806 lines
28 KiB
Rust
Raw Normal View History

Migrate `wasm-bindgen` to using `walrus` This commit moves `wasm-bindgen` the CLI tool from internally using `parity-wasm` for wasm parsing/serialization to instead use `walrus`. The `walrus` crate is something we've been working on recently with an aim to replace the usage of `parity-wasm` in `wasm-bindgen` to make the current CLI tool more maintainable as well as more future-proof. The `walrus` crate provides a much nicer AST to work with as well as a structured `Module`, whereas `parity-wasm` provides a very raw interface to the wasm module which isn't really appropriate for our use case. The many transformations and tweaks that wasm-bindgen does have a huge amount of ad-hoc index management to carefully craft a final wasm binary, but this is all entirely taken care for us with the `walrus` crate. Additionally, `wasm-bindgen` will ingest and rewrite the wasm file, often changing the binary offsets of functions. Eventually with DWARF debug information we'll need to be sure to preserve the debug information throughout the transformations that `wasm-bindgen` does today. This is practically impossible to do with the `parity-wasm` architecture, but `walrus` was designed from the get-go to solve this problem transparently in the `walrus` crate itself. (it doesn't today, but this is planned work) It is the intention that this does not end up regressing any `wasm-bindgen` use cases, neither in functionality or in speed. As a large change and refactoring, however, it's likely that at least something will arise! We'll want to continue to remain vigilant to any issues that come up with this commit. Note that the `gc` crate has been deleted as part of this change, as the `gc` crate is no longer necessary since `walrus` does it automatically. Additionally the `gc` crate was one of the main problems with preserving debug information as it often deletes wasm items! Finally, this also starts moving crates to the 2018 edition where necessary since `walrus` requires the 2018 edition, and in general it's more pleasant to work within the 2018 edition!
2019-01-31 09:54:23 -08:00
use crate::descriptor::{Descriptor, Function};
Add experimental support for the `anyref` type This commit adds experimental support to `wasm-bindgen` to emit and leverage the `anyref` native wasm type. This native type is still in a proposal status (the reference-types proposal). The intention of `anyref` is to be able to directly hold JS values in wasm and pass the to imported functions, namely to empower eventual host bindings (now renamed WebIDL bindings) integration where we can skip JS shims altogether for many imports. This commit doesn't actually affect wasm-bindgen's behavior at all as-is, but rather this support requires an opt-in env var to be configured. Once the support is stable in browsers it's intended that this will add a CLI switch for turning on this support, eventually defaulting it to `true` in the far future. The basic strategy here is to take the `stack` and `slab` globals in the generated JS glue and move them into wasm using a table. This new table in wasm is managed at the fringes via injected shims. At `wasm-bindgen`-time the CLI will rewrite exports and imports with shims that actually use `anyref` if needed, performing loads/stores inside the wasm module instead of externally in the wasm module. This should provide a boost over what we have today, but it's not a fantastic strategy long term. We have a more grand vision for `anyref` being a first-class type in the language, but that's on a much longer horizon and this is currently thought to be the best we can do in terms of integration in the near future. The stack/heap JS tables are combined into one wasm table. The stack starts at the end of the table and grows down with a stack pointer (also injected). The heap starts at the end and grows up (state managed in linear memory). The anyref transformation here will hook up various intrinsics in wasm-bindgen to the runtime functionality if the anyref supoprt is enabled. The main tricky treatment here was applied to closures, where we need JS to use a different function pointer than the one Rust gives it to use a JS function pointer empowered with anyref. This works by switching up a bit how descriptors work, embedding the shims to call inside descriptors rather than communicated at runtime. This means that we're accessing constant values in the generated JS and we can just update the constant value accessed.
2018-10-18 08:43:36 -07:00
use crate::js::js2rust::ExportedShim;
Migrate `wasm-bindgen` to using `walrus` This commit moves `wasm-bindgen` the CLI tool from internally using `parity-wasm` for wasm parsing/serialization to instead use `walrus`. The `walrus` crate is something we've been working on recently with an aim to replace the usage of `parity-wasm` in `wasm-bindgen` to make the current CLI tool more maintainable as well as more future-proof. The `walrus` crate provides a much nicer AST to work with as well as a structured `Module`, whereas `parity-wasm` provides a very raw interface to the wasm module which isn't really appropriate for our use case. The many transformations and tweaks that wasm-bindgen does have a huge amount of ad-hoc index management to carefully craft a final wasm binary, but this is all entirely taken care for us with the `walrus` crate. Additionally, `wasm-bindgen` will ingest and rewrite the wasm file, often changing the binary offsets of functions. Eventually with DWARF debug information we'll need to be sure to preserve the debug information throughout the transformations that `wasm-bindgen` does today. This is practically impossible to do with the `parity-wasm` architecture, but `walrus` was designed from the get-go to solve this problem transparently in the `walrus` crate itself. (it doesn't today, but this is planned work) It is the intention that this does not end up regressing any `wasm-bindgen` use cases, neither in functionality or in speed. As a large change and refactoring, however, it's likely that at least something will arise! We'll want to continue to remain vigilant to any issues that come up with this commit. Note that the `gc` crate has been deleted as part of this change, as the `gc` crate is no longer necessary since `walrus` does it automatically. Additionally the `gc` crate was one of the main problems with preserving debug information as it often deletes wasm items! Finally, this also starts moving crates to the 2018 edition where necessary since `walrus` requires the 2018 edition, and in general it's more pleasant to work within the 2018 edition!
2019-01-31 09:54:23 -08:00
use crate::js::{Context, ImportTarget, Js2Rust};
use failure::{bail, Error};
/// Helper struct for manufacturing a shim in JS used to translate Rust types to
/// JS, then invoking an imported JS function.
pub struct Rust2Js<'a, 'b: 'a> {
pub cx: &'a mut Context<'b>,
/// Arguments of the JS shim that we're generating, aka the variables passed
/// from Rust which are only numbers.
shim_arguments: Vec<String>,
/// Arguments which are forwarded to the imported JS function
js_arguments: Vec<String>,
/// Conversions that happen before we invoke the wasm function, such as
/// converting a string to a ptr/length pair.
prelude: String,
/// "Destructors" or cleanup that must happen after the wasm function
/// finishes. This is scheduled in a `finally` block.
finally: String,
/// Next global index to write to when passing arguments via the single
/// global stack.
global_idx: usize,
/// Index of the next argument for unique name generation purposes.
arg_idx: usize,
/// Expression used to generate the return value. The string "JS" in this
/// expression is replaced with the actual JS invocation eventually.
ret_expr: String,
/// Whether or not we're catching JS exceptions
catch: bool,
catch_and_rethrow: bool,
2018-08-19 13:39:16 +01:00
/// Whether or not the last argument is a slice representing variadic arguments.
variadic: bool,
Add experimental support for the `anyref` type This commit adds experimental support to `wasm-bindgen` to emit and leverage the `anyref` native wasm type. This native type is still in a proposal status (the reference-types proposal). The intention of `anyref` is to be able to directly hold JS values in wasm and pass the to imported functions, namely to empower eventual host bindings (now renamed WebIDL bindings) integration where we can skip JS shims altogether for many imports. This commit doesn't actually affect wasm-bindgen's behavior at all as-is, but rather this support requires an opt-in env var to be configured. Once the support is stable in browsers it's intended that this will add a CLI switch for turning on this support, eventually defaulting it to `true` in the far future. The basic strategy here is to take the `stack` and `slab` globals in the generated JS glue and move them into wasm using a table. This new table in wasm is managed at the fringes via injected shims. At `wasm-bindgen`-time the CLI will rewrite exports and imports with shims that actually use `anyref` if needed, performing loads/stores inside the wasm module instead of externally in the wasm module. This should provide a boost over what we have today, but it's not a fantastic strategy long term. We have a more grand vision for `anyref` being a first-class type in the language, but that's on a much longer horizon and this is currently thought to be the best we can do in terms of integration in the near future. The stack/heap JS tables are combined into one wasm table. The stack starts at the end of the table and grows down with a stack pointer (also injected). The heap starts at the end and grows up (state managed in linear memory). The anyref transformation here will hook up various intrinsics in wasm-bindgen to the runtime functionality if the anyref supoprt is enabled. The main tricky treatment here was applied to closures, where we need JS to use a different function pointer than the one Rust gives it to use a JS function pointer empowered with anyref. This works by switching up a bit how descriptors work, embedding the shims to call inside descriptors rather than communicated at runtime. This means that we're accessing constant values in the generated JS and we can just update the constant value accessed.
2018-10-18 08:43:36 -07:00
/// list of arguments that are anyref, and whether they're an owned anyref
/// or not.
pub anyref_args: Vec<(usize, bool)>,
pub ret_anyref: bool,
}
impl<'a, 'b> Rust2Js<'a, 'b> {
pub fn new(cx: &'a mut Context<'b>) -> Rust2Js<'a, 'b> {
Rust2Js {
cx,
shim_arguments: Vec::new(),
js_arguments: Vec::new(),
prelude: String::new(),
finally: String::new(),
global_idx: 0,
arg_idx: 0,
ret_expr: String::new(),
catch: false,
catch_and_rethrow: false,
2018-08-19 13:39:16 +01:00
variadic: false,
Add experimental support for the `anyref` type This commit adds experimental support to `wasm-bindgen` to emit and leverage the `anyref` native wasm type. This native type is still in a proposal status (the reference-types proposal). The intention of `anyref` is to be able to directly hold JS values in wasm and pass the to imported functions, namely to empower eventual host bindings (now renamed WebIDL bindings) integration where we can skip JS shims altogether for many imports. This commit doesn't actually affect wasm-bindgen's behavior at all as-is, but rather this support requires an opt-in env var to be configured. Once the support is stable in browsers it's intended that this will add a CLI switch for turning on this support, eventually defaulting it to `true` in the far future. The basic strategy here is to take the `stack` and `slab` globals in the generated JS glue and move them into wasm using a table. This new table in wasm is managed at the fringes via injected shims. At `wasm-bindgen`-time the CLI will rewrite exports and imports with shims that actually use `anyref` if needed, performing loads/stores inside the wasm module instead of externally in the wasm module. This should provide a boost over what we have today, but it's not a fantastic strategy long term. We have a more grand vision for `anyref` being a first-class type in the language, but that's on a much longer horizon and this is currently thought to be the best we can do in terms of integration in the near future. The stack/heap JS tables are combined into one wasm table. The stack starts at the end of the table and grows down with a stack pointer (also injected). The heap starts at the end and grows up (state managed in linear memory). The anyref transformation here will hook up various intrinsics in wasm-bindgen to the runtime functionality if the anyref supoprt is enabled. The main tricky treatment here was applied to closures, where we need JS to use a different function pointer than the one Rust gives it to use a JS function pointer empowered with anyref. This works by switching up a bit how descriptors work, embedding the shims to call inside descriptors rather than communicated at runtime. This means that we're accessing constant values in the generated JS and we can just update the constant value accessed.
