for process in wasm-bindgen. All JS generation has been refactored and updated to now process these lists of bindings instead of the previous `Descriptor`. In other words this commit takes `js2rust.rs` and `rust2js.rs` and first splits them in two. Interpretation of `Descriptor` and what to do for conversions is in the binding selection modules. The actual generation of JS from the binding selection is now performed by `incoming.rs` and `outgoing.rs`. To boot this also deduplicates all the code between the argument handling of `js2rust.rs` and return value handling of `rust2js.rs`. This means that to implement a new binding you only need to implement it one place and it's implemented for free in the other! This commit is not the end of the story though. I would like to add a mdoe to `wasm-bindgen` that literally emits a WebIDL bindings section. That's left for a third (and hopefully final) refactoring which is also intended to optimize generated JS for bindings. This commit currently loses the optimization where an imported is hooked up by value directly whenever a shim isn't needed. It's planned that the next refactoring to emit a webidl binding section that can be added back in. It shouldn't be too too hard hopefully since all the scaffolding is in place now. cc #1524
2019-06-10 07:09:51 -07:00
Function {
First refactor for WebIDL bindings This commit starts the `wasm-bindgen` CLI tool down the road to being a true polyfill for WebIDL bindings. This refactor is probably the first of a few, but is hopefully the largest and most sprawling and everything will be a bit more targeted from here on out. The goal of this refactoring is to separate out the massive `crates/cli-support/src/js/mod.rs` into a number of separate pieces of functionality. It currently takes care of basically everything including: * Binding intrinsics * Handling anyref transformations * Generating all JS for imports/exports * All the logic for how to import and how to name imports * Execution and management of wasm-bindgen closures Many of these are separable concerns and most overlap with WebIDL bindings. The internal refactoring here is intended to make it more clear who's responsible for what as well as making some existing operations much more straightforward. At a high-level, the following changes are done: 1. A `src/webidl.rs` module is introduced. The purpose of this module is to take all of the raw wasm-bindgen custom sections from the module and transform them into a WebIDL bindings section. This module has a placeholder `WebidlCustomSection` which is nowhere near the actual custom section but if you squint is in theory very similar. It's hoped that this will eventually become the true WebIDL custom section, currently being developed in an external crate. Currently, however, the WebIDL bindings custom section only covers a subset of the functionality we export to wasm-bindgen users. To avoid leaving them high and dry this module also contains an auxiliary custom section named `WasmBindgenAux`. This custom section isn't intended to have a binary format, but is intended to represent a theoretical custom section necessary to couple with WebIDL bindings to achieve all our desired functionality in `wasm-bindgen`. It'll never be standardized, but it'll also never be serialized :) 2. The `src/webidl.rs` module now takes over quite a bit of functionality from `src/js/mod.rs`. Namely it handles synthesis of an `export_map` and an `import_map` mapping export/import IDs to exactly what's expected to be hooked up there. This does not include type information (as that's in the bindings section) but rather includes things like "this is the method of class A" or "this import is from module `foo`" and things like that. These could arguably be subsumed by future JS features as well, but that's for another time! 3. All handling of wasm-bindgen "descriptor functions" now happens in a dedicated `src/descriptors.rs` module. The output of this module is its own custom section (intended to be immediately consumed by the WebIDL module) which is in theory what we want to ourselves emit one day but rustc isn't capable of doing so right now. 4. Invocations and generations of imports are completely overhauled. Using the `import_map` generated in the WebIDL step all imports are now handled much more precisely in one location rather than haphazardly throughout the module. This means we have precise information about each import of the module and we only modify exactly what we're looking at. This also vastly simplifies intrinsic generation since it's all simply a codegen part of the `rust2js.rs` module now. 5. Handling of direct imports which don't have a JS shim generated is slightly different from before and is intended to be future-compatible with WebIDL bindings in its full glory, but we'll need to update it to handle cases for constructors and method calls eventually as well. 6. Intrinsic definitions now live in their own file (`src/intrinsic.rs`) and have a separated definition for their symbol name and signature. The actual implementation of each intrinsic lives in `rust2js.rs` There's a number of TODO items to finish before this merges. This includes reimplementing the anyref pass and actually implementing import maps for other targets. Those will come soon in follow-up commits, but the entire `tests/wasm/main.rs` suite is currently passing and this seems like a good checkpoint.
2019-05-23 09:15:26 -07:00
arguments: vec![Descriptor::Ref(Box::new(Descriptor::Anyref))],
shim_idx: 0,
Second large refactor for WebIDL bindings This commit is the second, and hopefully last massive, refactor for using WebIDL bindings internally in `wasm-bindgen`. This commit actually fully executes on the task at hand, moving `wasm-bindgen` to internally using WebIDL bindings throughout its code generation, anyref passes, etc. This actually fixes a number of issues that have existed in the anyref pass for some time now! The main changes here are to basically remove the usage of `Descriptor` from generating JS bindings. Instead two new types are introduced: `NonstandardIncoming` and `NonstandardOutgoing` which are bindings lists used for incoming/outgoing bindings. These mirror the standard terminology and literally have variants which are the standard values. All `Descriptor` types are now mapped into lists of incoming/outgoing bindings and used for process in wasm-bindgen. All JS generation has been refactored and updated to now process these lists of bindings instead of the previous `Descriptor`. In other words this commit takes `js2rust.rs` and `rust2js.rs` and first splits them in two. Interpretation of `Descriptor` and what to do for conversions is in the binding selection modules. The actual generation of JS from the binding selection is now performed by `incoming.rs` and `outgoing.rs`. To boot this also deduplicates all the code between the argument handling of `js2rust.rs` and return value handling of `rust2js.rs`. This means that to implement a new binding you only need to implement it one place and it's implemented for free in the other! This commit is not the end of the story though. I would like to add a mdoe to `wasm-bindgen` that literally emits a WebIDL bindings section. That's left for a third (and hopefully final) refactoring which is also intended to optimize generated JS for bindings. This commit currently loses the optimization where an imported is hooked up by value directly whenever a shim isn't needed. It's planned that the next refactoring to emit a webidl binding section that can be added back in. It shouldn't be too too hard hopefully since all the scaffolding is in place now. cc #1524
2019-06-10 07:09:51 -07:00
ret: Descriptor::Boolean,
},
ast::WebidlFunctionKind::Static,
)?;
