When parsing for i18n messages, interpolated strings are
split into `Text` and `Placeholder` pieces. The method that
does this `_visitTextWithInterpolation()` was becoming too
complex. This commit refactors that method along with some
associated functions that it uses.
PR Close#39717
ngtsc has a robust suite of testing utilities, designed for in-memory
testing of a TypeScript compiler. Previously these utilities lived in the
`test` directory for the compiler-cli package.
This commit moves those utilities to an `ngtsc/testing` package, enabling
them to be depended on separately and opening the door for using them from
the upcoming language server testing infrastructure.
As part of this refactoring, the `fake_core` package (a lightweight API
replacement for @angular/core) is expanded to include functionality needed
for Language Service test use cases.
PR Close#39594
The result of utf-8 encoding a string was represented in a string, where
each individual character represented a single byte according to its
character code. All usages of this data were interested in the byte
itself, so this required conversion from a character back to its code.
This commit simply stores the individual bytes in array to avoid the
conversion. This yields a ~10% performance improvement for i18n message
ID computation.
PR Close#39694
Message ID computation makes extensive use of big integer
multiplications in order to translate the message's fingerprint into
a numerical representation. In large compilations with heavy use of i18n
this was showing up high in profiler sessions.
There are two factors contributing to the bottleneck:
1. a suboptimal big integer representation using strings, which requires
repeated allocation and conversion from a character to numeric digits
and back.
2. repeated computation of the necessary base-256 exponents and their
multiplication factors.
The first bottleneck is addressed using a representation that uses an
array of individual digits. This avoids repeated conversion and
allocation overhead is also greatly reduced, as adding two big integers
can now be done in-place with virtually no memory allocations.
The second point is addressed by a memoized exponentiation pool to
optimize the multiplication of a base-256 exponent.
As an additional optimization are the two 32-bit words now converted to
decimal per word, instead of going through an intermediate byte buffer
and doing the decimal conversion per byte.
The results of these optimizations depend a lot on the number of i18n
messages for which a message should be computed. Benchmarks have shown
that computing message IDs is now ~6x faster for 1,000 messages, ~14x
faster for 10,000 messages, and ~24x faster for 100,000 messages.
PR Close#39694
Similar to #39613, #39609, and #38898, we should store the `keySpan` for
Reference nodes so that we can accurately map from a template node to a
span in the original file. This is most notably an issue at the moment
for directive references `#ref="exportAs"`. The current behavior for the
language service when requesting information for the reference
is that it will return a text span that results in
highlighting the entire source when it should only highlight "ref" (test
added for this case as well).
PR Close#39616
Though we currently have the knowledge of where the `key` for an
event binding appears during parsing, we do not propagate this
information to the output AST. This means that once we produce the
template AST, we have no way of mapping a template position to the key
span alone. The best we can currently do is map back to the
`sourceSpan`. This presents problems downstream, specifically for the
language service, where we cannot provide correct information about a
position in a template because the AST is not granular enough.
This is essentially identical to the change from #38898, but for event
bindings rather than input bindings.
PR Close#39609
Similar to #39609 and #38898, though we currently have the knowledge of where the key for an
attribute appears during parsing, we do not propagate this
information to the output AST. This means that once we produce the
template AST, we have no way of mapping a template position to the key
span alone. The best we can currently do is map back to the
sourceSpan. This presents problems downstream, specifically for the
language service, where we cannot provide correct information about a
position in a template because the AST is not granular enough.
PR Close#39613
Tokenized text node may have leading whitespace skipped from their
source-span. But the source-span is used to compute where there are
interpolated blocks, resulting in placeholder nodes whose source-spans
are offset by the amount of skipped characters.
This fix uses the `fullStart` location of text source-spans for computing
the source-span of placeholders, so that they are accurate.
Fixes#39195
PR Close#39486
This commit ensures that when leading whitespace is skipped by
the tokenizer, the original start location (before skipping) is captured
in the `fullStart` property of the token's source-span.
PR Close#39486
In an i18n message, two placeholders next to each other must have
an "empty" message-part to separate them. Previously, the source-span
for this message-part was pointing to the wrong original location.
This caused problems in the generated source-maps and lead to extracted
i18n messages from being rendered incorrectly.
PR Close#39486
Currently expressions `$event.foo()` and `this.$event.foo()`, as well as `$any(foo)` and
`this.$any(foo)`, are treated as the same expression by the compiler, because `this` is considered
the same implicit receiver as when the receiver is omitted. This introduces the following issues:
1. Any time something called `$any` is used, it'll be stripped away, leaving only the first parameter.
2. If something called `$event` is used anywhere in a template, it'll be preserved as `$event`,
rather than being rewritten to `ctx.$event`, causing the value to undefined at runtime. This
applies to listener, property and text bindings.