2018-10-18 08:43:36 -07:00
anyref_args: Vec::new(),
ret_anyref: false,
}
}
pub fn catch(&mut self, catch: bool) -> &mut Self {
self.catch = catch;
self
}
pub fn catch_and_rethrow(&mut self, catch_and_rethrow: bool) -> &mut Self {
self.catch_and_rethrow = catch_and_rethrow;
self
}
2018-08-19 13:39:16 +01:00
pub fn variadic(&mut self, variadic: bool) -> &mut Self {
self.variadic = variadic;
self
}
/// Generates all bindings necessary for the signature in `Function`,
/// creating necessary argument conversions and return value processing.
pub fn process(&mut self, function: &Function) -> Result<&mut Self, Error> {
for arg in function.arguments.iter() {
self.argument(arg)?;
}
self.ret(&function.ret)?;
Ok(self)
}
2018-08-19 13:39:16 +01:00
/// Get a generated name for an argument.
fn shim_argument(&mut self) -> String {
let s = format!("arg{}", self.arg_idx);
self.arg_idx += 1;
self.shim_arguments.push(s.clone());
s
}
fn argument(&mut self, arg: &Descriptor) -> Result<(), Error> {
let abi = self.shim_argument();
Add support for optional slice types (#507) * Shard the `convert.rs` module into sub-modules Hopefully this'll make the organization a little nicer over time! * Start adding support for optional types This commit starts adding support for optional types to wasm-bindgen as arguments/return values to functions. The strategy here is to add two new traits, `OptionIntoWasmAbi` and `OptionFromWasmAbi`. These two traits are used as a blanket impl to implement `IntoWasmAbi` and `FromWasmAbi` for `Option<T>`. Some consequences of this design: * It should be possible to ensure `Option<SomeForeignType>` implements to/from wasm traits. This is because the option-based traits can be implemented for foreign types. * A specialized implementation is possible for all types, so there's no need for `Option<T>` to introduce unnecessary overhead. * Two new traits is a bit unforutnate but I can't currently think of an alternative design that works for the above two constraints, although it doesn't mean one doesn't exist! * The error messages for "can't use this type here" is actually halfway decent because it says these new traits need to be implemented, which provides a good place to document and talk about what's going on here! * Nested references like `Option<&T>` can't implement `FromWasmAbi`. This means that you can't define a function in Rust which takes `Option<&str>`. It may be possible to do this one day but it'll likely require more trait trickery than I'm capable of right now. * Add support for optional slices This commit adds support for optional slice types, things like strings and arrays. The null representation of these has a pointer value of 0, which should never happen in normal Rust. Otherwise the various plumbing is done throughout the tooling to enable these types in all locations. * Fix `takeObject` on global sentinels These don't have a reference count as they're always expected to work, so avoid actually dropping a reference on them. * Remove some no longer needed bindings * Add support for optional anyref types This commit adds support for optional imported class types. Each type imported with `#[wasm_bindgen]` automatically implements the relevant traits and now supports `Option<Foo>` in various argument/return positions. * Fix building without the `std` feature * Actually fix the build... * Add support for optional types to WebIDL Closes #502
2018-07-19 14:44:23 -05:00
let (arg, optional) = match arg {
Descriptor::Option(t) => (&**t, true),
_ => (arg, false),
};
if let Some(ty) = arg.vector_kind() {
let abi2 = self.shim_argument();
Add experimental support for the `anyref` type This commit adds experimental support to `wasm-bindgen` to emit and leverage the `anyref` native wasm type. This native type is still in a proposal status (the reference-types proposal). The intention of `anyref` is to be able to directly hold JS values in wasm and pass the to imported functions, namely to empower eventual host bindings (now renamed WebIDL bindings) integration where we can skip JS shims altogether for many imports. This commit doesn't actually affect wasm-bindgen's behavior at all as-is, but rather this support requires an opt-in env var to be configured. Once the support is stable in browsers it's intended that this will add a CLI switch for turning on this support, eventually defaulting it to `true` in the far future. The basic strategy here is to take the `stack` and `slab` globals in the generated JS glue and move them into wasm using a table. This new table in wasm is managed at the fringes via injected shims. At `wasm-bindgen`-time the CLI will rewrite exports and imports with shims that actually use `anyref` if needed, performing loads/stores inside the wasm module instead of externally in the wasm module. This should provide a boost over what we have today, but it's not a fantastic strategy long term. We have a more grand vision for `anyref` being a first-class type in the language, but that's on a much longer horizon and this is currently thought to be the best we can do in terms of integration in the near future. The stack/heap JS tables are combined into one wasm table. The stack starts at the end of the table and grows down with a stack pointer (also injected). The heap starts at the end and grows up (state managed in linear memory). The anyref transformation here will hook up various intrinsics in wasm-bindgen to the runtime functionality if the anyref supoprt is enabled. The main tricky treatment here was applied to closures, where we need JS to use a different function pointer than the one Rust gives it to use a JS function pointer empowered with anyref. This works by switching up a bit how descriptors work, embedding the shims to call inside descriptors rather than communicated at runtime. This means that we're accessing constant values in the generated JS and we can just update the constant value accessed.
2018-10-18 08:43:36 -07:00
let f = self.cx.expose_get_vector_from_wasm(ty)?;
2018-06-27 22:42:34 -07:00
self.prelude(&format!(
Add support for optional slice types (#507) * Shard the `convert.rs` module into sub-modules Hopefully this'll make the organization a little nicer over time! * Start adding support for optional types This commit starts adding support for optional types to wasm-bindgen as arguments/return values to functions. The strategy here is to add two new traits, `OptionIntoWasmAbi` and `OptionFromWasmAbi`. These two traits are used as a blanket impl to implement `IntoWasmAbi` and `FromWasmAbi` for `Option<T>`. Some consequences of this design: * It should be possible to ensure `Option<SomeForeignType>` implements to/from wasm traits. This is because the option-based traits can be implemented for foreign types. * A specialized implementation is possible for all types, so there's no need for `Option<T>` to introduce unnecessary overhead. * Two new traits is a bit unforutnate but I can't currently think of an alternative design that works for the above two constraints, although it doesn't mean one doesn't exist! * The error messages for "can't use this type here" is actually halfway decent because it says these new traits need to be implemented, which provides a good place to document and talk about what's going on here! * Nested references like `Option<&T>` can't implement `FromWasmAbi`. This means that you can't define a function in Rust which takes `Option<&str>`. It may be possible to do this one day but it'll likely require more trait trickery than I'm capable of right now. * Add support for optional slices This commit adds support for optional slice types, things like strings and arrays. The null representation of these has a pointer value of 0, which should never happen in normal Rust. Otherwise the various plumbing is done throughout the tooling to enable these types in all locations. * Fix `takeObject` on global sentinels These don't have a reference count as they're always expected to work, so avoid actually dropping a reference on them. * Remove some no longer needed bindings * Add support for optional anyref types This commit adds support for optional imported class types. Each type imported with `#[wasm_bindgen]` automatically implements the relevant traits and now supports `Option<Foo>` in various argument/return positions. * Fix building without the `std` feature * Actually fix the build... * Add support for optional types to WebIDL Closes #502
2018-07-19 14:44:23 -05:00
"let v{0} = {prefix}{func}({0}, {1});",
2018-06-27 22:42:34 -07:00
abi,
abi2,
Add support for optional slice types (#507) * Shard the `convert.rs` module into sub-modules Hopefully this'll make the organization a little nicer over time! * Start adding support for optional types This commit starts adding support for optional types to wasm-bindgen as arguments/return values to functions. The strategy here is to add two new traits, `OptionIntoWasmAbi` and `OptionFromWasmAbi`. These two traits are used as a blanket impl to implement `IntoWasmAbi` and `FromWasmAbi` for `Option<T>`. Some consequences of this design: * It should be possible to ensure `Option<SomeForeignType>` implements to/from wasm traits. This is because the option-based traits can be implemented for foreign types. * A specialized implementation is possible for all types, so there's no need for `Option<T>` to introduce unnecessary overhead. * Two new traits is a bit unforutnate but I can't currently think of an alternative design that works for the above two constraints, although it doesn't mean one doesn't exist! * The error messages for "can't use this type here" is actually halfway decent because it says these new traits need to be implemented, which provides a good place to document and talk about what's going on here! * Nested references like `Option<&T>` can't implement `FromWasmAbi`. This means that you can't define a function in Rust which takes `Option<&str>`. It may be possible to do this one day but it'll likely require more trait trickery than I'm capable of right now. * Add support for optional slices This commit adds support for optional slice types, things like strings and arrays. The null representation of these has a pointer value of 0, which should never happen in normal Rust. Otherwise the various plumbing is done throughout the tooling to enable these types in all locations. * Fix `takeObject` on global sentinels These don't have a reference count as they're always expected to work, so avoid actually dropping a reference on them. * Remove some no longer needed bindings * Add support for optional anyref types This commit adds support for optional imported class types. Each type imported with `#[wasm_bindgen]` automatically implements the relevant traits and now supports `Option<Foo>` in various argument/return positions. * Fix building without the `std` feature * Actually fix the build... * Add support for optional types to WebIDL Closes #502
2018-07-19 14:44:23 -05:00
func = f,
prefix = if optional {
format!("{} == 0 ? undefined : ", abi)
} else {
String::new()
},
2018-06-27 22:42:34 -07:00
));
if !arg.is_by_ref() && !arg.is_clamped_by_ref() {
2018-06-27 22:42:34 -07:00
self.prelude(&format!(
"\
Add support for optional slice types (#507) * Shard the `convert.rs` module into sub-modules Hopefully this'll make the organization a little nicer over time! * Start adding support for optional types This commit starts adding support for optional types to wasm-bindgen as arguments/return values to functions. The strategy here is to add two new traits, `OptionIntoWasmAbi` and `OptionFromWasmAbi`. These two traits are used as a blanket impl to implement `IntoWasmAbi` and `FromWasmAbi` for `Option<T>`. Some consequences of this design: * It should be possible to ensure `Option<SomeForeignType>` implements to/from wasm traits. This is because the option-based traits can be implemented for foreign types. * A specialized implementation is possible for all types, so there's no need for `Option<T>` to introduce unnecessary overhead. * Two new traits is a bit unforutnate but I can't currently think of an alternative design that works for the above two constraints, although it doesn't mean one doesn't exist! * The error messages for "can't use this type here" is actually halfway decent because it says these new traits need to be implemented, which provides a good place to document and talk about what's going on here! * Nested references like `Option<&T>` can't implement `FromWasmAbi`. This means that you can't define a function in Rust which takes `Option<&str>`. It may be possible to do this one day but it'll likely require more trait trickery than I'm capable of right now. * Add support for optional slices This commit adds support for optional slice types, things like strings and arrays. The null representation of these has a pointer value of 0, which should never happen in normal Rust. Otherwise the various plumbing is done throughout the tooling to enable these types in all locations. * Fix `takeObject` on global sentinels These don't have a reference count as they're always expected to work, so avoid actually dropping a reference on them. * Remove some no longer needed bindings * Add support for optional anyref types This commit adds support for optional imported class types. Each type imported with `#[wasm_bindgen]` automatically implements the relevant traits and now supports `Option<Foo>` in various argument/return positions. * Fix building without the `std` feature * Actually fix the build... * Add support for optional types to WebIDL Closes #502