First refactor for WebIDL bindings This commit starts the `wasm-bindgen` CLI tool down the road to being a true polyfill for WebIDL bindings. This refactor is probably the first of a few, but is hopefully the largest and most sprawling and everything will be a bit more targeted from here on out. The goal of this refactoring is to separate out the massive `crates/cli-support/src/js/mod.rs` into a number of separate pieces of functionality. It currently takes care of basically everything including: * Binding intrinsics * Handling anyref transformations * Generating all JS for imports/exports * All the logic for how to import and how to name imports * Execution and management of wasm-bindgen closures Many of these are separable concerns and most overlap with WebIDL bindings. The internal refactoring here is intended to make it more clear who's responsible for what as well as making some existing operations much more straightforward. At a high-level, the following changes are done: 1. A `src/webidl.rs` module is introduced. The purpose of this module is to take all of the raw wasm-bindgen custom sections from the module and transform them into a WebIDL bindings section. This module has a placeholder `WebidlCustomSection` which is nowhere near the actual custom section but if you squint is in theory very similar. It's hoped that this will eventually become the true WebIDL custom section, currently being developed in an external crate. Currently, however, the WebIDL bindings custom section only covers a subset of the functionality we export to wasm-bindgen users. To avoid leaving them high and dry this module also contains an auxiliary custom section named `WasmBindgenAux`. This custom section isn't intended to have a binary format, but is intended to represent a theoretical custom section necessary to couple with WebIDL bindings to achieve all our desired functionality in `wasm-bindgen`. It'll never be standardized, but it'll also never be serialized :) 2. The `src/webidl.rs` module now takes over quite a bit of functionality from `src/js/mod.rs`. Namely it handles synthesis of an `export_map` and an `import_map` mapping export/import IDs to exactly what's expected to be hooked up there. This does not include type information (as that's in the bindings section) but rather includes things like "this is the method of class A" or "this import is from module `foo`" and things like that. These could arguably be subsumed by future JS features as well, but that's for another time! 3. All handling of wasm-bindgen "descriptor functions" now happens in a dedicated `src/descriptors.rs` module. The output of this module is its own custom section (intended to be immediately consumed by the WebIDL module) which is in theory what we want to ourselves emit one day but rustc isn't capable of doing so right now. 4. Invocations and generations of imports are completely overhauled. Using the `import_map` generated in the WebIDL step all imports are now handled much more precisely in one location rather than haphazardly throughout the module. This means we have precise information about each import of the module and we only modify exactly what we're looking at. This also vastly simplifies intrinsic generation since it's all simply a codegen part of the `rust2js.rs` module now. 5. Handling of direct imports which don't have a JS shim generated is slightly different from before and is intended to be future-compatible with WebIDL bindings in its full glory, but we'll need to update it to handle cases for constructors and method calls eventually as well. 6. Intrinsic definitions now live in their own file (`src/intrinsic.rs`) and have a separated definition for their symbol name and signature. The actual implementation of each intrinsic lives in `rust2js.rs` There's a number of TODO items to finish before this merges. This includes reimplementing the anyref pass and actually implementing import maps for other targets. Those will come soon in follow-up commits, but the entire `tests/wasm/main.rs` suite is currently passing and this seems like a good checkpoint.
2019-05-23 09:15:26 -07:00
// And then save off that this function is is an instanceof shim for an
// imported item.
let import = self.determine_import(import, &type_.name)?;
self.aux
.import_map
.insert(import_id, AuxImport::Instanceof(import));
Ok(())
}
fn enum_(&mut self, enum_: decode::Enum<'_>) -> Result<(), Error> {
let aux = AuxEnum {
name: enum_.name.to_string(),
comments: concatenate_comments(&enum_.comments),
variants: enum_
.variants
.iter()
.map(|v| (v.name.to_string(), v.value))
.collect(),
};
self.aux.enums.push(aux);
Ok(())
}
fn struct_(&mut self, struct_: decode::Struct<'_>) -> Result<(), Error> {
for field in struct_.fields {
let getter = wasm_bindgen_shared::struct_field_get(&struct_.name, &field.name);
let setter = wasm_bindgen_shared::struct_field_set(&struct_.name, &field.name);
let descriptor = match self.descriptors.remove(&getter) {
None => continue,
Some(d) => d,
};
// Register a webidl transformation for the getter
let (getter_id, _) = self.function_exports[&getter];
let getter_descriptor = Function {
arguments: vec![Descriptor::I32],
First refactor for WebIDL bindings This commit starts the `wasm-bindgen` CLI tool down the road to being a true polyfill for WebIDL bindings. This refactor is probably the first of a few, but is hopefully the largest and most sprawling and everything will be a bit more targeted from here on out. The goal of this refactoring is to separate out the massive `crates/cli-support/src/js/mod.rs` into a number of separate pieces of functionality. It currently takes care of basically everything including: * Binding intrinsics * Handling anyref transformations * Generating all JS for imports/exports * All the logic for how to import and how to name imports * Execution and management of wasm-bindgen closures Many of these are separable concerns and most overlap with WebIDL bindings. The internal refactoring here is intended to make it more clear who's responsible for what as well as making some existing operations much more straightforward. At a high-level, the following changes are done: 1. A `src/webidl.rs` module is introduced. The purpose of this module is to take all of the raw wasm-bindgen custom sections from the module and transform them into a WebIDL bindings section. This module has a placeholder `WebidlCustomSection` which is nowhere near the actual custom section but if you squint is in theory very similar. It's hoped that this will eventually become the true WebIDL custom section, currently being developed in an external crate. Currently, however, the WebIDL bindings custom section only covers a subset of the functionality we export to wasm-bindgen users. To avoid leaving them high and dry this module also contains an auxiliary custom section named `WasmBindgenAux`. This custom section isn't intended to have a binary format, but is intended to represent a theoretical custom section necessary to couple with WebIDL bindings to achieve all our desired functionality in `wasm-bindgen`. It'll never be standardized, but it'll also never be serialized :) 2. The `src/webidl.rs` module now takes over quite a bit of functionality from `src/js/mod.rs`. Namely it handles synthesis of an `export_map` and an `import_map` mapping export/import IDs to exactly what's expected to be hooked up there. This does not include type information (as that's in the bindings section) but rather includes things like "this is the method of class A" or "this import is from module `foo`" and things like that. These could arguably be subsumed by future JS features as well, but that's for another time! 3. All handling of wasm-bindgen "descriptor functions" now happens in a dedicated `src/descriptors.rs` module. The output of this module is its own custom section (intended to be immediately consumed by the WebIDL module) which is in theory what we want to ourselves emit one day but rustc isn't capable of doing so right now. 4. Invocations and generations of imports are completely overhauled. Using the `import_map` generated in the WebIDL step all imports are now handled much more precisely in one location rather than haphazardly throughout the module. This means we have precise information about each import of the module and we only modify exactly what we're looking at. This also vastly simplifies intrinsic generation since it's all simply a codegen part of the `rust2js.rs` module now. 5. Handling of direct imports which don't have a JS shim generated is slightly different from before and is intended to be future-compatible with WebIDL bindings in its full glory, but we'll need to update it to handle cases for constructors and method calls eventually as well. 6. Intrinsic definitions now live in their own file (`src/intrinsic.rs`) and have a separated definition for their symbol name and signature. The actual implementation of each intrinsic lives in `rust2js.rs` There's a number of TODO items to finish before this merges. This includes reimplementing the anyref pass and actually implementing import maps for other targets. Those will come soon in follow-up commits, but the entire `tests/wasm/main.rs` suite is currently passing and this seems like a good checkpoint.