These changes resolve the first issue and part of the second one by preserving anything that
is accessed through `this`, even if it's one of the "special" ones like `$any` or `$event`.
Furthermore, these changes only expose the `$event` global variable inside event listeners,
whereas previously it was available everywhere.
Fixes#30278.
PR Close#39323
This commit handles the following cases:
- incomplete pipes in a pipe chain
- incomplete arguments in a pipe chain
- incomplete arguments provided to a pipe
- nested pipes
The idea is to unconditionally recover on the presence of a pipe, which
should be okay because expression parsing can be independently between
pipes.
PR Close#39437
There is no actionable change in this commit other than to pretty-print
EOF tokens. Actual parsing of unterminated pipes is already supported,
this just adds a test for it.
Part of #38596
PR Close#39113
At a high level, the current shadow DOM shim logic works by escaping the content of a CSS rule
(e.g. `div {color: red;}` becomes `div {%BLOCK%}`), using a regex to parse out things like the
selector and the rule body, and then re-adding the content after the selector has been modified.
The problem is that the regex has to be very broad in order capture all of the different use cases,
which can cause it to match strings suffixed with a semi-colon in some places where it shouldn't,
like this URL from Google Fonts `https://fonts.googleapis.com/css2?family=Roboto:wght@400;500&display=swap`.
Most of the time this is fine, because the logic that escapes the rule content to `%BLOCK%` will
have converted it to something that won't be matched by the regex. However, it breaks down for rules
like `@import` which don't have a body, but can still have quoted content with characters that can
match the regex.
These changes resolve the issue by making a second pass over the escaped string and replacing all
of the remaining quoted content with `%QUOTED%` before parsing it with the regex. Once everything
has been processed, we make a final pass where we restore the quoted content.
In a previous iteration of this PR, I went with a shorter approach which narrowed down the
regex so that it doesn't capture rules without a body. It fixed the issue, but it also ended
up breaking some of the more contrived unit test cases. I decided not to pursue it further, because
we would've ended up with a very long and brittle regex that likely would've broken in even weirder
ways.
Fixes#38587.
PR Close#38716
Removes `ViewEncapsulation.Native` which has been deprecated for several major versions.
BREAKING CHANGES:
* `ViewEncapsulation.Native` has been removed. Use `ViewEncapsulation.ShadowDom` instead. Existing
usages will be updated automatically by `ng update`.
PR Close#38882
Prior to this change, expressions within ICUs would have a source span
corresponding with the whole ICU. This commit narrows down the source
spans of these expressions to the exact location in the source file, as
a prerequisite for reporting type check errors within these expressions.
PR Close#39072
The expression parser already has support for recovering on malformed
property reads, but did not have tests describing the recovered ast in
such cases. This commit adds tests to demonstrate such cases; in
particular, the recovered ast is a full PropertyRead but with an empty
property name. This is likely the most preferred option, as it does not
constrain consumers of the AST to what the property name should look
like. Furthermore, we cannot mark the property name as empty in any
other way (e.g. an EmptyExpr) because the property name, as of present,
is a string field rather than an AST itself.
Note that tokens past a malformed property read are not preserved in the
AST (for example in `foo.1234`, `1234` is not preserved in the AST).
This is because the extra tokens do not belong to the singular
expression created by the property read, and there is not a meaningful
way to interpret a secondary expression in a single parsed expression.
Part of #38596
PR Close#38998
This patch refactors the interpolation parser to do so iteratively
rather than using a regex. Doing so prepares us for supporting granular
recovery on poorly-formed interpolations, for example when an
interpolation does not terminate (`{{ 1 + 2`) or is not terminated
properly (`{{ 1 + 2 {{ 2 + 3 }}`).
Part of #38596
PR Close#38977
This patch adds support for recovering well-formed (and near-complete)
ASTs for semantically malformed keyed reads and keyed writes. See the
added tests for details on the types of semantics we can now recover;
in particular, notice that some assumptions are made about the form of
a keyed read/write intended by a user. For example, in the malformed
expression `a[1 + = 2`, we assume that the user meant to write a binary
expression for the key of `a`, and assign that key the value `2`. In
particular, we now parse this as `a[1 + <empty expression>] = 2`. There
are some different interpretations that can be made here, but I think
this is reasonable.
