It changes the following:
- `Symbol` object has `[[Construct]]` internal method
- Fix and comment step 1 in `Symbol` constructor.
- Implement step 1 in `BigInt` constructor.
This Pull Request fixes a faulty cast for `Array.splice`.
Negative values for delete_count were being directly casted to usize, which was not the intended behavior. This pull request fixes the issue by using a fallible conversion, defaulting to 0 if the conversion fails.
It changes the following:
- Replace cast in `Array.splice` prototype method with fallible conversion.
This Pull Request changes the following:
- Remove syntax error for unicode escaped characters in keywords from the lexer.
- Adjust the lexer tokens for keywords to indicate if they contain unicode escaped characters.
- Throw syntax errors in parser, when keywords cannot contain unicode escaped characters.
This Pull Request changes the following:
- Allow `Initializer` after `ArrayBindingPattern` in `FormalParameter`
- Refactor `Initializer` detection in `FormalParameter` to avoid clones
This Pull Request changes the following:
- Allow `PropertyName`s in `BindingProperty`in `ObjectBindingPattern`. Previously only `BindingIdentifier`s where allowed.
`ArrayBuffer.isView()` should check whether the object contains a `[[ViewedArrayBuffer]]` internal slot, which `DataView` has.
It changes the following:
- Fix `ArrayBuffer.isView()`
This Pull Request fixes/closes #337.
It changes the following:
- Implement class declaration parsing.
- Implement class expression parsing.
- Implement class execution.
There are still some features like `super` missing and there are some early errors that are not implemented yet. But I think it makes sense to merge this, as we can branch out the missing features from here.
This Pull Request fixes `byteLength` for `ArrayBuffer`. It should be an accessor property rather than a method, per the spec.
It changes the following:
- Removes `byteLength` method for `ArrayBuffer` built-in.
- Add `byteLength` accessor property for `ArrayBuffer`.
- Change `byte_length` function name to `get_byte_length`, to match other function names used for accessor properties.
<!---
Thank you for contributing to Boa! Please fill out the template below, and remove or add any
information as you feel neccesary.
--->
This Pull Request fixes/closes #1989.
It changes the following:
- Implements From<f32> for JsValue
Acked-by: Taylor Sutton <tsutton125@gmail.com>
This Pull Request fixes/closes #1998
The call to retrieve operands modifies pc, setting it to the index of
the *next* instruction. So, we save its initial value and use that
for printing.
This Pull Request lets true/false/null be used as object property identifiers, when using dot assignment.
`foo.null = 'bar';`
It changes the following:
- AST parsing of member expressions
This Pull Request fixes length properties on multiple array prototype methods that were including rest parameters in the count. More tests should pass.
It changes the following:
- Length properties on some array prototype methods
This Pull Request makes the non-octal-decimal-integer test pass. The test would previously fail for values with multiple leading zeroes.
It changes the following:
- Number lexer
`Node::DoWhileLoop` ast node had a buggy bytecode generation where `self.patch_jump(exit)` was called after emitting `LoopEnd` opcode. This would patch the loop exit to the instruction following the do while code, which would panic in cases where do while was enclosed in a block statement.
This Pull Request fixes#1929.
It changes the following:
- Patch jump before emitting `Opcode::LoopEnd`
- Add test which has do while statement inside a block statement to demonstrate that the change fixes the panic.
This Pull Request fixes/closes #1962.
It changes the following:
- When executing arithmetic operations on `JsValue`s, try to use integer operations and fallback to `f64` operations on error.
- Adds tests for serde_json conversions from integer operations.
Some of the fields in AST structs were both
1. Arrays
2. Marked as 'flatten'
This is illegal per serde docs (and doesn't really make sense).
The fix is to remove the attribute.
See: https://serde.rs/attr-flatten.htmlFixes: #1920
Co-authored-by: Taras Boiko <me@tboiko.com>
This Pull Request fixes/closes #1942.
`Date.prototype.toDateString` should return a value representing the local date. The Rust `Date` inner value represents UTC time, so it should be adjusted to local time before formatting (see equivalent conversions performed by `to_string` and `to_time_string`).
To verify this is working as intended, run the test suite with your OS timezone set to `GMT+0`, then again with `GMT+10`. The test `date_proto_to_date_string` should pass for each. For me (Ubuntu via WSL), this can be done with `sudo dpkg-reconfigure tzdata`.
This PR also fixes a couple other test cases that used the wrong month value (as noted at the top of the file, JS months are 0-based while `chrono` months are 1-based).
Fixes `BigInt` and `Number` comparison, and vice versa. Before we were removing the decimal point of the floating-point number which was causing cases like `0.000001 > 0n` (or `0n < 0.000001`) to fail.
It changes the following:
- Refreshes the vm and debugging docs to represent the current state
- Fix some bytecode trace output
- Rename a field in the `CodeBlock`
<!---
Thank you for contributing to Boa! Please fill out the template below, and remove or add any
information as you feel neccesary.
--->
This Pull Request fixes/closes #1784.
There're still a few tests failing, notably:
- `iter-set-elem-prop-non-writable` - we don't have generator functions implemented
- `calling-from-valid-1-noStrict`, `iter-map-fn-this-non-strict` - `thisArg` in non-strict mode, when undefined, should be inherited (that's what I'm guessing, I haven't confirmed this, but strict counterparts do pass with `thisArg` being `undefined`)
- `source-array-boundary`, `elements-deleted-after` - ~~Not sure yet, still investigating, but they also include thisArg, so perhaps function calling has an underlying issue?~~ Failing because `this` on the top level evaluates to an empty object instead of containing everything from the top scope
Co-authored-by: HalidOdat <halidodat@gmail.com>
This fixes a bug with the postfix increment and decrement. Before those operators would return the left-hand-side value, but the spec specifies they should return ToNumeric(left-had-side value).