2019-05-23 09:15:26 -07:00
shim_idx: 0,
ret: descriptor.clone(),
};
Second large refactor for WebIDL bindings This commit is the second, and hopefully last massive, refactor for using WebIDL bindings internally in `wasm-bindgen`. This commit actually fully executes on the task at hand, moving `wasm-bindgen` to internally using WebIDL bindings throughout its code generation, anyref passes, etc. This actually fixes a number of issues that have existed in the anyref pass for some time now! The main changes here are to basically remove the usage of `Descriptor` from generating JS bindings. Instead two new types are introduced: `NonstandardIncoming` and `NonstandardOutgoing` which are bindings lists used for incoming/outgoing bindings. These mirror the standard terminology and literally have variants which are the standard values. All `Descriptor` types are now mapped into lists of incoming/outgoing bindings and used for process in wasm-bindgen. All JS generation has been refactored and updated to now process these lists of bindings instead of the previous `Descriptor`. In other words this commit takes `js2rust.rs` and `rust2js.rs` and first splits them in two. Interpretation of `Descriptor` and what to do for conversions is in the binding selection modules. The actual generation of JS from the binding selection is now performed by `incoming.rs` and `outgoing.rs`. To boot this also deduplicates all the code between the argument handling of `js2rust.rs` and return value handling of `rust2js.rs`. This means that to implement a new binding you only need to implement it one place and it's implemented for free in the other! This commit is not the end of the story though. I would like to add a mdoe to `wasm-bindgen` that literally emits a WebIDL bindings section. That's left for a third (and hopefully final) refactoring which is also intended to optimize generated JS for bindings. This commit currently loses the optimization where an imported is hooked up by value directly whenever a shim isn't needed. It's planned that the next refactoring to emit a webidl binding section that can be added back in. It shouldn't be too too hard hopefully since all the scaffolding is in place now. cc #1524
2019-06-10 07:09:51 -07:00
bindings::register_export(
self.module,
&mut self.bindings,
getter_id,
getter_descriptor,
)?;
First refactor for WebIDL bindings This commit starts the `wasm-bindgen` CLI tool down the road to being a true polyfill for WebIDL bindings. This refactor is probably the first of a few, but is hopefully the largest and most sprawling and everything will be a bit more targeted from here on out. The goal of this refactoring is to separate out the massive `crates/cli-support/src/js/mod.rs` into a number of separate pieces of functionality. It currently takes care of basically everything including: * Binding intrinsics * Handling anyref transformations * Generating all JS for imports/exports * All the logic for how to import and how to name imports * Execution and management of wasm-bindgen closures Many of these are separable concerns and most overlap with WebIDL bindings. The internal refactoring here is intended to make it more clear who's responsible for what as well as making some existing operations much more straightforward. At a high-level, the following changes are done: 1. A `src/webidl.rs` module is introduced. The purpose of this module is to take all of the raw wasm-bindgen custom sections from the module and transform them into a WebIDL bindings section. This module has a placeholder `WebidlCustomSection` which is nowhere near the actual custom section but if you squint is in theory very similar. It's hoped that this will eventually become the true WebIDL custom section, currently being developed in an external crate. Currently, however, the WebIDL bindings custom section only covers a subset of the functionality we export to wasm-bindgen users. To avoid leaving them high and dry this module also contains an auxiliary custom section named `WasmBindgenAux`. This custom section isn't intended to have a binary format, but is intended to represent a theoretical custom section necessary to couple with WebIDL bindings to achieve all our desired functionality in `wasm-bindgen`. It'll never be standardized, but it'll also never be serialized :) 2. The `src/webidl.rs` module now takes over quite a bit of functionality from `src/js/mod.rs`. Namely it handles synthesis of an `export_map` and an `import_map` mapping export/import IDs to exactly what's expected to be hooked up there. This does not include type information (as that's in the bindings section) but rather includes things like "this is the method of class A" or "this import is from module `foo`" and things like that. These could arguably be subsumed by future JS features as well, but that's for another time! 3. All handling of wasm-bindgen "descriptor functions" now happens in a dedicated `src/descriptors.rs` module. The output of this module is its own custom section (intended to be immediately consumed by the WebIDL module) which is in theory what we want to ourselves emit one day but rustc isn't capable of doing so right now. 4. Invocations and generations of imports are completely overhauled. Using the `import_map` generated in the WebIDL step all imports are now handled much more precisely in one location rather than haphazardly throughout the module. This means we have precise information about each import of the module and we only modify exactly what we're looking at. This also vastly simplifies intrinsic generation since it's all simply a codegen part of the `rust2js.rs` module now. 5. Handling of direct imports which don't have a JS shim generated is slightly different from before and is intended to be future-compatible with WebIDL bindings in its full glory, but we'll need to update it to handle cases for constructors and method calls eventually as well. 6. Intrinsic definitions now live in their own file (`src/intrinsic.rs`) and have a separated definition for their symbol name and signature. The actual implementation of each intrinsic lives in `rust2js.rs` There's a number of TODO items to finish before this merges. This includes reimplementing the anyref pass and actually implementing import maps for other targets. Those will come soon in follow-up commits, but the entire `tests/wasm/main.rs` suite is currently passing and this seems like a good checkpoint.