2018-07-19 14:44:23 -05:00
{start}
v{0} = v{0}.slice();
wasm.__wbindgen_free({0}, {1} * {size});
{end}\
2018-06-27 22:42:34 -07:00
",
abi,
abi2,
Add support for optional slice types (#507) * Shard the `convert.rs` module into sub-modules Hopefully this'll make the organization a little nicer over time! * Start adding support for optional types This commit starts adding support for optional types to wasm-bindgen as arguments/return values to functions. The strategy here is to add two new traits, `OptionIntoWasmAbi` and `OptionFromWasmAbi`. These two traits are used as a blanket impl to implement `IntoWasmAbi` and `FromWasmAbi` for `Option<T>`. Some consequences of this design: * It should be possible to ensure `Option<SomeForeignType>` implements to/from wasm traits. This is because the option-based traits can be implemented for foreign types. * A specialized implementation is possible for all types, so there's no need for `Option<T>` to introduce unnecessary overhead. * Two new traits is a bit unforutnate but I can't currently think of an alternative design that works for the above two constraints, although it doesn't mean one doesn't exist! * The error messages for "can't use this type here" is actually halfway decent because it says these new traits need to be implemented, which provides a good place to document and talk about what's going on here! * Nested references like `Option<&T>` can't implement `FromWasmAbi`. This means that you can't define a function in Rust which takes `Option<&str>`. It may be possible to do this one day but it'll likely require more trait trickery than I'm capable of right now. * Add support for optional slices This commit adds support for optional slice types, things like strings and arrays. The null representation of these has a pointer value of 0, which should never happen in normal Rust. Otherwise the various plumbing is done throughout the tooling to enable these types in all locations. * Fix `takeObject` on global sentinels These don't have a reference count as they're always expected to work, so avoid actually dropping a reference on them. * Remove some no longer needed bindings * Add support for optional anyref types This commit adds support for optional imported class types. Each type imported with `#[wasm_bindgen]` automatically implements the relevant traits and now supports `Option<Foo>` in various argument/return positions. * Fix building without the `std` feature * Actually fix the build... * Add support for optional types to WebIDL Closes #502
2018-07-19 14:44:23 -05:00
size = ty.size(),
start = if optional {
format!("if ({} !== 0) {{", abi)
} else {
String::new()
},
Add support for optional slice types (#507) * Shard the `convert.rs` module into sub-modules Hopefully this'll make the organization a little nicer over time! * Start adding support for optional types This commit starts adding support for optional types to wasm-bindgen as arguments/return values to functions. The strategy here is to add two new traits, `OptionIntoWasmAbi` and `OptionFromWasmAbi`. These two traits are used as a blanket impl to implement `IntoWasmAbi` and `FromWasmAbi` for `Option<T>`. Some consequences of this design: * It should be possible to ensure `Option<SomeForeignType>` implements to/from wasm traits. This is because the option-based traits can be implemented for foreign types. * A specialized implementation is possible for all types, so there's no need for `Option<T>` to introduce unnecessary overhead. * Two new traits is a bit unforutnate but I can't currently think of an alternative design that works for the above two constraints, although it doesn't mean one doesn't exist! * The error messages for "can't use this type here" is actually halfway decent because it says these new traits need to be implemented, which provides a good place to document and talk about what's going on here! * Nested references like `Option<&T>` can't implement `FromWasmAbi`. This means that you can't define a function in Rust which takes `Option<&str>`. It may be possible to do this one day but it'll likely require more trait trickery than I'm capable of right now. * Add support for optional slices This commit adds support for optional slice types, things like strings and arrays. The null representation of these has a pointer value of 0, which should never happen in normal Rust. Otherwise the various plumbing is done throughout the tooling to enable these types in all locations. * Fix `takeObject` on global sentinels These don't have a reference count as they're always expected to work, so avoid actually dropping a reference on them. * Remove some no longer needed bindings * Add support for optional anyref types This commit adds support for optional imported class types. Each type imported with `#[wasm_bindgen]` automatically implements the relevant traits and now supports `Option<Foo>` in various argument/return positions. * Fix building without the `std` feature * Actually fix the build... * Add support for optional types to WebIDL Closes #502
2018-07-19 14:44:23 -05:00
end = if optional { "}" } else { "" },
2018-06-27 22:42:34 -07:00
));
self.cx.require_internal_export("__wbindgen_free")?;
}
self.js_arguments.push(format!("v{}", abi));
2018-06-27 22:42:34 -07:00
return Ok(());
}
Add support for optional slice types (#507) * Shard the `convert.rs` module into sub-modules Hopefully this'll make the organization a little nicer over time! * Start adding support for optional types This commit starts adding support for optional types to wasm-bindgen as arguments/return values to functions. The strategy here is to add two new traits, `OptionIntoWasmAbi` and `OptionFromWasmAbi`. These two traits are used as a blanket impl to implement `IntoWasmAbi` and `FromWasmAbi` for `Option<T>`. Some consequences of this design: * It should be possible to ensure `Option<SomeForeignType>` implements to/from wasm traits. This is because the option-based traits can be implemented for foreign types. * A specialized implementation is possible for all types, so there's no need for `Option<T>` to introduce unnecessary overhead. * Two new traits is a bit unforutnate but I can't currently think of an alternative design that works for the above two constraints, although it doesn't mean one doesn't exist! * The error messages for "can't use this type here" is actually halfway decent because it says these new traits need to be implemented, which provides a good place to document and talk about what's going on here! * Nested references like `Option<&T>` can't implement `FromWasmAbi`. This means that you can't define a function in Rust which takes `Option<&str>`. It may be possible to do this one day but it'll likely require more trait trickery than I'm capable of right now. * Add support for optional slices This commit adds support for optional slice types, things like strings and arrays. The null representation of these has a pointer value of 0, which should never happen in normal Rust. Otherwise the various plumbing is done throughout the tooling to enable these types in all locations. * Fix `takeObject` on global sentinels These don't have a reference count as they're always expected to work, so avoid actually dropping a reference on them. * Remove some no longer needed bindings * Add support for optional anyref types This commit adds support for optional imported class types. Each type imported with `#[wasm_bindgen]` automatically implements the relevant traits and now supports `Option<Foo>` in various argument/return positions. * Fix building without the `std` feature * Actually fix the build... * Add support for optional types to WebIDL Closes #502
2018-07-19 14:44:23 -05:00
// No need to special case `optional` here because `takeObject` will
// naturally work.
if arg.is_anyref() {
Add experimental support for the `anyref` type This commit adds experimental support to `wasm-bindgen` to emit and leverage the `anyref` native wasm type. This native type is still in a proposal status (the reference-types proposal). The intention of `anyref` is to be able to directly hold JS values in wasm and pass the to imported functions, namely to empower eventual host bindings (now renamed WebIDL bindings) integration where we can skip JS shims altogether for many imports. This commit doesn't actually affect wasm-bindgen's behavior at all as-is, but rather this support requires an opt-in env var to be configured. Once the support is stable in browsers it's intended that this will add a CLI switch for turning on this support, eventually defaulting it to `true` in the far future. The basic strategy here is to take the `stack` and `slab` globals in the generated JS glue and move them into wasm using a table. This new table in wasm is managed at the fringes via injected shims. At `wasm-bindgen`-time the CLI will rewrite exports and imports with shims that actually use `anyref` if needed, performing loads/stores inside the wasm module instead of externally in the wasm module. This should provide a boost over what we have today, but it's not a fantastic strategy long term. We have a more grand vision for `anyref` being a first-class type in the language, but that's on a much longer horizon and this is currently thought to be the best we can do in terms of integration in the near future. The stack/heap JS tables are combined into one wasm table. The stack starts at the end of the table and grows down with a stack pointer (also injected). The heap starts at the end and grows up (state managed in linear memory). The anyref transformation here will hook up various intrinsics in wasm-bindgen to the runtime functionality if the anyref supoprt is enabled. The main tricky treatment here was applied to closures, where we need JS to use a different function pointer than the one Rust gives it to use a JS function pointer empowered with anyref. This works by switching up a bit how descriptors work, embedding the shims to call inside descriptors rather than communicated at runtime. This means that we're accessing constant values in the generated JS and we can just update the constant value accessed.
2018-10-18 08:43:36 -07:00
let arg = self.cx.take_object(&abi);
self.js_arguments.push(arg);
self.anyref_args.push((self.arg_idx - 1, true));
return Ok(());
} else if arg.is_ref_anyref() {
Add experimental support for the `anyref` type This commit adds experimental support to `wasm-bindgen` to emit and leverage the `anyref` native wasm type. This native type is still in a proposal status (the reference-types proposal). The intention of `anyref` is to be able to directly hold JS values in wasm and pass the to imported functions, namely to empower eventual host bindings (now renamed WebIDL bindings) integration where we can skip JS shims altogether for many imports. This commit doesn't actually affect wasm-bindgen's behavior at all as-is, but rather this support requires an opt-in env var to be configured. Once the support is stable in browsers it's intended that this will add a CLI switch for turning on this support, eventually defaulting it to `true` in the far future. The basic strategy here is to take the `stack` and `slab` globals in the generated JS glue and move them into wasm using a table. This new table in wasm is managed at the fringes via injected shims. At `wasm-bindgen`-time the CLI will rewrite exports and imports with shims that actually use `anyref` if needed, performing loads/stores inside the wasm module instead of externally in the wasm module. This should provide a boost over what we have today, but it's not a fantastic strategy long term. We have a more grand vision for `anyref` being a first-class type in the language, but that's on a much longer horizon and this is currently thought to be the best we can do in terms of integration in the near future. The stack/heap JS tables are combined into one wasm table. The stack starts at the end of the table and grows down with a stack pointer (also injected). The heap starts at the end and grows up (state managed in linear memory). The anyref transformation here will hook up various intrinsics in wasm-bindgen to the runtime functionality if the anyref supoprt is enabled. The main tricky treatment here was applied to closures, where we need JS to use a different function pointer than the one Rust gives it to use a JS function pointer empowered with anyref. This works by switching up a bit how descriptors work, embedding the shims to call inside descriptors rather than communicated at runtime. This means that we're accessing constant values in the generated JS and we can just update the constant value accessed.