<!---
Thank you for contributing to Boa! Please fill out the template below, and remove or add any
information as you feel neccesary.
--->
This Pull Request fixes some bugs related to try blocks:
- Fixes a panic when a finally block contained variable declarations. (Thanks to @VTCAKAVSMoACE for the report!)
- Fixes a bug where try blocks in the last position of a statement list didn't return its inner last value as the result of the evaluation.
- Add tests for both cases and two other common cases.
- Extract and cleanup some code.
<!---
Thank you for contributing to Boa! Please fill out the template below, and remove or add any
information as you feel neccesary.
--->
This Pull Request fixes/closes #1897.
It changes the following:
- Rejects uninitialized const declarations inside the init value of a for loop statement.
- Adds test for the case.
This PR changes the following:
- Fixes the panics induced by incorrect continues.
- Adds tests which demonstrate the various panics induced.
- Actually rustfmts correctly?
This PR changes the following:
- More elegantly handles illegal access statements in compile_access
- Adds a slew of previously unhandled illegal access test cases
### Caveats
It is very, very likely that you will want to simply restrict unary and assignment operations in the AST. However, this prevents crashes in the meantime with a error that is just slightly less detailed than if it were implemented in AST.
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 PR add `Number.parseInt` and `Number.parseFloat` which according to spec are clones of the global objects `parseInt` and `parseFloat`.
It also fixes the last failing test of the `NativeError` feature with this we get 100% spec complaint `NativeError`s 🎉
It changes the following:
- Add `Number.parseInt()`
- Add `Number.parseFloat()`
- Fix length of `AggregateError`
- Fix length of `Reflect.setPrototypeOf`
This PR is also related to #577
Changes:
- Implements `IteratorValue` (`IteratorResult::value()`)
- Implements `IteratorComplete` (`IteratorResult::complete()`)
- Implements `IteratorStep` (`IteratorRecord::step()`)
- Makes `IteratorNext` (`IteratorRecord::next()`) spec compliant
- Deprecates/removes `JsValue::get_field()`.
The ECMAScript 2022 specification removes the `toInteger` method, and replaces it with `toIntegerOrInfinity`, which is arguably better for us since the `JsValue::toInteger` returns an `f64`, which is pretty confusing at times.
This pull request removes the `JsValue::to_integer` method, replaces all its calls by `JsValue::to_integer_or_infinity` or others per the spec and documents several methods from the `string` builtin.
We store string `PropertyKey`s with two enums `String` and `Index` for performance reasons, but the spec does not differentiate between string and index property keys so before conversion to `JsValue` we have to convert to a string.
This was failing tests like `Reflect.ownKeys([true, "", 1])` because it was returning (integer numbers) `[1, 2, 3]` instead of `['1', '2', '3']`
Before the `%NativeError%` objects (like `TypeError`, `ReferenceError`, etc) `[[prototype]]` field was set to `Function.prototype` but this is wrong it should be the `Error` constructor object itself.
This makes the `%NativeError%`s 100% spec compliant :)
(except `AggregateError` because its not implemented)
This cleans up the vm stack on a function return. This is needed because when errors are thrown, there may be still values on the stack that cannot be cleaned up via regular control flow. The change should have little performance impact, because the `Vec::truncate` function does not alter the capacity of the vector.
This PR fixes the bug when passing `undefined` as radix in `Number.prototype.toString()` which was throwing a range error, instead of setting the radix to `10`
[spec](https://tc39.es/ecma262/#sec-number.prototype.tostring):
> **Note**
>
> The optional radix should be an [integral Number](https://tc39.es/ecma262/#integral-number) value in the inclusive range 2𝔽 to 36𝔽. If radix is undefined then 10𝔽 is used as the value of radix.
This Pull Request fixes/closes #1864.
It changes the following:
- Removed `JsBigInt` from `Const` nodes, using a boxed `BigInt` instead.
- Modifies the `JsObject` variant so that it has a similar structure to other variants, where the internal structure is private.
The size of `JsValue` stays in 2 64-bit words (in a 64-bit system at least), and the size of `Const` also stays the same.
I have noticed that we clone tokens too much in the parser, so I was thinking that we should implement a by-value getter for `kind()`. Something like `kind_unwrap()`.
This Pull Request is related to #1873.
It changes the following:
- Removes the panic in case a label is not found. I think this should be an early syntax error, but at least we shouldn't panic while we fix the issue with the labels.
I think we should solve the issue with labeled statements for 0.15.
This PR adds support for arrays with empty elements e.g. `["", false, , , ]`. Before we were filling the empty places with `undefined`, but this is wrong according to [spec](https://tc39.es/ecma262/#sec-runtime-semantics-arrayaccumulation) there shouldn't be undefined with a index at that place, instead only `length` is incremented. So `[,,,].length == 3` and operations like `[,,,,].indexOf(undefined) == -1`, `[,,,,].lastIndexOf(undefined) == -1` etc.
Halid Odat
lupd
raskad
Jordan Last
Taylor Sutton
NorbertGarfield
pd
jedel1043
Taras Boiko
Aaron Ross
Nikodem Rabuliński
Addison Crump
Iban Eguia