The actual changes in the parser code are fairly minimal (a nice
surprise!); the biggest addition is a `writeContext` that marks whether
the `=` operator can serve as a recovery point after error detection.
Part of #38596
PR Close#39004
Currently it is impossible to determine the source of a binding that
generates `BoundAttribute` because all bound attributes generated from a
microsyntax expression share the same source span.
For example, in
```html
<div *ngFor="let item of items; trackBy: trackByFn"></div>
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
source span for all `BoundAttribute`s generated from microsyntax
```
the `BoundAttribute` for both `ngForOf` and `ngForTrackBy`
share the same source span.
A lot of hacks were necessary in View Engine language service to work
around this limitation. It was done by inspecting the whole source span
then figuring out the relative position of the cursor.
With this change, we introduce a flag to set the binding span as the
source span of the `ParsedProperty` in Ivy AST.
This flag is needed so that we don't have to change VE ASTs.
Note that in the binding parser, we already set `bindingSpan` as the
source span for a `ParsedVariable`, and `keySpan` as the source span for
a literal attribute. This change makes the Ivy AST more consistent by
propagating the binding span to `ParsedProperty` as well.
PR Close#39036
This commit updates the symbols in the TemplateTypeCheck API and methods
for retrieving them:
* Include `isComponent` and `selector` for directives so callers can determine which
attributes on an element map to the matched directives.
* Add a new `TextAttributeSymbol` and return this when requesting a symbol for a `TextAttribute`.
* When requesting a symbol for `PropertyWrite` and `MethodCall`, use the
`nameSpan` to retrieve symbols.
* Add fix to retrieve generic directives attached to elements/templates.
PR Close#38844
Now that we have `keySpan` for `BoundAttribute` (implemented in
https://github.com/angular/angular/pull/38898) we could do the same
for `Variable`.
This would allow us to distinguish the LHS and RHS from the whole source
span.
PR Close#38965
Some compiler tests take a long time to run, even using multiple
executors. A profiling session revealed that most time is spent in
parsing source files, especially the default libraries are expensive to
parse.
The default library files are constant across all tests, so this commit
introduces a shared cache of parsed source files of the default
libraries. This achieves a significant improvement for several targets
on my machine:
//packages/compiler-cli/test/compliance: from 23s to 5s.
//packages/compiler-cli/test/ngtsc: from 115s to 11s.
Note that the number of shards for the compliance tests has been halved,
as the extra shards no longer provide any speedup.
PR Close#38909
Though we currently have the knowledge of where the `key` for an
attribute binding appears during parsing, we do not propagate this
information to the output AST. This means that once we produce the
template AST, we have no way of mapping a template position to the key
span alone. The best we can currently do is map back to the
`sourceSpan`. This presents problems downstream, specifically for the
language service, where we cannot provide correct information about a
position in a template because the AST is not granular enough.
PR Close#38898
Using an interface makes the code cleaner and more readable.
This change also adds the `range` property to the type to be used
for source-mapping.
PR Close#38775
Let's say we have a code like
```html
<div<span>123</span>
```
Currently this gets parsed into a tree with the element tag `div<span`.
This has at least two downsides:
- An incorrect diagnostic that `</span>` doesn't close an element is
emitted.
- A consumer of the parse tree using it for editor services is unable to
provide correct completions for the opening `<span>` tag.
This patch attempts to fix both issues by instead parsing the code into
the same tree that would be parsed for `<div></div><span>123</span>`.
In particular, we do this by optimistically scanning an open tag as
usual, but if we do not notice a terminating '>', we mark the tag as
"incomplete". A parser then emits an error for the incomplete tag and
adds a synthetic (recovered) element node to the tree with the
incomplete open tag's name.
What's the downside of this? For one, a breaking change.
<ol>
<li>
The first breaking change is that `<` symbols that are ambiguously text
or opening tags will be parsed as opening tags instead of text in
element bodies. Take the code
```html
<p>a<b</p>
```
Clearly we cannot have the best of both worlds, and this patch chooses
to swap the parsing strategy to support the new feature. Of course, `<`
can still be inserted as text via the `<` entity.
</li>
</ol>
Part of #38596
PR Close#38681
The current tests print out the span numbers, which are really difficult to verify
since it requires manually going to the template string and looking at what
characters appear within those indexes. The better humanization would be
to use the toString method of the spans, which prints the span text itself
PR Close#38902
Common AST formats such as TS and Babel do not use a separate
node for comments, but instead attach comments to other AST nodes.
Previously this was worked around in TS by creating a `NotEmittedStatement`
AST node to attach the comment to. But Babel does not have this facility,
so it will not be a viable approach for the linker.