2018-10-18 08:43:36 -07:00
let arg = self.cx.get_object(&abi);
self.js_arguments.push(arg);
self.anyref_args.push((self.arg_idx - 1, false));
return Ok(());
Add support for optional slice types (#507) * Shard the `convert.rs` module into sub-modules Hopefully this'll make the organization a little nicer over time! * Start adding support for optional types This commit starts adding support for optional types to wasm-bindgen as arguments/return values to functions. The strategy here is to add two new traits, `OptionIntoWasmAbi` and `OptionFromWasmAbi`. These two traits are used as a blanket impl to implement `IntoWasmAbi` and `FromWasmAbi` for `Option<T>`. Some consequences of this design: * It should be possible to ensure `Option<SomeForeignType>` implements to/from wasm traits. This is because the option-based traits can be implemented for foreign types. * A specialized implementation is possible for all types, so there's no need for `Option<T>` to introduce unnecessary overhead. * Two new traits is a bit unforutnate but I can't currently think of an alternative design that works for the above two constraints, although it doesn't mean one doesn't exist! * The error messages for "can't use this type here" is actually halfway decent because it says these new traits need to be implemented, which provides a good place to document and talk about what's going on here! * Nested references like `Option<&T>` can't implement `FromWasmAbi`. This means that you can't define a function in Rust which takes `Option<&str>`. It may be possible to do this one day but it'll likely require more trait trickery than I'm capable of right now. * Add support for optional slices This commit adds support for optional slice types, things like strings and arrays. The null representation of these has a pointer value of 0, which should never happen in normal Rust. Otherwise the various plumbing is done throughout the tooling to enable these types in all locations. * Fix `takeObject` on global sentinels These don't have a reference count as they're always expected to work, so avoid actually dropping a reference on them. * Remove some no longer needed bindings * Add support for optional anyref types This commit adds support for optional imported class types. Each type imported with `#[wasm_bindgen]` automatically implements the relevant traits and now supports `Option<Foo>` in various argument/return positions. * Fix building without the `std` feature * Actually fix the build... * Add support for optional types to WebIDL Closes #502
2018-07-19 14:44:23 -05:00
}
if optional {
if arg.is_wasm_native() {
2018-08-03 16:28:35 +03:00
let value = self.shim_argument();
self.js_arguments.push(format!(
"{present} === 0 ? undefined : {value}",
value = value,
present = abi,
));
return Ok(());
2018-08-03 16:28:35 +03:00
}
if arg.is_abi_as_u32() {
self.js_arguments
.push(format!("{0} === 0xFFFFFF ? undefined : {0}", abi));
return Ok(());
2018-08-03 16:28:35 +03:00
}
if let Some(signed) = arg.get_64() {
let f = if signed {
self.cx.expose_int64_cvt_shim()
} else {
self.cx.expose_uint64_cvt_shim()
};
self.shim_argument();
let low = self.shim_argument();
let high = self.shim_argument();
let name = format!("n{}", abi);
self.prelude(&format!(
"
u32CvtShim[0] = {present} === 0 ? 0 : {low};
u32CvtShim[1] = {present} === 0 ? 0 : {high};
const {name} = {present} === 0 ? undefined : {f}[0];
2018-08-03 16:28:35 +03:00
",
present = abi,
low = low,
high = high,
f = f,
name = name,
));
self.js_arguments.push(name);
return Ok(());
}
2018-08-03 19:07:12 +03:00
match *arg {
Descriptor::Boolean => {
self.js_arguments
.push(format!("{0} === 0xFFFFFF ? undefined : {0} !== 0", abi));
return Ok(());
}
Descriptor::Enum { hole } => {
self.js_arguments
.push(format!("{0} === {1} ? undefined : {0}", abi, hole));
return Ok(());
}
2018-08-03 20:45:57 +03:00
Descriptor::Char => {
2019-04-16 10:52:27 -07:00
self.js_arguments.push(format!(
"{0} === 0xFFFFFF ? undefined : String.fromCodePoint({0})",
abi
));
return Ok(());
}
Descriptor::RustStruct(ref class) => {
self.cx.require_class_wrap(class);
let assign = format!(
"let c{0} = {0} === 0 ? undefined : {1}.__wrap({0});",
abi, class
);
self.prelude(&assign);
self.js_arguments.push(format!("c{}", abi));
return Ok(());
}
_ => bail!(
"unsupported optional argument type for calling JS function from Rust: {:?}",
arg
),
2018-08-03 19:07:12 +03:00
};
Add support for optional slice types (#507) * Shard the `convert.rs` module into sub-modules Hopefully this'll make the organization a little nicer over time! * Start adding support for optional types This commit starts adding support for optional types to wasm-bindgen as arguments/return values to functions. The strategy here is to add two new traits, `OptionIntoWasmAbi` and `OptionFromWasmAbi`. These two traits are used as a blanket impl to implement `IntoWasmAbi` and `FromWasmAbi` for `Option<T>`. Some consequences of this design: * It should be possible to ensure `Option<SomeForeignType>` implements to/from wasm traits. This is because the option-based traits can be implemented for foreign types. * A specialized implementation is possible for all types, so there's no need for `Option<T>` to introduce unnecessary overhead. * Two new traits is a bit unforutnate but I can't currently think of an alternative design that works for the above two constraints, although it doesn't mean one doesn't exist! * The error messages for "can't use this type here" is actually halfway decent because it says these new traits need to be implemented, which provides a good place to document and talk about what's going on here! * Nested references like `Option<&T>` can't implement `FromWasmAbi`. This means that you can't define a function in Rust which takes `Option<&str>`. It may be possible to do this one day but it'll likely require more trait trickery than I'm capable of right now. * Add support for optional slices This commit adds support for optional slice types, things like strings and arrays. The null representation of these has a pointer value of 0, which should never happen in normal Rust. Otherwise the various plumbing is done throughout the tooling to enable these types in all locations. * Fix `takeObject` on global sentinels These don't have a reference count as they're always expected to work, so avoid actually dropping a reference on them. * Remove some no longer needed bindings * Add support for optional anyref types This commit adds support for optional imported class types. Each type imported with `#[wasm_bindgen]` automatically implements the relevant traits and now supports `Option<Foo>` in various argument/return positions. * Fix building without the `std` feature * Actually fix the build... * Add support for optional types to WebIDL Closes #502
2018-07-19 14:44:23 -05:00
}
2018-08-03 16:28:35 +03:00
if let Some(signed) = arg.get_64() {
let f = if signed {
self.cx.expose_int64_cvt_shim()
} else {
self.cx.expose_uint64_cvt_shim()
};
2018-08-03 16:28:35 +03:00
let high = self.shim_argument();
let name = format!("n{}", abi);
2018-06-27 22:42:34 -07:00
self.prelude(&format!(
"\
u32CvtShim[0] = {low};
u32CvtShim[1] = {high};
const {name} = {f}[0];
2018-06-27 22:42:34 -07:00
",
2018-08-03 16:28:35 +03:00
low = abi,
high = high,
f = f,
name = name,
));
self.js_arguments.push(name);
2018-06-27 22:42:34 -07:00
return Ok(());
}
if let Some(class) = arg.rust_struct() {
if arg.is_by_ref() {
bail!("cannot invoke JS functions with custom ref types yet")
}
self.cx.require_class_wrap(class);
let assign = format!("let c{0} = {1}.__wrap({0});", abi, class);
self.prelude(&assign);
self.js_arguments.push(format!("c{}", abi));
2018-06-27 22:42:34 -07:00
return Ok(());
}
if let Some((f, mutable)) = arg.stack_closure() {
let arg2 = self.shim_argument();
Add experimental support for the `anyref` type This commit adds experimental support to `wasm-bindgen` to emit and leverage the `anyref` native wasm type. This native type is still in a proposal status (the reference-types proposal). The intention of `anyref` is to be able to directly hold JS values in wasm and pass the to imported functions, namely to empower eventual host bindings (now renamed WebIDL bindings) integration where we can skip JS shims altogether for many imports. This commit doesn't actually affect wasm-bindgen's behavior at all as-is, but rather this support requires an opt-in env var to be configured. Once the support is stable in browsers it's intended that this will add a CLI switch for turning on this support, eventually defaulting it to `true` in the far future. The basic strategy here is to take the `stack` and `slab` globals in the generated JS glue and move them into wasm using a table. This new table in wasm is managed at the fringes via injected shims. At `wasm-bindgen`-time the CLI will rewrite exports and imports with shims that actually use `anyref` if needed, performing loads/stores inside the wasm module instead of externally in the wasm module. This should provide a boost over what we have today, but it's not a fantastic strategy long term. We have a more grand vision for `anyref` being a first-class type in the language, but that's on a much longer horizon and this is currently thought to be the best we can do in terms of integration in the near future. The stack/heap JS tables are combined into one wasm table. The stack starts at the end of the table and grows down with a stack pointer (also injected). The heap starts at the end and grows up (state managed in linear memory). The anyref transformation here will hook up various intrinsics in wasm-bindgen to the runtime functionality if the anyref supoprt is enabled. The main tricky treatment here was applied to closures, where we need JS to use a different function pointer than the one Rust gives it to use a JS function pointer empowered with anyref. This works by switching up a bit how descriptors work, embedding the shims to call inside descriptors rather than communicated at runtime. This means that we're accessing constant values in the generated JS and we can just update the constant value accessed.
2018-10-18 08:43:36 -07:00
let mut shim = f.shim_idx;
let (js, _ts, _js_doc) = {
let mut builder = Js2Rust::new("", self.cx);
if mutable {
2018-06-27 22:42:34 -07:00
builder
.prelude("let a = this.a;\n")
.prelude("this.a = 0;\n")
.rust_argument("a")
.finally("this.a = a;\n");
} else {
builder.rust_argument("this.a");
}
2019-03-14 08:46:42 -03:00
builder.rust_argument("this.b").process(f, None)?.finish(
Add experimental support for the `anyref` type This commit adds experimental support to `wasm-bindgen` to emit and leverage the `anyref` native wasm type. This native type is still in a proposal status (the reference-types proposal). The intention of `anyref` is to be able to directly hold JS values in wasm and pass the to imported functions, namely to empower eventual host bindings (now renamed WebIDL bindings) integration where we can skip JS shims altogether for many imports. This commit doesn't actually affect wasm-bindgen's behavior at all as-is, but rather this support requires an opt-in env var to be configured. Once the support is stable in browsers it's intended that this will add a CLI switch for turning on this support, eventually defaulting it to `true` in the far future. The basic strategy here is to take the `stack` and `slab` globals in the generated JS glue and move them into wasm using a table. This new table in wasm is managed at the fringes via injected shims. At `wasm-bindgen`-time the CLI will rewrite exports and imports with shims that actually use `anyref` if needed, performing loads/stores inside the wasm module instead of externally in the wasm module. This should provide a boost over what we have today, but it's not a fantastic strategy long term. We have a more grand vision for `anyref` being a first-class type in the language, but that's on a much longer horizon and this is currently thought to be the best we can do in terms of integration in the near future. The stack/heap JS tables are combined into one wasm table. The stack starts at the end of the table and grows down with a stack pointer (also injected). The heap starts at the end and grows up (state managed in linear memory). The anyref transformation here will hook up various intrinsics in wasm-bindgen to the runtime functionality if the anyref supoprt is enabled. The main tricky treatment here was applied to closures, where we need JS to use a different function pointer than the one Rust gives it to use a JS function pointer empowered with anyref. This works by switching up a bit how descriptors work, embedding the shims to call inside descriptors rather than communicated at runtime. This means that we're accessing constant values in the generated JS and we can just update the constant value accessed.
2018-10-18 08:43:36 -07:00
"function",
"this.f",
ExportedShim::TableElement(&mut shim),
)
};
self.cx.function_table_needed = true;
self.global_idx();
2018-06-27 22:42:34 -07:00
self.prelude(&format!(
"\
let cb{0} = {js};\n\
cb{0}.f = wasm.__wbg_function_table.get({idx});\n\
cb{0}.a = {0};\n\
cb{0}.b = {1};\n\
2018-06-27 22:42:34 -07:00
",
abi,
arg2,
2018-06-27 22:42:34 -07:00
js = js,
Add experimental support for the `anyref` type This commit adds experimental support to `wasm-bindgen` to emit and leverage the `anyref` native wasm type. This native type is still in a proposal status (the reference-types proposal). The intention of `anyref` is to be able to directly hold JS values in wasm and pass the to imported functions, namely to empower eventual host bindings (now renamed WebIDL bindings) integration where we can skip JS shims altogether for many imports. This commit doesn't actually affect wasm-bindgen's behavior at all as-is, but rather this support requires an opt-in env var to be configured. Once the support is stable in browsers it's intended that this will add a CLI switch for turning on this support, eventually defaulting it to `true` in the far future. The basic strategy here is to take the `stack` and `slab` globals in the generated JS glue and move them into wasm using a table. This new table in wasm is managed at the fringes via injected shims. At `wasm-bindgen`-time the CLI will rewrite exports and imports with shims that actually use `anyref` if needed, performing loads/stores inside the wasm module instead of externally in the wasm module. This should provide a boost over what we have today, but it's not a fantastic strategy long term. We have a more grand vision for `anyref` being a first-class type in the language, but that's on a much longer horizon and this is currently thought to be the best we can do in terms of integration in the near future. The stack/heap JS tables are combined into one wasm table. The stack starts at the end of the table and grows down with a stack pointer (also injected). The heap starts at the end and grows up (state managed in linear memory). The anyref transformation here will hook up various intrinsics in wasm-bindgen to the runtime functionality if the anyref supoprt is enabled. The main tricky treatment here was applied to closures, where we need JS to use a different function pointer than the one Rust gives it to use a JS function pointer empowered with anyref. This works by switching up a bit how descriptors work, embedding the shims to call inside descriptors rather than communicated at runtime. This means that we're accessing constant values in the generated JS and we can just update the constant value accessed.