2019-05-23 09:15:26 -07:00
self.aux.export_map.insert(
getter_id,
AuxExport {
debug_name: format!("getter for `{}::{}`", struct_.name, field.name),
arg_names: None,
comments: concatenate_comments(&field.comments),
kind: AuxExportKind::Getter {
class: struct_.name.to_string(),
field: field.name.to_string(),
},
},
);
// If present, register information for the setter as well.
if field.readonly {
continue;
}
let (setter_id, _) = self.function_exports[&setter];
let setter_descriptor = Function {
arguments: vec![Descriptor::I32, descriptor],
First refactor for WebIDL bindings This commit starts the `wasm-bindgen` CLI tool down the road to being a true polyfill for WebIDL bindings. This refactor is probably the first of a few, but is hopefully the largest and most sprawling and everything will be a bit more targeted from here on out. The goal of this refactoring is to separate out the massive `crates/cli-support/src/js/mod.rs` into a number of separate pieces of functionality. It currently takes care of basically everything including: * Binding intrinsics * Handling anyref transformations * Generating all JS for imports/exports * All the logic for how to import and how to name imports * Execution and management of wasm-bindgen closures Many of these are separable concerns and most overlap with WebIDL bindings. The internal refactoring here is intended to make it more clear who's responsible for what as well as making some existing operations much more straightforward. At a high-level, the following changes are done: 1. A `src/webidl.rs` module is introduced. The purpose of this module is to take all of the raw wasm-bindgen custom sections from the module and transform them into a WebIDL bindings section. This module has a placeholder `WebidlCustomSection` which is nowhere near the actual custom section but if you squint is in theory very similar. It's hoped that this will eventually become the true WebIDL custom section, currently being developed in an external crate. Currently, however, the WebIDL bindings custom section only covers a subset of the functionality we export to wasm-bindgen users. To avoid leaving them high and dry this module also contains an auxiliary custom section named `WasmBindgenAux`. This custom section isn't intended to have a binary format, but is intended to represent a theoretical custom section necessary to couple with WebIDL bindings to achieve all our desired functionality in `wasm-bindgen`. It'll never be standardized, but it'll also never be serialized :) 2. The `src/webidl.rs` module now takes over quite a bit of functionality from `src/js/mod.rs`. Namely it handles synthesis of an `export_map` and an `import_map` mapping export/import IDs to exactly what's expected to be hooked up there. This does not include type information (as that's in the bindings section) but rather includes things like "this is the method of class A" or "this import is from module `foo`" and things like that. These could arguably be subsumed by future JS features as well, but that's for another time! 3. All handling of wasm-bindgen "descriptor functions" now happens in a dedicated `src/descriptors.rs` module. The output of this module is its own custom section (intended to be immediately consumed by the WebIDL module) which is in theory what we want to ourselves emit one day but rustc isn't capable of doing so right now. 4. Invocations and generations of imports are completely overhauled. Using the `import_map` generated in the WebIDL step all imports are now handled much more precisely in one location rather than haphazardly throughout the module. This means we have precise information about each import of the module and we only modify exactly what we're looking at. This also vastly simplifies intrinsic generation since it's all simply a codegen part of the `rust2js.rs` module now. 5. Handling of direct imports which don't have a JS shim generated is slightly different from before and is intended to be future-compatible with WebIDL bindings in its full glory, but we'll need to update it to handle cases for constructors and method calls eventually as well. 6. Intrinsic definitions now live in their own file (`src/intrinsic.rs`) and have a separated definition for their symbol name and signature. The actual implementation of each intrinsic lives in `rust2js.rs` There's a number of TODO items to finish before this merges. This includes reimplementing the anyref pass and actually implementing import maps for other targets. Those will come soon in follow-up commits, but the entire `tests/wasm/main.rs` suite is currently passing and this seems like a good checkpoint.
