In previous implementation when converting a number to PropertyKey, we
always converted it to string and only when the conversion to
PropertyKey::Index was not possible we returned that string.
This change allows using built-in `JsObject` wrappers with the `TryFromJs` logic, which makes it easier to derive it in complex objects containing known built-ins, being able to use all the power of the wrappers even inside complex structures.
It changes the following:
* Implements `TryFromJs` for all `JsObject` wrappers ~, except for `JsProxy` and `Js<Int>Array` wrappers~ -> This has now been added.
* Adds checker functions for individual typed array objects.
* Adds some missing checker functions from `Object` to `JsObject`
~The reason for not implementing it for `JsProxy` is that we don't have a way (as far as I can tell) to know if a `JsObject` is a `Proxy` object. The reason for the typed arrays (note that `JsTypedArray` implements `TryFromJs`) is that I didn't find an easy way to know which type of typed array a `JsObject` was. Do we have a way of using a `JsObject` to create a Rust `JsUint8Array` or the like?~
This Pull Request fixes#2317 and #1835, finally giving our engine proper realms 🥳.
It changes the following:
- Extracts the compile environment stack from `Realm` and into `Vm`.
- Adjusts the bytecompiler to accommodate this change.
- Adjusts `call/construct_internal` to accommodate this change. This also coincidentally fixed#2317, which I'm pretty happy about.
- Adjusts several APIs (`NativeJob`, `Realm`) and builtins (`eval`, initializers) to accommodate this change.
- Adjusts `JsNativeError`s to hold a reference to the Realm from which they were created. This only affects errors created within calls to function objects. Native calls don't need to set the realm because it's inherited by the next outer active function object. TLDR: `JsError` API stays the same, we just set the origin Realm of errors in `JsObject::call/construct_internal`.
This Pull Request fixes test [`assert-throws-same-realm.js`](eb44f67274/test/harness/assert-throws-same-realm.js).
It changes the following:
- Handles global variables through the global object, instead of the `context`.
- Adds an `active_function` field to the vm, which is used as the `NewTarget` when certain builtins aren't called with `new`.
- Adds a `realm_intrinsics` field to `Function`.
When a rust string literal is given for a property key, boa checks if it can be parsed as an index and converts the string into a utf16 slice. This PR rewrites each hard-coded property key as a utf16 slice so that we can bypass those conversions at runtime.
This improves QuickJS benchmark score 5% on average.
Richards: 35.4 -> 37.0
DeltaBlue: 35.0 -> 38.1
Crypto: 57.6 -> 59.6
RayTrace: 137 -> 146
EarleyBoyer: 131 -> 138
Splay: 98.3 -> 104
NavierStokes: 10.2 -> 10.2
Small (ish?) step towards having proper realm records
This PR changes the following:
- Moves `Intrinsics` to `Realm`.
- Cleans up the initialization logic of our intrinsics to not depend on `Context`, unblocking things like #2314.
- Adds hooks to initialize the global object and the global this per the corresponding [`InitializeHostDefinedRealm ( )`](https://tc39.es/ecma262/#sec-initializehostdefinedrealm) hook. Though, this is currently broken because the vm uses `GlobalPropertyMap` instead of the `JsObject` API to initialize global properties.
Follows from #2528, and should complement #2411 to implement the module import hooks.
~~Similarly to the Intl/ICU4X PR (#2478), this has a lot of trivial changes caused by the new lifetimes. I thought about passing the queue and the hooks by value, but it was very painful having to wrap everything with `Rc` in order to be accessible by the host.
In contrast, `&dyn` can be easily provided by the host and has the advantage of not requiring additional allocations, with the downside of adding two more lifetimes to our `Context`, but I think it's worth.~~ I was able to unify all lifetimes into the shortest one of the three, making our API just like before!
Changes:
- Added a new `HostHooks` trait and a `&dyn HostHooks` field to `Context`. This allows hosts to implement the trait for their custom type, then pass it to the context.