2018-10-18 08:43:36 -07:00
idx = shim,
2018-06-27 22:42:34 -07:00
));
self.finally(&format!("cb{0}.a = cb{0}.b = 0;", abi));
self.js_arguments.push(format!("cb{0}.bind(cb{0})", abi));
2018-06-27 22:42:34 -07:00
return Ok(());
}
if let Some(num) = arg.number() {
if num.is_u32() {
self.js_arguments.push(format!("{} >>> 0", abi));
} else {
self.js_arguments.push(abi);
}
return Ok(());
}
let invoc_arg = match *arg {
Descriptor::Boolean => format!("{} !== 0", abi),
2018-06-27 22:42:34 -07:00
Descriptor::Char => format!("String.fromCodePoint({})", abi),
_ => bail!(
"unsupported argument type for calling JS function from Rust: {:?}",
arg
),
};
self.js_arguments.push(invoc_arg);
Ok(())
}
fn ret(&mut self, ty: &Descriptor) -> Result<(), Error> {
if let Descriptor::Unit = ty {
self.ret_expr = "JS;".to_string();
return Ok(());
}
Add support for optional slice types (#507) * Shard the `convert.rs` module into sub-modules Hopefully this'll make the organization a little nicer over time! * Start adding support for optional types This commit starts adding support for optional types to wasm-bindgen as arguments/return values to functions. The strategy here is to add two new traits, `OptionIntoWasmAbi` and `OptionFromWasmAbi`. These two traits are used as a blanket impl to implement `IntoWasmAbi` and `FromWasmAbi` for `Option<T>`. Some consequences of this design: * It should be possible to ensure `Option<SomeForeignType>` implements to/from wasm traits. This is because the option-based traits can be implemented for foreign types. * A specialized implementation is possible for all types, so there's no need for `Option<T>` to introduce unnecessary overhead. * Two new traits is a bit unforutnate but I can't currently think of an alternative design that works for the above two constraints, although it doesn't mean one doesn't exist! * The error messages for "can't use this type here" is actually halfway decent because it says these new traits need to be implemented, which provides a good place to document and talk about what's going on here! * Nested references like `Option<&T>` can't implement `FromWasmAbi`. This means that you can't define a function in Rust which takes `Option<&str>`. It may be possible to do this one day but it'll likely require more trait trickery than I'm capable of right now. * Add support for optional slices This commit adds support for optional slice types, things like strings and arrays. The null representation of these has a pointer value of 0, which should never happen in normal Rust. Otherwise the various plumbing is done throughout the tooling to enable these types in all locations. * Fix `takeObject` on global sentinels These don't have a reference count as they're always expected to work, so avoid actually dropping a reference on them. * Remove some no longer needed bindings * Add support for optional anyref types This commit adds support for optional imported class types. Each type imported with `#[wasm_bindgen]` automatically implements the relevant traits and now supports `Option<Foo>` in various argument/return positions. * Fix building without the `std` feature * Actually fix the build... * Add support for optional types to WebIDL Closes #502
2018-07-19 14:44:23 -05:00
let (ty, optional) = match ty {
Descriptor::Option(t) => (&**t, true),
_ => (ty, false),
};
if ty.is_by_ref() {
bail!("cannot return a reference from JS to Rust")
}
if let Some(ty) = ty.vector_kind() {
let f = self.cx.pass_to_wasm_function(ty)?;
self.cx.expose_uint32_memory();
self.shim_arguments.insert(0, "ret".to_string());
Add support for optional slice types (#507) * Shard the `convert.rs` module into sub-modules Hopefully this'll make the organization a little nicer over time! * Start adding support for optional types This commit starts adding support for optional types to wasm-bindgen as arguments/return values to functions. The strategy here is to add two new traits, `OptionIntoWasmAbi` and `OptionFromWasmAbi`. These two traits are used as a blanket impl to implement `IntoWasmAbi` and `FromWasmAbi` for `Option<T>`. Some consequences of this design: * It should be possible to ensure `Option<SomeForeignType>` implements to/from wasm traits. This is because the option-based traits can be implemented for foreign types. * A specialized implementation is possible for all types, so there's no need for `Option<T>` to introduce unnecessary overhead. * Two new traits is a bit unforutnate but I can't currently think of an alternative design that works for the above two constraints, although it doesn't mean one doesn't exist! * The error messages for "can't use this type here" is actually halfway decent because it says these new traits need to be implemented, which provides a good place to document and talk about what's going on here! * Nested references like `Option<&T>` can't implement `FromWasmAbi`. This means that you can't define a function in Rust which takes `Option<&str>`. It may be possible to do this one day but it'll likely require more trait trickery than I'm capable of right now. * Add support for optional slices This commit adds support for optional slice types, things like strings and arrays. The null representation of these has a pointer value of 0, which should never happen in normal Rust. Otherwise the various plumbing is done throughout the tooling to enable these types in all locations. * Fix `takeObject` on global sentinels These don't have a reference count as they're always expected to work, so avoid actually dropping a reference on them. * Remove some no longer needed bindings * Add support for optional anyref types This commit adds support for optional imported class types. Each type imported with `#[wasm_bindgen]` automatically implements the relevant traits and now supports `Option<Foo>` in various argument/return positions. * Fix building without the `std` feature * Actually fix the build... * Add support for optional types to WebIDL Closes #502
2018-07-19 14:44:23 -05:00
let mut prelude = String::new();
let expr = if optional {
prelude.push_str("const val = JS;");
self.cx.expose_is_like_none();
format!("isLikeNone(val) ? [0, 0] : {}(val)", f)
} else {
format!("{}(JS)", f)
};
2018-06-27 22:42:34 -07:00
self.ret_expr = format!(
"\
Add support for optional slice types (#507) * Shard the `convert.rs` module into sub-modules Hopefully this'll make the organization a little nicer over time! * Start adding support for optional types This commit starts adding support for optional types to wasm-bindgen as arguments/return values to functions. The strategy here is to add two new traits, `OptionIntoWasmAbi` and `OptionFromWasmAbi`. These two traits are used as a blanket impl to implement `IntoWasmAbi` and `FromWasmAbi` for `Option<T>`. Some consequences of this design: * It should be possible to ensure `Option<SomeForeignType>` implements to/from wasm traits. This is because the option-based traits can be implemented for foreign types. * A specialized implementation is possible for all types, so there's no need for `Option<T>` to introduce unnecessary overhead. * Two new traits is a bit unforutnate but I can't currently think of an alternative design that works for the above two constraints, although it doesn't mean one doesn't exist! * The error messages for "can't use this type here" is actually halfway decent because it says these new traits need to be implemented, which provides a good place to document and talk about what's going on here! * Nested references like `Option<&T>` can't implement `FromWasmAbi`. This means that you can't define a function in Rust which takes `Option<&str>`. It may be possible to do this one day but it'll likely require more trait trickery than I'm capable of right now. * Add support for optional slices This commit adds support for optional slice types, things like strings and arrays. The null representation of these has a pointer value of 0, which should never happen in normal Rust. Otherwise the various plumbing is done throughout the tooling to enable these types in all locations. * Fix `takeObject` on global sentinels These don't have a reference count as they're always expected to work, so avoid actually dropping a reference on them. * Remove some no longer needed bindings * Add support for optional anyref types This commit adds support for optional imported class types. Each type imported with `#[wasm_bindgen]` automatically implements the relevant traits and now supports `Option<Foo>` in various argument/return positions. * Fix building without the `std` feature * Actually fix the build... * Add support for optional types to WebIDL Closes #502
2018-07-19 14:44:23 -05:00
{}
const retptr = {};
const retlen = WASM_VECTOR_LEN;
const mem = getUint32Memory();
mem[ret / 4] = retptr;
mem[ret / 4 + 1] = retlen;
",
prelude, expr
2018-06-27 22:42:34 -07:00
);
return Ok(());
}
Add support for optional slice types (#507) * Shard the `convert.rs` module into sub-modules Hopefully this'll make the organization a little nicer over time! * Start adding support for optional types This commit starts adding support for optional types to wasm-bindgen as arguments/return values to functions. The strategy here is to add two new traits, `OptionIntoWasmAbi` and `OptionFromWasmAbi`. These two traits are used as a blanket impl to implement `IntoWasmAbi` and `FromWasmAbi` for `Option<T>`. Some consequences of this design: * It should be possible to ensure `Option<SomeForeignType>` implements to/from wasm traits. This is because the option-based traits can be implemented for foreign types. * A specialized implementation is possible for all types, so there's no need for `Option<T>` to introduce unnecessary overhead. * Two new traits is a bit unforutnate but I can't currently think of an alternative design that works for the above two constraints, although it doesn't mean one doesn't exist! * The error messages for "can't use this type here" is actually halfway decent because it says these new traits need to be implemented, which provides a good place to document and talk about what's going on here! * Nested references like `Option<&T>` can't implement `FromWasmAbi`. This means that you can't define a function in Rust which takes `Option<&str>`. It may be possible to do this one day but it'll likely require more trait trickery than I'm capable of right now. * Add support for optional slices This commit adds support for optional slice types, things like strings and arrays. The null representation of these has a pointer value of 0, which should never happen in normal Rust. Otherwise the various plumbing is done throughout the tooling to enable these types in all locations. * Fix `takeObject` on global sentinels These don't have a reference count as they're always expected to work, so avoid actually dropping a reference on them. * Remove some no longer needed bindings * Add support for optional anyref types This commit adds support for optional imported class types. Each type imported with `#[wasm_bindgen]` automatically implements the relevant traits and now supports `Option<Foo>` in various argument/return positions. * Fix building without the `std` feature * Actually fix the build... * Add support for optional types to WebIDL Closes #502
2018-07-19 14:44:23 -05:00
if ty.is_anyref() {
Add experimental support for the `anyref` type This commit adds experimental support to `wasm-bindgen` to emit and leverage the `anyref` native wasm type. This native type is still in a proposal status (the reference-types proposal). The intention of `anyref` is to be able to directly hold JS values in wasm and pass the to imported functions, namely to empower eventual host bindings (now renamed WebIDL bindings) integration where we can skip JS shims altogether for many imports. This commit doesn't actually affect wasm-bindgen's behavior at all as-is, but rather this support requires an opt-in env var to be configured. Once the support is stable in browsers it's intended that this will add a CLI switch for turning on this support, eventually defaulting it to `true` in the far future. The basic strategy here is to take the `stack` and `slab` globals in the generated JS glue and move them into wasm using a table. This new table in wasm is managed at the fringes via injected shims. At `wasm-bindgen`-time the CLI will rewrite exports and imports with shims that actually use `anyref` if needed, performing loads/stores inside the wasm module instead of externally in the wasm module. This should provide a boost over what we have today, but it's not a fantastic strategy long term. We have a more grand vision for `anyref` being a first-class type in the language, but that's on a much longer horizon and this is currently thought to be the best we can do in terms of integration in the near future. The stack/heap JS tables are combined into one wasm table. The stack starts at the end of the table and grows down with a stack pointer (also injected). The heap starts at the end and grows up (state managed in linear memory). The anyref transformation here will hook up various intrinsics in wasm-bindgen to the runtime functionality if the anyref supoprt is enabled. The main tricky treatment here was applied to closures, where we need JS to use a different function pointer than the one Rust gives it to use a JS function pointer empowered with anyref. This works by switching up a bit how descriptors work, embedding the shims to call inside descriptors rather than communicated at runtime. This means that we're accessing constant values in the generated JS and we can just update the constant value accessed.