2019-05-23 09:15:26 -07:00
shim_idx: 0,
ret: Descriptor::Unit,
};
Second large refactor for WebIDL bindings This commit is the second, and hopefully last massive, refactor for using WebIDL bindings internally in `wasm-bindgen`. This commit actually fully executes on the task at hand, moving `wasm-bindgen` to internally using WebIDL bindings throughout its code generation, anyref passes, etc. This actually fixes a number of issues that have existed in the anyref pass for some time now! The main changes here are to basically remove the usage of `Descriptor` from generating JS bindings. Instead two new types are introduced: `NonstandardIncoming` and `NonstandardOutgoing` which are bindings lists used for incoming/outgoing bindings. These mirror the standard terminology and literally have variants which are the standard values. All `Descriptor` types are now mapped into lists of incoming/outgoing bindings and used for process in wasm-bindgen. All JS generation has been refactored and updated to now process these lists of bindings instead of the previous `Descriptor`. In other words this commit takes `js2rust.rs` and `rust2js.rs` and first splits them in two. Interpretation of `Descriptor` and what to do for conversions is in the binding selection modules. The actual generation of JS from the binding selection is now performed by `incoming.rs` and `outgoing.rs`. To boot this also deduplicates all the code between the argument handling of `js2rust.rs` and return value handling of `rust2js.rs`. This means that to implement a new binding you only need to implement it one place and it's implemented for free in the other! This commit is not the end of the story though. I would like to add a mdoe to `wasm-bindgen` that literally emits a WebIDL bindings section. That's left for a third (and hopefully final) refactoring which is also intended to optimize generated JS for bindings. This commit currently loses the optimization where an imported is hooked up by value directly whenever a shim isn't needed. It's planned that the next refactoring to emit a webidl binding section that can be added back in. It shouldn't be too too hard hopefully since all the scaffolding is in place now. cc #1524
2019-06-10 07:09:51 -07:00
bindings::register_export(
self.module,
&mut self.bindings,
setter_id,
setter_descriptor,
)?;
First refactor for WebIDL bindings This commit starts the `wasm-bindgen` CLI tool down the road to being a true polyfill for WebIDL bindings. This refactor is probably the first of a few, but is hopefully the largest and most sprawling and everything will be a bit more targeted from here on out. The goal of this refactoring is to separate out the massive `crates/cli-support/src/js/mod.rs` into a number of separate pieces of functionality. It currently takes care of basically everything including: * Binding intrinsics * Handling anyref transformations * Generating all JS for imports/exports * All the logic for how to import and how to name imports * Execution and management of wasm-bindgen closures Many of these are separable concerns and most overlap with WebIDL bindings. The internal refactoring here is intended to make it more clear who's responsible for what as well as making some existing operations much more straightforward. At a high-level, the following changes are done: 1. A `src/webidl.rs` module is introduced. The purpose of this module is to take all of the raw wasm-bindgen custom sections from the module and transform them into a WebIDL bindings section. This module has a placeholder `WebidlCustomSection` which is nowhere near the actual custom section but if you squint is in theory very similar. It's hoped that this will eventually become the true WebIDL custom section, currently being developed in an external crate. Currently, however, the WebIDL bindings custom section only covers a subset of the functionality we export to wasm-bindgen users. To avoid leaving them high and dry this module also contains an auxiliary custom section named `WasmBindgenAux`. This custom section isn't intended to have a binary format, but is intended to represent a theoretical custom section necessary to couple with WebIDL bindings to achieve all our desired functionality in `wasm-bindgen`. It'll never be standardized, but it'll also never be serialized :) 2. The `src/webidl.rs` module now takes over quite a bit of functionality from `src/js/mod.rs`. Namely it handles synthesis of an `export_map` and an `import_map` mapping export/import IDs to exactly what's expected to be hooked up there. This does not include type information (as that's in the bindings section) but rather includes things like "this is the method of class A" or "this import is from module `foo`" and things like that. These could arguably be subsumed by future JS features as well, but that's for another time! 3. All handling of wasm-bindgen "descriptor functions" now happens in a dedicated `src/descriptors.rs` module. The output of this module is its own custom section (intended to be immediately consumed by the WebIDL module) which is in theory what we want to ourselves emit one day but rustc isn't capable of doing so right now. 4. Invocations and generations of imports are completely overhauled. Using the `import_map` generated in the WebIDL step all imports are now handled much more precisely in one location rather than haphazardly throughout the module. This means we have precise information about each import of the module and we only modify exactly what we're looking at. This also vastly simplifies intrinsic generation since it's all simply a codegen part of the `rust2js.rs` module now. 5. Handling of direct imports which don't have a JS shim generated is slightly different from before and is intended to be future-compatible with WebIDL bindings in its full glory, but we'll need to update it to handle cases for constructors and method calls eventually as well. 6. Intrinsic definitions now live in their own file (`src/intrinsic.rs`) and have a separated definition for their symbol name and signature. The actual implementation of each intrinsic lives in `rust2js.rs` There's a number of TODO items to finish before this merges. This includes reimplementing the anyref pass and actually implementing import maps for other targets. Those will come soon in follow-up commits, but the entire `tests/wasm/main.rs` suite is currently passing and this seems like a good checkpoint.
2019-05-23 09:15:26 -07:00
self.aux.export_map.insert(
setter_id,
AuxExport {
debug_name: format!("setter for `{}::{}`", struct_.name, field.name),
arg_names: None,
comments: concatenate_comments(&field.comments),
kind: AuxExportKind::Setter {
class: struct_.name.to_string(),
field: field.name.to_string(),
},
},
);
}
let aux = AuxStruct {
name: struct_.name.to_string(),
comments: concatenate_comments(&struct_.comments),
};
self.aux.structs.push(aux);
let wrap_constructor = wasm_bindgen_shared::new_function(struct_.name);
if let Some((import_id, _id)) = self.function_imports.get(&wrap_constructor) {
self.aux.import_map.insert(
*import_id,
AuxImport::WrapInExportedClass(struct_.name.to_string()),
);
let binding = Function {
shim_idx: 0,
arguments: vec![Descriptor::I32],
ret: Descriptor::Anyref,
};