- Added a new `JobQueue` trait and a `&dyn JobQueue` field to our `Context`, allowing custom event loops and other fun things.
- Added two simple implementations of `JobQueue`: `IdleJobQueue` which does nothing and `SimpleJobQueue` which runs all jobs until all successfully complete or until any of them throws an error.
- Modified `boa_cli` to run all jobs until the queue is empty, even if a job returns `Err`. This also prints all errors to the user.
The section about `Symbol` on the [specification](https://tc39.es/ecma262/#sec-ecmascript-language-types-symbol-type) says:
> The Symbol type is the set of all non-String values that may be used as the key of an Object property ([6.1.7](https://tc39.es/ecma262/#sec-object-type)).
Each possible Symbol value is unique and immutable.
Our previous implementation of `JsSymbol` used `Rc` and a thread local `Cell<usize>`. However, this meant that two different symbols in two different threads could share the same hash, making symbols not unique.
Also, the [GlobalSymbolRegistry](https://tc39.es/ecma262/#table-globalsymbolregistry-record-fields) is meant to be shared by all realms, including realms that are not in the same thread as the main one; this forces us to replace our current thread local global symbol registry with a thread-safe one that uses `DashMap` for concurrent access. However, the global symbol registry uses `JsString`s as keys and values, which forces us to either use `Vec<u16>` instead (wasteful and needs to allocate to convert to `JsString` on each access) or make `JsString` thread-safe with an atomic counter. For this reason, I implemented the second option.
This PR changes the following:
- Makes `JsSymbol` thread-safe by using Arc instead of Rc, and making `SYMBOL_HASH_COUNT` an `AtomicU64`.
- ~~Makes `JsString` thread-safe by using `AtomicUsize` instead of `Cell<usize>` for its ref count.~~ EDIT: Talked with @jasonwilliams and we decided to use `Box<[u16]>` for the global registry instead, because this won't penalize common usage of `JsString`, which is used a LOT more than `JsSymbol`.
- Makes the `GLOBAL_SYMBOL_REGISTRY` truly global, using `DashMap` as our global map that is shared by all threads.
- Replaces some thread locals with thread-safe alternatives, such as static arrays and static indices.
- Various improvements to all related code for this.
This Pull Request fixes various bugs related to classes.
The biggest changes are:
- Changed private names to be unique across multiple classes.
- Changed private name resolution to work via a visitor after a class is parsed. The way class early errors are defined makes it impossible to perform private name resolution while parsing.
- Added function names to class methods.
- Added class name binding to method function environments.
- Separated opcodes for `static` and non-`static` class method definitions to make the above operations possible.
There are still some bugs and further issues with classes but this is already a lot.
As part of the new modules PR, I was working on implementing the [host hooks](https://tc39.es/ecma262/#sec-host-hooks-summary) for the module import hooks and custom job queues. However, the promises module needed a bit of a refactor in order to couple with the new API. So, I thought it was a good idea to separate the promises refactor into its own PR, since the other PR is already big as it is.
- Replaced some usages of `JobCallback` with a new `NativeJob` that isn't traced by the GC, since those closures are always rooted and executed by the `Context` globally. This will also allow hosts to pass their custom jobs to the job queue, and maybe could also accept futures in the Future (pun intended 😆).
- Refactored several functions to account for the `HostPromiseRejectionTracker` hook which needs the promise `JsObject`.
- Rewrote some patterns with newer Rust idioms.
Just a general cleanup of the APIs of our `Context`.
- Reordered the `pub` and `pub(crate)/fn` methods to have a clear separation between our public and private APIs.
- Removed the call method and added it to `JsValue` instead, which semantically makes a bit more sense.
- Removed the `construct_object` method, and added an utility method `new` to `JsObject` instead.
- Rewrote some patterns I found while rewriting the calls of the removed function.