2018-10-18 08:43:36 -07:00
if self.cx.config.anyref {
if optional {
self.cx.expose_add_to_anyref_table()?;
self.cx.expose_is_like_none();
self.ret_expr = "
const val = JS;
return isLikeNone(val) ? 0 : addToAnyrefTable(val);
"
.to_string();
} else {
self.ret_anyref = true;
self.ret_expr = "return JS;".to_string()
}
Add support for optional slice types (#507) * Shard the `convert.rs` module into sub-modules Hopefully this'll make the organization a little nicer over time! * Start adding support for optional types This commit starts adding support for optional types to wasm-bindgen as arguments/return values to functions. The strategy here is to add two new traits, `OptionIntoWasmAbi` and `OptionFromWasmAbi`. These two traits are used as a blanket impl to implement `IntoWasmAbi` and `FromWasmAbi` for `Option<T>`. Some consequences of this design: * It should be possible to ensure `Option<SomeForeignType>` implements to/from wasm traits. This is because the option-based traits can be implemented for foreign types. * A specialized implementation is possible for all types, so there's no need for `Option<T>` to introduce unnecessary overhead. * Two new traits is a bit unforutnate but I can't currently think of an alternative design that works for the above two constraints, although it doesn't mean one doesn't exist! * The error messages for "can't use this type here" is actually halfway decent because it says these new traits need to be implemented, which provides a good place to document and talk about what's going on here! * Nested references like `Option<&T>` can't implement `FromWasmAbi`. This means that you can't define a function in Rust which takes `Option<&str>`. It may be possible to do this one day but it'll likely require more trait trickery than I'm capable of right now. * Add support for optional slices This commit adds support for optional slice types, things like strings and arrays. The null representation of these has a pointer value of 0, which should never happen in normal Rust. Otherwise the various plumbing is done throughout the tooling to enable these types in all locations. * Fix `takeObject` on global sentinels These don't have a reference count as they're always expected to work, so avoid actually dropping a reference on them. * Remove some no longer needed bindings * Add support for optional anyref types This commit adds support for optional imported class types. Each type imported with `#[wasm_bindgen]` automatically implements the relevant traits and now supports `Option<Foo>` in various argument/return positions. * Fix building without the `std` feature * Actually fix the build... * Add support for optional types to WebIDL Closes #502
2018-07-19 14:44:23 -05:00
} else {
Add experimental support for the `anyref` type This commit adds experimental support to `wasm-bindgen` to emit and leverage the `anyref` native wasm type. This native type is still in a proposal status (the reference-types proposal). The intention of `anyref` is to be able to directly hold JS values in wasm and pass the to imported functions, namely to empower eventual host bindings (now renamed WebIDL bindings) integration where we can skip JS shims altogether for many imports. This commit doesn't actually affect wasm-bindgen's behavior at all as-is, but rather this support requires an opt-in env var to be configured. Once the support is stable in browsers it's intended that this will add a CLI switch for turning on this support, eventually defaulting it to `true` in the far future. The basic strategy here is to take the `stack` and `slab` globals in the generated JS glue and move them into wasm using a table. This new table in wasm is managed at the fringes via injected shims. At `wasm-bindgen`-time the CLI will rewrite exports and imports with shims that actually use `anyref` if needed, performing loads/stores inside the wasm module instead of externally in the wasm module. This should provide a boost over what we have today, but it's not a fantastic strategy long term. We have a more grand vision for `anyref` being a first-class type in the language, but that's on a much longer horizon and this is currently thought to be the best we can do in terms of integration in the near future. The stack/heap JS tables are combined into one wasm table. The stack starts at the end of the table and grows down with a stack pointer (also injected). The heap starts at the end and grows up (state managed in linear memory). The anyref transformation here will hook up various intrinsics in wasm-bindgen to the runtime functionality if the anyref supoprt is enabled. The main tricky treatment here was applied to closures, where we need JS to use a different function pointer than the one Rust gives it to use a JS function pointer empowered with anyref. This works by switching up a bit how descriptors work, embedding the shims to call inside descriptors rather than communicated at runtime. This means that we're accessing constant values in the generated JS and we can just update the constant value accessed.
2018-10-18 08:43:36 -07:00
self.cx.expose_add_heap_object();
if optional {
self.cx.expose_is_like_none();
self.ret_expr = "
const val = JS;
return isLikeNone(val) ? 0 : addHeapObject(val);
"
.to_string();
} else {
self.ret_expr = "return addHeapObject(JS);".to_string()
}
Add support for optional slice types (#507) * Shard the `convert.rs` module into sub-modules Hopefully this'll make the organization a little nicer over time! * Start adding support for optional types This commit starts adding support for optional types to wasm-bindgen as arguments/return values to functions. The strategy here is to add two new traits, `OptionIntoWasmAbi` and `OptionFromWasmAbi`. These two traits are used as a blanket impl to implement `IntoWasmAbi` and `FromWasmAbi` for `Option<T>`. Some consequences of this design: * It should be possible to ensure `Option<SomeForeignType>` implements to/from wasm traits. This is because the option-based traits can be implemented for foreign types. * A specialized implementation is possible for all types, so there's no need for `Option<T>` to introduce unnecessary overhead. * Two new traits is a bit unforutnate but I can't currently think of an alternative design that works for the above two constraints, although it doesn't mean one doesn't exist! * The error messages for "can't use this type here" is actually halfway decent because it says these new traits need to be implemented, which provides a good place to document and talk about what's going on here! * Nested references like `Option<&T>` can't implement `FromWasmAbi`. This means that you can't define a function in Rust which takes `Option<&str>`. It may be possible to do this one day but it'll likely require more trait trickery than I'm capable of right now. * Add support for optional slices This commit adds support for optional slice types, things like strings and arrays. The null representation of these has a pointer value of 0, which should never happen in normal Rust. Otherwise the various plumbing is done throughout the tooling to enable these types in all locations. * Fix `takeObject` on global sentinels These don't have a reference count as they're always expected to work, so avoid actually dropping a reference on them. * Remove some no longer needed bindings * Add support for optional anyref types This commit adds support for optional imported class types. Each type imported with `#[wasm_bindgen]` automatically implements the relevant traits and now supports `Option<Foo>` in various argument/return positions. * Fix building without the `std` feature * Actually fix the build... * Add support for optional types to WebIDL Closes #502
2018-07-19 14:44:23 -05:00
}
return Ok(());
Add support for optional slice types (#507) * Shard the `convert.rs` module into sub-modules Hopefully this'll make the organization a little nicer over time! * Start adding support for optional types This commit starts adding support for optional types to wasm-bindgen as arguments/return values to functions. The strategy here is to add two new traits, `OptionIntoWasmAbi` and `OptionFromWasmAbi`. These two traits are used as a blanket impl to implement `IntoWasmAbi` and `FromWasmAbi` for `Option<T>`. Some consequences of this design: * It should be possible to ensure `Option<SomeForeignType>` implements to/from wasm traits. This is because the option-based traits can be implemented for foreign types. * A specialized implementation is possible for all types, so there's no need for `Option<T>` to introduce unnecessary overhead. * Two new traits is a bit unforutnate but I can't currently think of an alternative design that works for the above two constraints, although it doesn't mean one doesn't exist! * The error messages for "can't use this type here" is actually halfway decent because it says these new traits need to be implemented, which provides a good place to document and talk about what's going on here! * Nested references like `Option<&T>` can't implement `FromWasmAbi`. This means that you can't define a function in Rust which takes `Option<&str>`. It may be possible to do this one day but it'll likely require more trait trickery than I'm capable of right now. * Add support for optional slices This commit adds support for optional slice types, things like strings and arrays. The null representation of these has a pointer value of 0, which should never happen in normal Rust. Otherwise the various plumbing is done throughout the tooling to enable these types in all locations. * Fix `takeObject` on global sentinels These don't have a reference count as they're always expected to work, so avoid actually dropping a reference on them. * Remove some no longer needed bindings * Add support for optional anyref types This commit adds support for optional imported class types. Each type imported with `#[wasm_bindgen]` automatically implements the relevant traits and now supports `Option<Foo>` in various argument/return positions. * Fix building without the `std` feature * Actually fix the build... * Add support for optional types to WebIDL Closes #502
2018-07-19 14:44:23 -05:00
}
if optional {
self.cx.expose_is_like_none();
if ty.is_wasm_native() {
2018-08-03 16:28:35 +03:00
self.cx.expose_uint32_memory();
match ty {
Descriptor::I32 => self.cx.expose_int32_memory(),
Descriptor::U32 => (),
Descriptor::F32 => self.cx.expose_f32_memory(),
Descriptor::F64 => self.cx.expose_f64_memory(),
_ => (),
};
self.shim_arguments.insert(0, "ret".to_string());
self.ret_expr = format!(
"
const val = JS;
getUint32Memory()[ret / 4] = !isLikeNone(val);
{mem}[ret / {size} + 1] = isLikeNone(val) ? 0 : val;
2018-08-03 16:28:35 +03:00
",
size = match ty {
Descriptor::I32 => 4,
Descriptor::U32 => 4,
Descriptor::F32 => 4,
Descriptor::F64 => 8,
_ => unreachable!(),
},
mem = match ty {
Descriptor::I32 => "getInt32Memory()",
Descriptor::U32 => "getUint32Memory()",
Descriptor::F32 => "getFloat32Memory()",
Descriptor::F64 => "getFloat64Memory()",
_ => unreachable!(),
}
);
return Ok(());
}
if ty.is_abi_as_u32() {
2018-08-03 16:28:35 +03:00
self.ret_expr = "
const val = JS;
return isLikeNone(val) ? 0xFFFFFF : val;
2018-11-27 12:07:59 -08:00
"
.to_string();
2018-08-03 16:28:35 +03:00
return Ok(());
}
if let Some(signed) = ty.get_64() {
self.cx.expose_uint32_memory();
let f = if signed {
self.cx.expose_int64_memory();
"getInt64Memory"
} else {
self.cx.expose_uint64_memory();
"getUint64Memory"
};
self.shim_arguments.insert(0, "ret".to_string());
self.ret_expr = format!(
"
const val = JS;
getUint32Memory()[ret / 4] = !isLikeNone(val);
{}()[ret / 8 + 1] = isLikeNone(val) ? BigInt(0) : val;
2018-08-03 16:28:35 +03:00
",
f
);
return Ok(());
}
2018-08-03 19:07:12 +03:00
match *ty {
Descriptor::Boolean => {
self.ret_expr = "
const val = JS;
return isLikeNone(val) ? 0xFFFFFF : val ? 1 : 0;
2018-11-27 12:07:59 -08:00
"
.to_string();
}
2018-08-03 20:45:57 +03:00
Descriptor::Char => {
self.ret_expr = "
const val = JS;
return isLikeNone(val) ? 0xFFFFFF : val.codePointAt(0);
2018-11-27 12:07:59 -08:00
"
.to_string();
}
Descriptor::Enum { hole } => {
Migrate `wasm-bindgen` to using `walrus` This commit moves `wasm-bindgen` the CLI tool from internally using `parity-wasm` for wasm parsing/serialization to instead use `walrus`. The `walrus` crate is something we've been working on recently with an aim to replace the usage of `parity-wasm` in `wasm-bindgen` to make the current CLI tool more maintainable as well as more future-proof. The `walrus` crate provides a much nicer AST to work with as well as a structured `Module`, whereas `parity-wasm` provides a very raw interface to the wasm module which isn't really appropriate for our use case. The many transformations and tweaks that wasm-bindgen does have a huge amount of ad-hoc index management to carefully craft a final wasm binary, but this is all entirely taken care for us with the `walrus` crate. Additionally, `wasm-bindgen` will ingest and rewrite the wasm file, often changing the binary offsets of functions. Eventually with DWARF debug information we'll need to be sure to preserve the debug information throughout the transformations that `wasm-bindgen` does today. This is practically impossible to do with the `parity-wasm` architecture, but `walrus` was designed from the get-go to solve this problem transparently in the `walrus` crate itself. (it doesn't today, but this is planned work) It is the intention that this does not end up regressing any `wasm-bindgen` use cases, neither in functionality or in speed. As a large change and refactoring, however, it's likely that at least something will arise! We'll want to continue to remain vigilant to any issues that come up with this commit. Note that the `gc` crate has been deleted as part of this change, as the `gc` crate is no longer necessary since `walrus` does it automatically. Additionally the `gc` crate was one of the main problems with preserving debug information as it often deletes wasm items! Finally, this also starts moving crates to the 2018 edition where necessary since `walrus` requires the 2018 edition, and in general it's more pleasant to work within the 2018 edition!