Second large refactor for WebIDL bindings This commit is the second, and hopefully last massive, refactor for using WebIDL bindings internally in `wasm-bindgen`. This commit actually fully executes on the task at hand, moving `wasm-bindgen` to internally using WebIDL bindings throughout its code generation, anyref passes, etc. This actually fixes a number of issues that have existed in the anyref pass for some time now! The main changes here are to basically remove the usage of `Descriptor` from generating JS bindings. Instead two new types are introduced: `NonstandardIncoming` and `NonstandardOutgoing` which are bindings lists used for incoming/outgoing bindings. These mirror the standard terminology and literally have variants which are the standard values. All `Descriptor` types are now mapped into lists of incoming/outgoing bindings and used for process in wasm-bindgen. All JS generation has been refactored and updated to now process these lists of bindings instead of the previous `Descriptor`. In other words this commit takes `js2rust.rs` and `rust2js.rs` and first splits them in two. Interpretation of `Descriptor` and what to do for conversions is in the binding selection modules. The actual generation of JS from the binding selection is now performed by `incoming.rs` and `outgoing.rs`. To boot this also deduplicates all the code between the argument handling of `js2rust.rs` and return value handling of `rust2js.rs`. This means that to implement a new binding you only need to implement it one place and it's implemented for free in the other! This commit is not the end of the story though. I would like to add a mdoe to `wasm-bindgen` that literally emits a WebIDL bindings section. That's left for a third (and hopefully final) refactoring which is also intended to optimize generated JS for bindings. This commit currently loses the optimization where an imported is hooked up by value directly whenever a shim isn't needed. It's planned that the next refactoring to emit a webidl binding section that can be added back in. It shouldn't be too too hard hopefully since all the scaffolding is in place now. cc #1524
2019-06-10 07:09:51 -07:00
bindings::register_import(
self.module,
&mut self.bindings,
*import_id,
binding,
ast::WebidlFunctionKind::Static,
)?;
First refactor for WebIDL bindings This commit starts the `wasm-bindgen` CLI tool down the road to being a true polyfill for WebIDL bindings. This refactor is probably the first of a few, but is hopefully the largest and most sprawling and everything will be a bit more targeted from here on out. The goal of this refactoring is to separate out the massive `crates/cli-support/src/js/mod.rs` into a number of separate pieces of functionality. It currently takes care of basically everything including: * Binding intrinsics * Handling anyref transformations * Generating all JS for imports/exports * All the logic for how to import and how to name imports * Execution and management of wasm-bindgen closures Many of these are separable concerns and most overlap with WebIDL bindings. The internal refactoring here is intended to make it more clear who's responsible for what as well as making some existing operations much more straightforward. At a high-level, the following changes are done: 1. A `src/webidl.rs` module is introduced. The purpose of this module is to take all of the raw wasm-bindgen custom sections from the module and transform them into a WebIDL bindings section. This module has a placeholder `WebidlCustomSection` which is nowhere near the actual custom section but if you squint is in theory very similar. It's hoped that this will eventually become the true WebIDL custom section, currently being developed in an external crate. Currently, however, the WebIDL bindings custom section only covers a subset of the functionality we export to wasm-bindgen users. To avoid leaving them high and dry this module also contains an auxiliary custom section named `WasmBindgenAux`. This custom section isn't intended to have a binary format, but is intended to represent a theoretical custom section necessary to couple with WebIDL bindings to achieve all our desired functionality in `wasm-bindgen`. It'll never be standardized, but it'll also never be serialized :) 2. The `src/webidl.rs` module now takes over quite a bit of functionality from `src/js/mod.rs`. Namely it handles synthesis of an `export_map` and an `import_map` mapping export/import IDs to exactly what's expected to be hooked up there. This does not include type information (as that's in the bindings section) but rather includes things like "this is the method of class A" or "this import is from module `foo`" and things like that. These could arguably be subsumed by future JS features as well, but that's for another time! 3. All handling of wasm-bindgen "descriptor functions" now happens in a dedicated `src/descriptors.rs` module. The output of this module is its own custom section (intended to be immediately consumed by the WebIDL module) which is in theory what we want to ourselves emit one day but rustc isn't capable of doing so right now. 4. Invocations and generations of imports are completely overhauled. Using the `import_map` generated in the WebIDL step all imports are now handled much more precisely in one location rather than haphazardly throughout the module. This means we have precise information about each import of the module and we only modify exactly what we're looking at. This also vastly simplifies intrinsic generation since it's all simply a codegen part of the `rust2js.rs` module now. 5. Handling of direct imports which don't have a JS shim generated is slightly different from before and is intended to be future-compatible with WebIDL bindings in its full glory, but we'll need to update it to handle cases for constructors and method calls eventually as well. 6. Intrinsic definitions now live in their own file (`src/intrinsic.rs`) and have a separated definition for their symbol name and signature. The actual implementation of each intrinsic lives in `rust2js.rs` There's a number of TODO items to finish before this merges. This includes reimplementing the anyref pass and actually implementing import maps for other targets. Those will come soon in follow-up commits, but the entire `tests/wasm/main.rs` suite is currently passing and this seems like a good checkpoint.
2019-05-23 09:15:26 -07:00
}
Ok(())
}
fn determine_import(&self, import: &decode::Import<'_>, item: &str) -> Result<JsImport, Error> {
let is_local_snippet = match import.module {
decode::ImportModule::Named(s) => self.aux.local_modules.contains_key(s),
decode::ImportModule::RawNamed(_) => false,
decode::ImportModule::Inline(_) => true,
decode::ImportModule::None => false,
};
// Similar to `--target no-modules`, only allow vendor prefixes
// basically for web apis, shouldn't be necessary for things like npm
// packages or other imported items.
let vendor_prefixes = self.vendor_prefixes.get(item);
if let Some(vendor_prefixes) = vendor_prefixes {
assert!(vendor_prefixes.len() > 0);
if is_local_snippet {
bail!(
"local JS snippets do not support vendor prefixes for \
the import of `{}` with a polyfill of `{}`",
item,
&vendor_prefixes[0]
);
}
if let decode::ImportModule::Named(module) = &import.module {
bail!(
"import of `{}` from `{}` has a polyfill of `{}` listed, but
vendor prefixes aren't supported when importing from modules",
item,
module,
&vendor_prefixes[0],
);
}
if let Some(ns) = &import.js_namespace {
bail!(
"import of `{}` through js namespace `{}` isn't supported \
right now when it lists a polyfill",
item,
ns
);
}
return Ok(JsImport {
name: JsImportName::VendorPrefixed {
name: item.to_string(),
prefixes: vendor_prefixes.clone(),
},
fields: Vec::new(),
});
}
let (name, fields) = match import.js_namespace {
Some(ns) => (ns, vec![item.to_string()]),
None => (item, Vec::new()),
};
let name = match import.module {
decode::ImportModule::Named(module) if is_local_snippet => JsImportName::LocalModule {
module: module.to_string(),
name: name.to_string(),
},
decode::ImportModule::Named(module) | decode::ImportModule::RawNamed(module) => {
JsImportName::Module {
module: module.to_string(),
name: name.to_string(),
}
}
decode::ImportModule::Inline(idx) => {
let offset = self
.aux
.snippets
.get(self.unique_crate_identifier)
.map(|s| s.len())
.unwrap_or(0);
JsImportName::InlineJs {
unique_crate_identifier: self.unique_crate_identifier.to_string(),
snippet_idx_in_crate: idx as usize + offset,
name: name.to_string(),
}
}
decode::ImportModule::None => JsImportName::Global {
name: name.to_string(),
},
};
Ok(JsImport { name, fields })
}
/// Perform a small verification pass over the module to perform some
/// internal sanity checks.