This commit refactors the output AST, to remove the `CommentStmt` and
`JSDocCommentStmt` nodes. Instead statements have a collection of
`leadingComments` that are rendered/attached to the final AST nodes
when being translated or printed.
PR Close#38811
Recent work on compiler internals in #38539 led to an unexpected failure,
where a pipe used exclusively inside of an ICU would no longer be
emitted into the compilation output. This caused runtime errors due to
missing pipes.
The issue occurred because the change in #38539 would determine the set
of used pipes up-front, independent from the template compilation using
the `R3TargetBinder`. However, `R3TargetBinder` did not consider
expressions within ICUs, so any pipe usages within those expressions
would not be detected. This fix unblocks #38539 and also concerns
upcoming linker work, given that prelink compilations would not go
through full template compilation but only `R3TargetBinder`.
PR Close#38810
In a microsyntax expressions, some attributes are not bound after
desugaring. For example,
```html
<div *ngFor="let item of items">
</div>
```
gets desugared to
```html
<ng-template ngFor let-items [ngForOf]="items">
</ngtemplate>
```
In this case, `ngFor` should be a literal attribute with no RHS value.
Therefore, its source span should be just the `keySpan` and not the
source span of the original template node.
This allows language service to precisely pinpoint different spans in a
microsyntax to provide accurate information.
PR Close#38766
Previously, localized strings had very limited or incorrect source-mapping
information available.
Now the i18n AST nodes and related output AST nodes include source-span
information about message-parts and placeholders - including closing tag
placeholders.
This information is then used when generating the final localized string
ASTs to ensure that the correct source-mapping is rendered.
See #38588 (comment)
PR Close#38645
The `MessagePiece` and derived classes, `LiteralPiece` and `PlaceholderPiece`
need to be referenced in the `LocalizedString` output AST class, so that we
can render the source-spans of each piece.
PR Close#38645
The `TagPlaceholder` can contain children, in which case there are two source
spans of interest: the opening tag and the closing tag. This commit now allows
the closing tag source-span to be tracked, so that it can be used later in
source-mapping.
PR Close#38645
The `R3TargetBinder` accepts an interface for directive metadata which
declares types for `input` and `output` objects. These types convey the
mapping between the property names for an input or output and the
corresponding property name on the component class. Due to
`R3TargetBinder`'s requirements, this mapping was specified with property
names as keys and field names as values.
However, because of duck typing, this interface was accidentally satisifed
by the opposite mapping, of field names to property names, that was produced
in other parts of the compiler. This form more naturally represents the data
model for inputs.
Rather than accept the field -> property mapping and invert it, this commit
introduces a new abstraction for such mappings which is bidirectional,
eliminating the ambiguous plain object type. This mapping uses new,
unambiguous terminology ("class property name" and "binding property name")
and can be used to satisfy both the needs of the binder as well as those of
the template type-checker (field -> property).
A new test ensures that the input/output metadata produced by the compiler
during analysis is directly compatible with the binder via this unambiguous
new interface.
PR Close#38685
Previously, the compiler was not able to display template parsing errors as
true `ts.Diagnostic`s that point inside the template. Instead, it would
throw an actual `Error`, and "crash" with a stack trace containing the
template errors.
Not only is this a poor user experience, but it causes the Language Service
to also crash as the user is editing a template (in actuality the LS has to
work around this bug).
With this commit, such parsing errors are converted to true template
diagnostics with appropriate span information to be displayed contextually
along with all other diagnostics. This majorly improves the user experience
and unblocks the Language Service from having to deal with the compiler
"crashing" to report errors.
PR Close#38576
Previously, the `sourceSpan` and `startSourceSpan` were the same
object, which meant that you had the following situation:
```
element = <div>some content</div>
sourceSpan = <div>
startSourceSpan = <div>
endSourceSpan = </div>
```
This made `sourceSpan` redundant and meant that if you
wanted a span for the whole element including its content
and closing tag, it had to be computed.
Now `sourceSpan` is separated from `startSourceSpan`
resulting in:
```
element = <div>some content</div>
sourceSpan = <div>some content</div>
startSourceSpan = <div>
endSourceSpan = </div>
```
PR Close#38581
Previously, the `startSourceSpan` property could be null
but in reality it is always well defined - except for a legacy
case in the old i18n extraction/merging code, where the
typings for source-spans are already being undermined.
Making this property non-null, simplifies code elsewhere
in the project.
PR Close#38581
Previously the lexer was responsible for deciding whether an "inline"
template should also have its line-endings normalized.