2019-01-31 09:54:23 -08:00
self.ret_expr = format!(
"
const val = JS;
return isLikeNone(val) ? {} : val;
Migrate `wasm-bindgen` to using `walrus` This commit moves `wasm-bindgen` the CLI tool from internally using `parity-wasm` for wasm parsing/serialization to instead use `walrus`. The `walrus` crate is something we've been working on recently with an aim to replace the usage of `parity-wasm` in `wasm-bindgen` to make the current CLI tool more maintainable as well as more future-proof. The `walrus` crate provides a much nicer AST to work with as well as a structured `Module`, whereas `parity-wasm` provides a very raw interface to the wasm module which isn't really appropriate for our use case. The many transformations and tweaks that wasm-bindgen does have a huge amount of ad-hoc index management to carefully craft a final wasm binary, but this is all entirely taken care for us with the `walrus` crate. Additionally, `wasm-bindgen` will ingest and rewrite the wasm file, often changing the binary offsets of functions. Eventually with DWARF debug information we'll need to be sure to preserve the debug information throughout the transformations that `wasm-bindgen` does today. This is practically impossible to do with the `parity-wasm` architecture, but `walrus` was designed from the get-go to solve this problem transparently in the `walrus` crate itself. (it doesn't today, but this is planned work) It is the intention that this does not end up regressing any `wasm-bindgen` use cases, neither in functionality or in speed. As a large change and refactoring, however, it's likely that at least something will arise! We'll want to continue to remain vigilant to any issues that come up with this commit. Note that the `gc` crate has been deleted as part of this change, as the `gc` crate is no longer necessary since `walrus` does it automatically. Additionally the `gc` crate was one of the main problems with preserving debug information as it often deletes wasm items! Finally, this also starts moving crates to the 2018 edition where necessary since `walrus` requires the 2018 edition, and in general it's more pleasant to work within the 2018 edition!
2019-01-31 09:54:23 -08:00
",
hole
);
}
Descriptor::RustStruct(ref class) => {
// Like below, assert the type
self.ret_expr = format!(
"\
const val = JS;
if (isLikeNone(val))
return 0;
if (!(val instanceof {0})) {{
throw new Error('expected value of type {0}');
}}
const ret = val.ptr;
val.ptr = 0;
return ret;\
",
class
);
}
_ => bail!(
"unsupported optional return type for calling JS function from Rust: {:?}",
ty
),
2018-08-03 19:07:12 +03:00
};
return Ok(());
Add support for optional slice types (#507) * Shard the `convert.rs` module into sub-modules Hopefully this'll make the organization a little nicer over time! * Start adding support for optional types This commit starts adding support for optional types to wasm-bindgen as arguments/return values to functions. The strategy here is to add two new traits, `OptionIntoWasmAbi` and `OptionFromWasmAbi`. These two traits are used as a blanket impl to implement `IntoWasmAbi` and `FromWasmAbi` for `Option<T>`. Some consequences of this design: * It should be possible to ensure `Option<SomeForeignType>` implements to/from wasm traits. This is because the option-based traits can be implemented for foreign types. * A specialized implementation is possible for all types, so there's no need for `Option<T>` to introduce unnecessary overhead. * Two new traits is a bit unforutnate but I can't currently think of an alternative design that works for the above two constraints, although it doesn't mean one doesn't exist! * The error messages for "can't use this type here" is actually halfway decent because it says these new traits need to be implemented, which provides a good place to document and talk about what's going on here! * Nested references like `Option<&T>` can't implement `FromWasmAbi`. This means that you can't define a function in Rust which takes `Option<&str>`. It may be possible to do this one day but it'll likely require more trait trickery than I'm capable of right now. * Add support for optional slices This commit adds support for optional slice types, things like strings and arrays. The null representation of these has a pointer value of 0, which should never happen in normal Rust. Otherwise the various plumbing is done throughout the tooling to enable these types in all locations. * Fix `takeObject` on global sentinels These don't have a reference count as they're always expected to work, so avoid actually dropping a reference on them. * Remove some no longer needed bindings * Add support for optional anyref types This commit adds support for optional imported class types. Each type imported with `#[wasm_bindgen]` automatically implements the relevant traits and now supports `Option<Foo>` in various argument/return positions. * Fix building without the `std` feature * Actually fix the build... * Add support for optional types to WebIDL Closes #502
2018-07-19 14:44:23 -05:00
}
if ty.number().is_some() {
self.ret_expr = "return JS;".to_string();
2018-06-27 22:42:34 -07:00
return Ok(());
}
2018-08-03 16:28:35 +03:00
if let Some(signed) = ty.get_64() {
let f = if signed {
self.cx.expose_int64_memory();
"getInt64Memory"
} else {
self.cx.expose_uint64_memory();
"getUint64Memory"
};
self.shim_arguments.insert(0, "ret".to_string());
2018-06-27 22:42:34 -07:00
self.ret_expr = format!(
"\
const val = JS;\n\
{}()[ret / 8] = val;\n\
",
f
);
return Ok(());
}
if let Some(class) = ty.rust_struct() {
if ty.is_by_ref() {
bail!("cannot invoke JS functions returning custom ref types yet")
}
// Insert an assertion to the type of the returned value as
// otherwise this will cause memory unsafety on the Rust side of
// things.
self.ret_expr = format!(
"\
const val = JS;
if (!(val instanceof {0})) {{
throw new Error('expected value of type {0}');
}}
const ret = val.ptr;
val.ptr = 0;
return ret;\
",
class
);
return Ok(());
}
self.ret_expr = match *ty {
Descriptor::Boolean => "return JS;".to_string(),
2018-06-15 12:55:37 -05:00
Descriptor::Char => "return JS.codePointAt(0);".to_string(),
_ => bail!(
"unsupported return type for calling JS function from Rust: {:?}",
ty
),
};
Ok(())
}
/// Returns whether this shim won't actually do anything when called other
/// than forward the invocation somewhere else.
///
/// This is used as an optimization to wire up imports directly where
/// possible and avoid a shim in some circumstances.
pub fn is_noop(&self) -> bool {
let Rust2Js {
// fields which may affect whether we do nontrivial work
catch,
catch_and_rethrow,
finally,
js_arguments,
prelude,
ret_expr,
variadic,
shim_arguments,
// all other fields, listed explicitly here so if one is added we'll
// trigger a nonexhaustive error.
arg_idx: _,
cx: _,
global_idx: _,
Add experimental support for the `anyref` type This commit adds experimental support to `wasm-bindgen` to emit and leverage the `anyref` native wasm type. This native type is still in a proposal status (the reference-types proposal). The intention of `anyref` is to be able to directly hold JS values in wasm and pass the to imported functions, namely to empower eventual host bindings (now renamed WebIDL bindings) integration where we can skip JS shims altogether for many imports. This commit doesn't actually affect wasm-bindgen's behavior at all as-is, but rather this support requires an opt-in env var to be configured. Once the support is stable in browsers it's intended that this will add a CLI switch for turning on this support, eventually defaulting it to `true` in the far future. The basic strategy here is to take the `stack` and `slab` globals in the generated JS glue and move them into wasm using a table. This new table in wasm is managed at the fringes via injected shims. At `wasm-bindgen`-time the CLI will rewrite exports and imports with shims that actually use `anyref` if needed, performing loads/stores inside the wasm module instead of externally in the wasm module. This should provide a boost over what we have today, but it's not a fantastic strategy long term. We have a more grand vision for `anyref` being a first-class type in the language, but that's on a much longer horizon and this is currently thought to be the best we can do in terms of integration in the near future. The stack/heap JS tables are combined into one wasm table. The stack starts at the end of the table and grows down with a stack pointer (also injected). The heap starts at the end and grows up (state managed in linear memory). The anyref transformation here will hook up various intrinsics in wasm-bindgen to the runtime functionality if the anyref supoprt is enabled. The main tricky treatment here was applied to closures, where we need JS to use a different function pointer than the one Rust gives it to use a JS function pointer empowered with anyref. This works by switching up a bit how descriptors work, embedding the shims to call inside descriptors rather than communicated at runtime. This means that we're accessing constant values in the generated JS and we can just update the constant value accessed.
2018-10-18 08:43:36 -07:00
anyref_args: _,
ret_anyref: _,
} = self;
!catch &&
!catch_and_rethrow &&
!variadic &&
prelude.is_empty() &&
finally.is_empty() &&
// make sure our faux return expression is "simple" by not
// performing any sort of transformation on the return value
(ret_expr == "JS;" || ret_expr == "return JS;") &&
// similarly we want to make sure that all the arguments are simply
// forwarded from the shim we would generate to the import,
// requiring no transformations
js_arguments == shim_arguments
}
Add experimental support for the `anyref` type This commit adds experimental support to `wasm-bindgen` to emit and leverage the `anyref` native wasm type. This native type is still in a proposal status (the reference-types proposal). The intention of `anyref` is to be able to directly hold JS values in wasm and pass the to imported functions, namely to empower eventual host bindings (now renamed WebIDL bindings) integration where we can skip JS shims altogether for many imports. This commit doesn't actually affect wasm-bindgen's behavior at all as-is, but rather this support requires an opt-in env var to be configured. Once the support is stable in browsers it's intended that this will add a CLI switch for turning on this support, eventually defaulting it to `true` in the far future. The basic strategy here is to take the `stack` and `slab` globals in the generated JS glue and move them into wasm using a table. This new table in wasm is managed at the fringes via injected shims. At `wasm-bindgen`-time the CLI will rewrite exports and imports with shims that actually use `anyref` if needed, performing loads/stores inside the wasm module instead of externally in the wasm module. This should provide a boost over what we have today, but it's not a fantastic strategy long term. We have a more grand vision for `anyref` being a first-class type in the language, but that's on a much longer horizon and this is currently thought to be the best we can do in terms of integration in the near future. The stack/heap JS tables are combined into one wasm table. The stack starts at the end of the table and grows down with a stack pointer (also injected). The heap starts at the end and grows up (state managed in linear memory). The anyref transformation here will hook up various intrinsics in wasm-bindgen to the runtime functionality if the anyref supoprt is enabled. The main tricky treatment here was applied to closures, where we need JS to use a different function pointer than the one Rust gives it to use a JS function pointer empowered with anyref. This works by switching up a bit how descriptors work, embedding the shims to call inside descriptors rather than communicated at runtime. This means that we're accessing constant values in the generated JS and we can just update the constant value accessed.