This Pull Request fixes/closes #1180. (I'll open a tracking issue for the progress)
It changes the following:
- Redesigns the internal API of Intl to (hopefully!) make it easier to implement a service.
- Implements the `Intl.Locale` service.
- Implements the `Intl.Collator` service.
- Implements the `Intl.ListFormat` service.
On the subject of the failing tests. Some of them are caused by missing locale data in the `icu_testdata` crate; we would need to regenerate that with the missing locales, or vendor a custom default data.
On the other hand, there are some tests that are bugs from the ICU4X crate. The repo https://github.com/jedel1043/icu4x-test262 currently tracks the found bugs when running test262. I'll sync with the ICU4X team to try to fix those.
cc @sffc
Per the [Standard Library development guide](https://std-dev-guide.rust-lang.org/code-considerations/performance/inline.html):
> You can add `#[inline]`:
>
> - To public, small, non-generic functions.
>
> You shouldn't need `#[inline]`:
> - On methods that have any generics in scope.
> - On methods on traits that don't have a default implementation.
>
> `#[inline]` can always be introduced later, so if you're in doubt they can just be removed.
This PR follows this guideline to reduce the number of `#[inline]` annotations in our code, removing the annotation in:
- Non-public functions
- Generic functions
- Medium and big functions.
Hopefully this shouldn't impact our perf at all, but let's wait to see the benchmark results.
This Pull Request restructures the lint deny/warn/allow lists in `boa_engine`. It adds a lot of documentation to pup functions. There are still a few clippy lints that are not fixed, mainly regarding casting of number types. Fixing those lints effectiveley would in some cases probably require bigger refactors.
This should probably wait for #2449 to be merged, because that PR already fixes that lints regarding the `Date` built-in.
This Pull Request updates the codebase to the newest version of rustc (1.65.0).
It changes the following:
- Bumps `rust-version` to 1.65.0.
- Rewrites some snippets to use the new let else, ok_or_else and some other utils.
- Removes the `rustdoc::missing_doc_code_examples` allow lint from our codebase. (Context: https://github.com/rust-lang/rust/pull/101732)
This Pull Request replaces `contains`, `contains_arguments`, `has_direct_super` and `function_contains_super` with visitors. (~1000 removed lines!)
Also, the new visitor implementation caught a bug where we weren't setting the home object of async functions, generators and async generators for methods of classes, which caused a stack overflow on `super` calls, and I think that's pretty cool!
Next is `var_declared_names`, `lexically_declared_names` and friends, which will be on another PR.
This is an experiment that tries to migrate the codebase from eager `Error` objects to lazy ones.
In short words, this redefines `JsResult = Result<JsValue, JsError>`, where `JsError` is a brand new type that stores only the essential part of an error type, and only transforms those errors to `JsObject`s on demand (when having to pass them as arguments to functions or store them inside async/generators).
This change is pretty big, because it unblocks a LOT of code from having to take a `&mut Context` on each call. It also paves the road for possibly making `JsError` a proper variant of `JsValue`, which can be a pretty big optimization for try/catch.
A downside of this is that it exposes some brand new error types to our public API. However, we can now implement `Error` on `JsError`, making our `JsResult` type a bit more inline with Rust's best practices.
~Will mark this as draft, since it's missing some documentation and a lot of examples, but~ it's pretty much feature complete. As always, any comments about the design are very much appreciated!