Now this decision is made higher up in the call stack to allow more
flexibility in the parser/lexer.
PR Close#38581
Prior to this change, the unary + and - operators would be parsed as `x - 0`
and `0 - x` respectively. The runtime semantics of these expressions are
equivalent, however they may introduce inaccurate template type checking
errors as the literal type is lost, for example:
```ts
@Component({
template: `<button [disabled]="isAdjacent(-1)"></button>`
})
export class Example {
isAdjacent(direction: -1 | 1): boolean { return false; }
}
```
would incorrectly report a type-check error:
> error TS2345: Argument of type 'number' is not assignable to parameter
of type '-1 | 1'.
Additionally, the translated expression for the unary + operator would be
considered as arithmetic expression with an incompatible left-hand side:
> error TS2362: The left-hand side of an arithmetic operation must be of
type 'any', 'number', 'bigint' or an enum type.
To resolve this issues, the implicit transformation should be avoided.
This commit adds a new unary AST node to represent these expressions,
allowing for more accurate type-checking.
Fixes#20845Fixes#36178
PR Close#37918
Within an angular template, when a character entity is unable to be parsed, previously a generic
unexpected character error was thrown. This does not properly express the issue that was discovered
as the issue is actually caused by the discovered character making the whole of the entity unparsable.
The compiler will now instead inform via the error message what string was attempted to be parsed
and what it was attempted to be parsed as.
Example, for this template:
```
<p>
ģp
</p>
```
Before this change:
`Unexpected character "p"`
After this change:
`Unable to parse entity "ģp" - hexadecimal character reference entities must end with ";"`
Fixes#26067
PR Close#38319
This commit fixes a bug in View Engine whereby the compiler errorneously
thinks that a method of a component has decorator metadata when that
method is one of those in `Object.prototype`, for example `toString`.
This bug is discovered in v10.0.4 of `@angular/language-service` after
the default bundle format was switched from ES5 to ES2015.
ES5 output:
```js
if (propMetadata[propName]) {
decorators.push.apply(decorators, __spread(propMetadata[propName]));
}
```
ES2015 output:
```js
if (propMetadata[propName]) {
decorators.push(...propMetadata[propName]);
}
```
The bug was not discovered in ES5 because the polyfill for the spread
operator happily accepts parameters that do not have the `iterable`
symbol:
```js
function __spread() {
for (var ar = [], i = 0; i < arguments.length; i++)
ar = ar.concat(__read(arguments[i]));
return ar;
}
```
whereas in es2015 it’ll fail since the iterable symbol is not present in
`propMetadata['toString']` which evaluates to a function.
Fixes https://github.com/angular/vscode-ng-language-service/issues/859
PR Close#38292
HTML is very lenient when it comes to closing elements, so Angular's parser has
rules that specify which elements are implicitly closed when closing a tag.
The parser keeps track of the nesting of tag names using a stack and parsing
a closing tag will pop as many elements off the stack as possible, provided
that the elements can be implicitly closed.
For example, consider the following templates:
- `<div><br></div>`, the `<br>` is implicitly closed when parsing `</div>`,
because `<br>` is a void element.
- `<div><p></div>`, the `<p>` is implicitly closed when parsing `</div>`,
as `<p>` is allowed to be closed by the closing of its parent element.
- `<ul><li>A <li>B</ul>`, the first `<li>` is implicitly closed when parsing
the second `<li>`, whereas the second `<li>` would be implicitly closed when
parsing the `</ul>`.
In all the cases above the parsed structure would be correct, however the source
span of the closing `</div>` would incorrectly be assigned to the element that
is implicitly closed. The problem was that closing an element would associate
the source span with the element at the top of the stack, however this may not
be the element that is actually being closed if some elements would be
implicitly closed.
This commit fixes the issue by assigning the end source span with the element
on the stack that is actually being closed. Any implicitly closed elements that
are popped off the stack will not be assigned an end source span, as the
implicit closing implies that no ending element is present.
Note that there is a difference between self-closed elements such as `<input/>`
and implicitly closed elements such as `<input>`. The former does have an end
source span (identical to its start source span) whereas the latter does not.
Fixes#36118
Resolves FW-2004
PR Close#38126
Trotyl
Charles Lyding
Krzysztof Grzybek
Bjarki
Marcono1234
Pete Bacon Darwin
Alex Rickabaugh
JoostK
Andrew Scott
Kristiyan Kostadinov
ayazhafiz
Keen Yee Liau
Keen Yee Liau
JiaLiPassion
Joey Perrott