fn verify(&self) -> Result<(), Error> {
let mut imports_counted = 0;
for import in self.module.imports.iter() {
if import.module != PLACEHOLDER_MODULE {
continue;
}
match import.kind {
walrus::ImportKind::Function(_) => {}
_ => bail!("import from `{}` was not a function", PLACEHOLDER_MODULE),
}
// Ensure that everything imported from the `__wbindgen_placeholder__`
// module has a location listed as to where it's expected to be
// imported from.
if !self.aux.import_map.contains_key(&import.id()) {
bail!(
"import of `{}` doesn't have an import map item listed",
import.name
);
}
// Also make sure there's a binding listed for it.
if !self.bindings.imports.contains_key(&import.id()) {
bail!("import of `{}` doesn't have a binding listed", import.name);
}
imports_counted += 1;
}
// Make sure there's no extraneous bindings that weren't actually
// imported in the module.
if self.aux.import_map.len() != imports_counted {
bail!("import map is larger than the number of imports");
}
if self.bindings.imports.len() != imports_counted {
bail!("import binding map is larger than the number of imports");
}
// Make sure the export map and export bindings map contain the same
// number of entries.
for id in self.bindings.exports.keys() {
if !self.aux.export_map.contains_key(id) {
bail!("bindings map has an entry that the export map does not");
}
}
if self.bindings.exports.len() != self.aux.export_map.len() {
bail!("export map and export bindings map have different sizes");
}
Ok(())
}
}
Second large refactor for WebIDL bindings This commit is the second, and hopefully last massive, refactor for using WebIDL bindings internally in `wasm-bindgen`. This commit actually fully executes on the task at hand, moving `wasm-bindgen` to internally using WebIDL bindings throughout its code generation, anyref passes, etc. This actually fixes a number of issues that have existed in the anyref pass for some time now! The main changes here are to basically remove the usage of `Descriptor` from generating JS bindings. Instead two new types are introduced: `NonstandardIncoming` and `NonstandardOutgoing` which are bindings lists used for incoming/outgoing bindings. These mirror the standard terminology and literally have variants which are the standard values. All `Descriptor` types are now mapped into lists of incoming/outgoing bindings and used for process in wasm-bindgen. All JS generation has been refactored and updated to now process these lists of bindings instead of the previous `Descriptor`. In other words this commit takes `js2rust.rs` and `rust2js.rs` and first splits them in two. Interpretation of `Descriptor` and what to do for conversions is in the binding selection modules. The actual generation of JS from the binding selection is now performed by `incoming.rs` and `outgoing.rs`. To boot this also deduplicates all the code between the argument handling of `js2rust.rs` and return value handling of `rust2js.rs`. This means that to implement a new binding you only need to implement it one place and it's implemented for free in the other! This commit is not the end of the story though. I would like to add a mdoe to `wasm-bindgen` that literally emits a WebIDL bindings section. That's left for a third (and hopefully final) refactoring which is also intended to optimize generated JS for bindings. This commit currently loses the optimization where an imported is hooked up by value directly whenever a shim isn't needed. It's planned that the next refactoring to emit a webidl binding section that can be added back in. It shouldn't be too too hard hopefully since all the scaffolding is in place now. cc #1524
2019-06-10 07:09:51 -07:00
impl walrus::CustomSection for NonstandardWebidlSection {
First refactor for WebIDL bindings This commit starts the `wasm-bindgen` CLI tool down the road to being a true polyfill for WebIDL bindings. This refactor is probably the first of a few, but is hopefully the largest and most sprawling and everything will be a bit more targeted from here on out. The goal of this refactoring is to separate out the massive `crates/cli-support/src/js/mod.rs` into a number of separate pieces of functionality. It currently takes care of basically everything including: * Binding intrinsics * Handling anyref transformations * Generating all JS for imports/exports * All the logic for how to import and how to name imports * Execution and management of wasm-bindgen closures Many of these are separable concerns and most overlap with WebIDL bindings. The internal refactoring here is intended to make it more clear who's responsible for what as well as making some existing operations much more straightforward. At a high-level, the following changes are done: 1. A `src/webidl.rs` module is introduced. The purpose of this module is to take all of the raw wasm-bindgen custom sections from the module and transform them into a WebIDL bindings section. This module has a placeholder `WebidlCustomSection` which is nowhere near the actual custom section but if you squint is in theory very similar. It's hoped that this will eventually become the true WebIDL custom section, currently being developed in an external crate. Currently, however, the WebIDL bindings custom section only covers a subset of the functionality we export to wasm-bindgen users. To avoid leaving them high and dry this module also contains an auxiliary custom section named `WasmBindgenAux`. This custom section isn't intended to have a binary format, but is intended to represent a theoretical custom section necessary to couple with WebIDL bindings to achieve all our desired functionality in `wasm-bindgen`. It'll never be standardized, but it'll also never be serialized :) 2. The `src/webidl.rs` module now takes over quite a bit of functionality from `src/js/mod.rs`. Namely it handles synthesis of an `export_map` and an `import_map` mapping export/import IDs to exactly what's expected to be hooked up there. This does not include type information (as that's in the bindings section) but rather includes things like "this is the method of class A" or "this import is from module `foo`" and things like that. These could arguably be subsumed by future JS features as well, but that's for another time! 3. All handling of wasm-bindgen "descriptor functions" now happens in a dedicated `src/descriptors.rs` module. The output of this module is its own custom section (intended to be immediately consumed by the WebIDL module) which is in theory what we want to ourselves emit one day but rustc isn't capable of doing so right now. 4. Invocations and generations of imports are completely overhauled. Using the `import_map` generated in the WebIDL step all imports are now handled much more precisely in one location rather than haphazardly throughout the module. This means we have precise information about each import of the module and we only modify exactly what we're looking at. This also vastly simplifies intrinsic generation since it's all simply a codegen part of the `rust2js.rs` module now. 5. Handling of direct imports which don't have a JS shim generated is slightly different from before and is intended to be future-compatible with WebIDL bindings in its full glory, but we'll need to update it to handle cases for constructors and method calls eventually as well. 6. Intrinsic definitions now live in their own file (`src/intrinsic.rs`) and have a separated definition for their symbol name and signature. The actual implementation of each intrinsic lives in `rust2js.rs` There's a number of TODO items to finish before this merges. This includes reimplementing the anyref pass and actually implementing import maps for other targets. Those will come soon in follow-up commits, but the entire `tests/wasm/main.rs` suite is currently passing and this seems like a good checkpoint.
