[webkit-changes] [WebKit/WebKit] 0e7b8c: Add enough BNF support for all CSS property longha...

[Sam Weinig]

  Branch: refs/heads/main
  Home:   https://github.com/WebKit/WebKit
  Commit: 0e7b8c42e6a0955ba6664560ecd90648f03148a1
      https://github.com/WebKit/WebKit/commit/0e7b8c42e6a0955ba6664560ecd90648f03148a1
  Author: Sam Weinig <weinig at apple.com>
  Date:   2023-01-02 (Mon, 02 Jan 2023)

  Changed paths:
    M Source/WebCore/css/CSSProperties.json
    M Source/WebCore/css/process-css-properties.py
    M Tools/Scripts/webkitpy/style/checkers/jsonchecker.py

  Log Message:
  -----------
  Add enough BNF support for all CSS property longhand grammars
https://bugs.webkit.org/show_bug.cgi?id=249980
rdar://103795673

Reviewed by Antti Koivisto.

Adds the remaining BNF parsing support in process-css-properties.py to
parse all the CSS property and descriptor definitions specified by CSS
specifications. This just adds the parsing of the grammars themselves
though, and generation of the actual C++ parsers is still to come. But,
with all the grammars now supported, we can enforce that a grammar is
at least added to CSSProperties.json for each longhand. And in fact,
what this change does is add a new "parser-grammar-unused" field that
is required for all longhands that are currently using a custom parser
function, as well as a "parser-grammar-unused-reason" field that has
the reason it is still using the custom parser function, usually due
to some kind of term we don't support generation for yet.

The majority of this change is going through each property and descriptor
and adding the unused grammars and reasons, as well as updating the
specification metadata and "values" arrays. For the "values" arrays,
we now have a hard check that any keyword the grammar supports is listed
in the "values" array as well. Eventually, we should remove the "values"
array entirely, and derive it from the grammars, but to do that, we need
to update the "css-status" website to support parsing the grammars as
well, so for now, we just enforce they are in sync.

A few other smaller changes include:

- Requiring "parser-grammar", "parser-function" or "skip-parser" for
each longhand ("custom-parser" is only used for shorthands, and should
be removed in a subsequent change).
- Renaming "parser-requires-*" fields to "parser-function-requires-*"
to better indicate that these only are used when "parser-function" is
used as well.
- Renaming "custom-parser-allows-number-or-integer-input" to
"parser-function-allows-number-or-integer-input" for the same reason.
- Added a new --dump-unused-grammars argument to process-css-properties.py
to print out all the properties and descriptors with unused grammars
as well as the "reasons".
- Added a new --check-unused-grammars-values argument to process-css-properties.py
to check "values" arrays against the unused grammars. This is not
checked by default as at least initially, I just want the unused
grammars to be checked for whether they parse.
- Added support for literals (e.g. the '[' the <line-names> production in css-grid),
functional notation (e.g. 'rect( [ <length> | auto ]#{4} )' in css-masking)
and functional references (e.g. '<rect()>' in css-masking as well) to the
BNF parser.
- Add support for all the BNF nodes to the Term class cluster, adding
LiteralTerm, GroupTerm, OptionalTerm, BoundedRepetitionTerm, FixedSizeRepetitionTerm,
and FunctionTerm. Also renamed and expanded CommaSeparatedRepetitionTerm
to UnboundedRepetitionTerm.

* Source/WebCore/css/CSSProperties.json:
* Source/WebCore/css/process-css-properties.py:
* Tools/Scripts/webkitpy/style/checkers/jsonchecker.py:

Canonical link: https://commits.webkit.org/258381@main