Note: Since there are a lot of changes which are essentially just rewriting `context.throw` to `JsNativeError::%type%`, I'll leave an "index" of the most important changes here:
- [boa_engine/src/error.rs](https://github.com/boa-dev/boa/pull/2283/files#diff-f15f2715655440626eefda5c46193d29856f4949ad37380c129a8debc6b82f26)
- [boa_engine/src/builtins/error/mod.rs](https://github.com/boa-dev/boa/pull/2283/files#diff-3eb1e4b4b5c7210eb98192a5277f5a239148423c6b970c4ae05d1b267f8f1084)
- [boa_tester/src/exec/mod.rs](https://github.com/boa-dev/boa/pull/2283/files#diff-fc3d7ad7b5e64574258c9febbe56171f3309b74e0c8da35238a76002f3ee34d9)
This PR fixes the length/index types from `usize` to `u64`, because in JavaScript the length can be from `0` to `2^53 - 1` it cannot be represented in a `usize`. On `64-bit` architectures this is fine because it is already a unsigned 64bit number and these changes essentially do nothing. But on 32bit architectures this was causing the length to be truncated giving an unexpected result.
fixes/closes #2182
It changes the following:
- Make length a `u64` to be able to represent all possible length values
- Change `JsValue::to_length()` t return `u64`
- Change `JsValue::to_index()` t return `u64`
This PR adds a safe wrapper around JavaScript `JsSet` from `builtins::set`, and is being tracked at #2098.
Implements following methods
- [x] `Set.prototype.size`
- [x] `Set.prototype.add(value)`
- [x] `Set.prototype.clear()`
- [x] `Set.prototype.delete(value)`
- [x] `Set.prototype.has(value)`
- [x] `Set.prototype.forEach(callbackFn[, thisArg])`
Implement wrapper for `builtins::set_iterator`, to be used by following.
- [x] `Set.prototype.values()`
- [x] `Set.prototype.keys()`
- [x] `Set.prototype.entries()`
*Note: Are there any other functions that should be added?
Also adds `set_create()` and made `get_size()` public in `builtins::set`.
Building up to #186, this PR extracts an `Intrinsics` struct from `Context`, facilitating a lot the extraction of a `Realm` struct.
Also, it adapts the `BuiltIn` trait to be useful for builtins that don't expose a global property on initialization (`Generator`, `TypedArray`, etc.)
It changes the following:
- Creates an `Intrinsics` struct and refactors `Context` to transfer its intrinsic related fields to `Intrinsics`.
- Renames some methods and parameters to better describe their functionality.
- Makes `BuiltIn::init` return `Option<JsValue>` to skip global property initialization if the builtin initialization returns `None`
This Pull Request is related to #577 .
It changes the following:
- Remove `JsValue::set_field`
- Remove `JsValue::set_property`
- Remove almost all uses of `JsValue::get_field`
- Use `.get_v()` instead of `get_field` according to spec in `serialize_json_property`
- Remove `Array::new_array()`
- Remove `Array::add_to_array_object()`
This PR adds some Clippy lints. Mainly, it adds the list of pedantic lints excluding some lints that were causing too many warnings. I also denied some useful restriction and pedantic lints, to make sure we use `Self` all the possible times (for better maintainability), and that we pass elements by reference where possible, for example, or that the documentation is properly written.
This might even have some small performance gains.
I also added a perfect hash function for the CLI keywords, which should be more efficient than a `HashSet`. This is something we could use elsewhere too.
It changes the following:
- Add handling for proxy objects to the abstract `is_array` operation.
- Implement the abstract `is_array` operation for `JsValue` and `JsObject` to avoid clones.
- Fix some builtin function lengths.
Previously when we had the `context.throw_` methods (like `context.thtrow_type_error()`) they were limited as to where we could call them, e.i. a function that returned `JsResult<JsValue>`. So we had to call the `context.construct_` methods with an explicit `Err()` enum wrap to throw in functions that returned non-jsvalues (which happens a lot).
Now, with this PR the throw methods have a generic `JsResult<R>` return that can return in any `JsResult<T>` returning function. Which cleans the API and makes the user experience a bit better.
```rust
return Err(context.construct_type_error("..."));
// to
return context.throw_type_error("...");
```
This implements most of the missing vm operations and fixes most of the panics.
The early errors for declarations in `for in` and `for of` loops are moved to the parser. The content of those `Node`s is changed accordingly.