2018-10-18 08:43:36 -07:00
pub fn finish(&mut self, invoc: &ImportTarget, shim: &str) -> Result<String, Error> {
let mut ret = String::new();
ret.push_str("function(");
ret.push_str(&self.shim_arguments.join(", "));
if self.catch {
if self.shim_arguments.len() > 0 {
ret.push_str(", ")
}
ret.push_str("exnptr");
}
ret.push_str(") {\n");
2018-06-15 12:55:37 -05:00
ret.push_str(&self.prelude);
let variadic = self.variadic;
let ret_expr = &self.ret_expr;
let handle_variadic = |invoc: &str, js_arguments: &[String]| {
let ret = if variadic {
let (last_arg, args) = match js_arguments.split_last() {
Some(pair) => pair,
None => bail!("a function with no arguments cannot be variadic"),
};
if args.len() > 0 {
ret_expr.replace(
"JS",
&format!("{}({}, ...{})", invoc, args.join(", "), last_arg),
)
} else {
2018-11-30 13:04:05 -08:00
ret_expr.replace("JS", &format!("{}(...{})", invoc, last_arg))
}
} else {
2018-11-30 13:04:05 -08:00
ret_expr.replace("JS", &format!("{}({})", invoc, js_arguments.join(", ")))
};
Ok(ret)
};
Add experimental support for the `anyref` type This commit adds experimental support to `wasm-bindgen` to emit and leverage the `anyref` native wasm type. This native type is still in a proposal status (the reference-types proposal). The intention of `anyref` is to be able to directly hold JS values in wasm and pass the to imported functions, namely to empower eventual host bindings (now renamed WebIDL bindings) integration where we can skip JS shims altogether for many imports. This commit doesn't actually affect wasm-bindgen's behavior at all as-is, but rather this support requires an opt-in env var to be configured. Once the support is stable in browsers it's intended that this will add a CLI switch for turning on this support, eventually defaulting it to `true` in the far future. The basic strategy here is to take the `stack` and `slab` globals in the generated JS glue and move them into wasm using a table. This new table in wasm is managed at the fringes via injected shims. At `wasm-bindgen`-time the CLI will rewrite exports and imports with shims that actually use `anyref` if needed, performing loads/stores inside the wasm module instead of externally in the wasm module. This should provide a boost over what we have today, but it's not a fantastic strategy long term. We have a more grand vision for `anyref` being a first-class type in the language, but that's on a much longer horizon and this is currently thought to be the best we can do in terms of integration in the near future. The stack/heap JS tables are combined into one wasm table. The stack starts at the end of the table and grows down with a stack pointer (also injected). The heap starts at the end and grows up (state managed in linear memory). The anyref transformation here will hook up various intrinsics in wasm-bindgen to the runtime functionality if the anyref supoprt is enabled. The main tricky treatment here was applied to closures, where we need JS to use a different function pointer than the one Rust gives it to use a JS function pointer empowered with anyref. This works by switching up a bit how descriptors work, embedding the shims to call inside descriptors rather than communicated at runtime. This means that we're accessing constant values in the generated JS and we can just update the constant value accessed.
2018-10-18 08:43:36 -07:00
let js_arguments = &self.js_arguments;
let fixed = |desc: &str, class: &Option<String>, amt: usize| {
if variadic {
bail!("{} cannot be variadic", desc);
}
match (class, js_arguments.len()) {
2018-11-30 13:04:05 -08:00
(None, n) if n == amt + 1 => Ok((js_arguments[0].clone(), &js_arguments[1..])),
(None, _) => bail!("setters must have {} arguments", amt + 1),
(Some(class), n) if n == amt => Ok((class.clone(), &js_arguments[..])),
(Some(_), _) => bail!("static setters must have {} arguments", amt),
}
2018-08-19 13:39:16 +01:00
};
let mut invoc = match invoc {
2018-11-27 12:07:59 -08:00
ImportTarget::Function(f) => handle_variadic(&f, &self.js_arguments)?,
ImportTarget::Constructor(c) => {
handle_variadic(&format!("new {}", c), &self.js_arguments)?
}
2018-11-27 12:07:59 -08:00
ImportTarget::Method(f) => handle_variadic(&format!("{}.call", f), &self.js_arguments)?,
ImportTarget::StructuralMethod(f) => {
let (receiver, args) = match self.js_arguments.split_first() {
Some(pair) => pair,
None => bail!("methods must have at least one argument"),
};
handle_variadic(&format!("{}.{}", receiver, f), args)?
}
ImportTarget::StructuralGetter(class, field) => {
let (receiver, _) = fixed("getter", class, 0)?;
let expr = format!("{}.{}", receiver, field);
self.ret_expr.replace("JS", &expr)
}
ImportTarget::StructuralSetter(class, field) => {
let (receiver, val) = fixed("setter", class, 1)?;
let expr = format!("{}.{} = {}", receiver, field, val[0]);
self.ret_expr.replace("JS", &expr)
}
ImportTarget::StructuralIndexingGetter(class) => {
let (receiver, field) = fixed("indexing getter", class, 1)?;
let expr = format!("{}[{}]", receiver, field[0]);
self.ret_expr.replace("JS", &expr)
}
ImportTarget::StructuralIndexingSetter(class) => {
let (receiver, field) = fixed("indexing setter", class, 2)?;
let expr = format!("{}[{}] = {}", receiver, field[0], field[1]);
self.ret_expr.replace("JS", &expr)
}
ImportTarget::StructuralIndexingDeleter(class) => {
let (receiver, field) = fixed("indexing deleter", class, 1)?;
let expr = format!("delete {}[{}]", receiver, field[0]);
self.ret_expr.replace("JS", &expr)
}
};
if self.catch {
Add experimental support for the `anyref` type This commit adds experimental support to `wasm-bindgen` to emit and leverage the `anyref` native wasm type. This native type is still in a proposal status (the reference-types proposal). The intention of `anyref` is to be able to directly hold JS values in wasm and pass the to imported functions, namely to empower eventual host bindings (now renamed WebIDL bindings) integration where we can skip JS shims altogether for many imports. This commit doesn't actually affect wasm-bindgen's behavior at all as-is, but rather this support requires an opt-in env var to be configured. Once the support is stable in browsers it's intended that this will add a CLI switch for turning on this support, eventually defaulting it to `true` in the far future. The basic strategy here is to take the `stack` and `slab` globals in the generated JS glue and move them into wasm using a table. This new table in wasm is managed at the fringes via injected shims. At `wasm-bindgen`-time the CLI will rewrite exports and imports with shims that actually use `anyref` if needed, performing loads/stores inside the wasm module instead of externally in the wasm module. This should provide a boost over what we have today, but it's not a fantastic strategy long term. We have a more grand vision for `anyref` being a first-class type in the language, but that's on a much longer horizon and this is currently thought to be the best we can do in terms of integration in the near future. The stack/heap JS tables are combined into one wasm table. The stack starts at the end of the table and grows down with a stack pointer (also injected). The heap starts at the end and grows up (state managed in linear memory). The anyref transformation here will hook up various intrinsics in wasm-bindgen to the runtime functionality if the anyref supoprt is enabled. The main tricky treatment here was applied to closures, where we need JS to use a different function pointer than the one Rust gives it to use a JS function pointer empowered with anyref. This works by switching up a bit how descriptors work, embedding the shims to call inside descriptors rather than communicated at runtime. This means that we're accessing constant values in the generated JS and we can just update the constant value accessed.
2018-10-18 08:43:36 -07:00
self.cx.expose_handle_error()?;
2018-06-27 22:42:34 -07:00
invoc = format!(
"\
try {{\n\
{}
}} catch (e) {{\n\
2019-02-04 02:08:08 +01:00
handleError(exnptr, e);\n\
}}\
",
2019-02-04 02:08:08 +01:00
&invoc
2018-06-27 22:42:34 -07:00
);
} else if self.catch_and_rethrow {
invoc = format!(
"\
try {{\n\
{}
}} catch (e) {{\n\
let error = (function () {{
try {{
return e instanceof Error
? `${{e.message}}\n\nStack:\n${{e.stack}}`
: e.toString();
}} catch(_) {{
return \"<failed to stringify thrown value>\";
}}
}}());
console.error(\"wasm-bindgen: imported JS function `{}` that \
was not marked as `catch` threw an error:\", \
error);
throw e;
}}\
",
&invoc,
shim,
);
}
if self.finally.len() > 0 {
2018-06-27 22:42:34 -07:00
invoc = format!(
"\
try {{\n\
{}
}} finally {{\n\
{}
}}\
",
2018-06-27 22:42:34 -07:00
&invoc, &self.finally
);
}
2018-06-15 12:55:37 -05:00
ret.push_str(&invoc);
2018-06-15 12:55:37 -05:00
ret.push_str("\n}\n");
Add experimental support for the `anyref` type This commit adds experimental support to `wasm-bindgen` to emit and leverage the `anyref` native wasm type. This native type is still in a proposal status (the reference-types proposal). The intention of `anyref` is to be able to directly hold JS values in wasm and pass the to imported functions, namely to empower eventual host bindings (now renamed WebIDL bindings) integration where we can skip JS shims altogether for many imports. This commit doesn't actually affect wasm-bindgen's behavior at all as-is, but rather this support requires an opt-in env var to be configured. Once the support is stable in browsers it's intended that this will add a CLI switch for turning on this support, eventually defaulting it to `true` in the far future. The basic strategy here is to take the `stack` and `slab` globals in the generated JS glue and move them into wasm using a table. This new table in wasm is managed at the fringes via injected shims. At `wasm-bindgen`-time the CLI will rewrite exports and imports with shims that actually use `anyref` if needed, performing loads/stores inside the wasm module instead of externally in the wasm module. This should provide a boost over what we have today, but it's not a fantastic strategy long term. We have a more grand vision for `anyref` being a first-class type in the language, but that's on a much longer horizon and this is currently thought to be the best we can do in terms of integration in the near future. The stack/heap JS tables are combined into one wasm table. The stack starts at the end of the table and grows down with a stack pointer (also injected). The heap starts at the end and grows up (state managed in linear memory). The anyref transformation here will hook up various intrinsics in wasm-bindgen to the runtime functionality if the anyref supoprt is enabled. The main tricky treatment here was applied to closures, where we need JS to use a different function pointer than the one Rust gives it to use a JS function pointer empowered with anyref. This works by switching up a bit how descriptors work, embedding the shims to call inside descriptors rather than communicated at runtime. This means that we're accessing constant values in the generated JS and we can just update the constant value accessed.
2018-10-18 08:43:36 -07:00
if self.ret_anyref || self.anyref_args.len() > 0 {
// Some return values go at the the beginning of the argument list
// (they force a return pointer). Handle that here by offsetting all
// our arg indices by one, but throw in some sanity checks for if
// this ever changes.
if let Some(start) = self.shim_arguments.get(0) {
if start == "ret" {
assert!(!self.ret_anyref);
if let Some(next) = self.shim_arguments.get(1) {
assert_eq!(next, "arg0");
}
for (idx, _) in self.anyref_args.iter_mut() {
*idx += 1;
}
} else {
assert_eq!(start, "arg0");
}
}
self.cx.anyref.import_xform(
"__wbindgen_placeholder__",
shim,
&self.anyref_args,
self.ret_anyref,
);
}
Ok(ret)
}
fn global_idx(&mut self) -> usize {
let ret = self.global_idx;
self.global_idx += 1;
ret
}
fn prelude(&mut self, s: &str) -> &mut Self {
for line in s.lines() {
self.prelude.push_str(line);
self.prelude.push_str("\n");
}
self
}
fn finally(&mut self, s: &str) -> &mut Self {
for line in s.lines() {
self.finally.push_str(line);
self.finally.push_str("\n");
}
self
}
}