2019-05-23 09:15:26 -07:00
fn name(&self) -> &str {
"webidl custom section"
}
fn data(&self, _: &walrus::IdsToIndices) -> Cow<[u8]> {
panic!("shouldn't emit custom sections just yet");
}
}
impl walrus::CustomSection for WasmBindgenAux {
fn name(&self) -> &str {
"wasm-bindgen custom section"
}
fn data(&self, _: &walrus::IdsToIndices) -> Cow<[u8]> {
panic!("shouldn't emit custom sections just yet");
}
}
fn extract_programs<'a>(
module: &mut Module,
program_storage: &'a mut Vec<Vec<u8>>,
) -> Result<Vec<decode::Program<'a>>, Error> {
let my_version = wasm_bindgen_shared::version();
assert!(program_storage.is_empty());
while let Some(raw) = module.customs.remove_raw("__wasm_bindgen_unstable") {
log::debug!(
"custom section '{}' looks like a wasm bindgen section",
raw.name
);
program_storage.push(raw.data);
}
let mut ret = Vec::new();
for program in program_storage.iter() {
let mut payload = &program[..];
while let Some(data) = get_remaining(&mut payload) {
// Historical versions of wasm-bindgen have used JSON as the custom
// data section format. Newer versions, however, are using a custom
// serialization protocol that looks much more like the wasm spec.
//
// We, however, want a sanity check to ensure that if we're running
// against the wrong wasm-bindgen we get a nicer error than an
// internal decode error. To that end we continue to verify a tiny
// bit of json at the beginning of each blob before moving to the
// next blob. This should keep us compatible with older wasm-bindgen
// instances as well as forward-compatible for now.
//
// Note, though, that as `wasm-pack` picks up steam it's hoped we
// can just delete this entirely. The `wasm-pack` project already
// manages versions for us, so we in theory should need this check
// less and less over time.
if let Some(their_version) = verify_schema_matches(data)? {
bail!(
"
it looks like the Rust project used to create this wasm file was linked against
a different version of wasm-bindgen than this binary:
rust wasm file: {}
this binary: {}
Currently the bindgen format is unstable enough that these two version must
exactly match, so it's required that these two version are kept in sync by
either updating the wasm-bindgen dependency or this binary. You should be able
to update the wasm-bindgen dependency with:
cargo update -p wasm-bindgen
or you can update the binary with
cargo install -f wasm-bindgen-cli
if this warning fails to go away though and you're not sure what to do feel free
to open an issue at https://github.com/rustwasm/wasm-bindgen/issues!
",
their_version,
my_version,
);
}
let next = get_remaining(&mut payload).unwrap();
log::debug!("found a program of length {}", next.len());
ret.push(<decode::Program as decode::Decode>::decode_all(next));
}
}
Ok(ret)
}
fn get_remaining<'a>(data: &mut &'a [u8]) -> Option<&'a [u8]> {
if data.len() == 0 {
return None;
}
let len = ((data[0] as usize) << 0)
| ((data[1] as usize) << 8)
| ((data[2] as usize) << 16)
| ((data[3] as usize) << 24);
let (a, b) = data[4..].split_at(len);
*data = b;
Some(a)
}
fn verify_schema_matches<'a>(data: &'a [u8]) -> Result<Option<&'a str>, Error> {
macro_rules! bad {
() => {
bail!("failed to decode what looked like wasm-bindgen data")
};
}
let data = match str::from_utf8(data) {
Ok(s) => s,
Err(_) => bad!(),
};
log::debug!("found version specifier {}", data);
if !data.starts_with("{") || !data.ends_with("}") {
bad!()
}
let needle = "\"schema_version\":\"";
let rest = match data.find(needle) {
Some(i) => &data[i + needle.len()..],
None => bad!(),
};
let their_schema_version = match rest.find("\"") {
Some(i) => &rest[..i],
None => bad!(),
};
if their_schema_version == wasm_bindgen_shared::SCHEMA_VERSION {
return Ok(None);
}
let needle = "\"version\":\"";
let rest = match data.find(needle) {
Some(i) => &data[i + needle.len()..],
None => bad!(),
};
let their_version = match rest.find("\"") {
Some(i) => &rest[..i],
None => bad!(),
};
Ok(Some(their_version))
}
fn concatenate_comments(comments: &[&str]) -> String {
comments
.iter()
.map(|s| s.trim_matches('"'))
.collect::<Vec<_>>()
.join("\n")
}
/// Do we need to generate JS glue shims for these incoming bindings?
pub fn incoming_do_not_require_glue(
exprs: &[NonstandardIncoming],
from_webidl_tys: &[ast::WebidlTypeRef],
to_wasm_tys: &[walrus::ValType],
) -> bool {
exprs.len() == from_webidl_tys.len()
&& exprs.len() == to_wasm_tys.len()
&& exprs
.iter()
.zip(from_webidl_tys)
.zip(to_wasm_tys)
.enumerate()
.all(|(i, ((expr, from_webidl_ty), to_wasm_ty))| match expr {
NonstandardIncoming::Standard(e) => e.is_expressible_in_js_without_webidl_bindings(
*from_webidl_ty,
*to_wasm_ty,
i as u32,
),
_ => false,
})
}
/// Do we need to generate JS glue shims for these outgoing bindings?
pub fn outgoing_do_not_require_glue(
exprs: &[NonstandardOutgoing],
from_wasm_tys: &[walrus::ValType],
to_webidl_tys: &[ast::WebidlTypeRef],
) -> bool {
exprs.len() == from_wasm_tys.len()
&& exprs.len() == to_webidl_tys.len()
&& exprs
.iter()
.zip(from_wasm_tys)
.zip(to_webidl_tys)
.enumerate()
.all(|(i, ((expr, from_wasm_ty), to_webidl_ty))| match expr {
NonstandardOutgoing::Standard(e) => e.is_expressible_in_js_without_webidl_bindings(
*from_wasm_ty,
*to_webidl_ty,
i as u32,
),
_ => false,
})
}