After the latest fixes, I tried updating the submodules, and can see we still have some syntax test failures, this time relating to embed...escape:
FAILED testdata/Packages\Java\syntax_test_java.java: 817
FAILED testdata/Packages\Markdown\syntax_test_markdown.md: 61
We get the same failures in Sublime Text, when rewriting the embed...escape rules to with_prototype like syntect's yaml_load is doing, so the good news is that syntects parser is working correctly:
JavaDoc.sublime-syntax:
--- Shipped Packages/Java/JavaDoc.sublime-syntax 2018-10-11 19:11:54
+++ Packages/Java/JavaDoc.sublime-syntax 2018-10-30 11:28:20
@@ -20,11 +20,12 @@
- meta_include_prototype: false
- match: /\*\*
scope: comment.block.documentation.javadoc punctuation.definition.comment.begin.javadoc
- embed: contents
- embed_scope: comment.block.documentation.javadoc text.html.javadoc
- escape: \*/
- escape_captures:
- 0: comment.block.documentation.javadoc punctuation.definition.comment.end.javadoc
+ push:
+ - [{ meta_include_prototype: false }, { meta_content_scope: 'comment.block.documentation.javadoc text.html.javadoc' }, { match: '\*/', scope: comment.block.documentation.javadoc punctuation.definition.comment.end.javadoc, pop: true }]
+ - contents
+ with_prototype:
+ - match: (?=\*/)
+ pop: trueMarkdown.sublime-syntax:
--- Shipped Packages/Markdown/Markdown.sublime-syntax 2018-10-11 19:11:54
+++ Packages/Markdown/Markdown.sublime-syntax 2018-10-30 09:44:46
@@ -1106,12 +1106,12 @@ contexts:
0: meta.code-fence.definition.begin.html-php.markdown-gfm
2: punctuation.definition.raw.code-fence.begin.markdown
5: constant.other.language-name.markdown
- embed: scope:embedding.php
- embed_scope: markup.raw.code-fence.html-php.markdown-gfm
- escape: '{{code_fence_escape}}'
- escape_captures:
- 0: meta.code-fence.definition.end.html-php.markdown-gfm
- 1: punctuation.definition.raw.code-fence.end.markdown
+ push:
+ - [{ meta_include_prototype: false }, { meta_content_scope: 'markup.raw.code-fence.html-php.markdown-gfm' }, { match: '{{code_fence_escape}}', captures: { 0: meta.code-fence.definition.end.html-php.markdown-gfm, 1: punctuation.definition.raw.code-fence.end.markdown }, pop: true }]
+ - scope:embedding.php
+ with_prototype:
+ - match: (?={{code_fence_escape}})
+ pop: true
- match: |-
(?x)
{{fenced_code_block_start}}For Java(Doc), the workaround suggested at #160 (comment) works.
For Markdown, the following syntax test is enough to see the problem:
| SYNTAX TEST "Packages/Markdown/Markdown.sublime-syntax"
```html+php
<div></div>
|^^^ entity.name.tag.block.any.html
<?php
|^^^^ punctuation.section.embedded.begin.php
var_dump(expression);
| ^^^^^^ support.function.var.php
```
|^^ punctuation.definition.raw.code-fence.end.markdown
Everything except the last line's assertions pass. Adding ?> before the closing backticks/code fence allows the backticks to be correctly recognized as the closing code fence instead of PHP interpolated string punctuation, but we will need a solution so that the way syntect handles embed actions will work 100% correctly in all situations.
Altering the PHP Source syntax definition can help to work around it, which (although obviously not a solution,) suggests that, for some reason, the with_prototype (rewritten from the embed in the Markdown syntax definition) is not being considered at this point (in both Sublime and syntect):
--- Shipped Packages/PHP/PHP Source.sublime-syntax 2018-10-11 19:11:54
+++ Packages/PHP/PHP Source.sublime-syntax 2018-10-30 13:36:58
@@ -109,6 +109,8 @@
- include: statements
expressions:
+ - match: (?=```)
+ pop: true
- include: comments
- match: (?i)\b((?:require|include)(?:_once)?)\b\s*
captures:
After the latest fixes, I tried updating the submodules, and can see we still have some syntax test failures, this time relating to
embed...escape:We get the same failures in Sublime Text, when rewriting the
embed...escaperules towith_prototypelikesyntect'syaml_loadis doing, so the good news is thatsyntects parser is working correctly:JavaDoc.sublime-syntax:Markdown.sublime-syntax:For Java(Doc), the workaround suggested at #160 (comment) works.
For Markdown, the following syntax test is enough to see the problem:
Everything except the last line's assertions pass. Adding
?>before the closing backticks/code fence allows the backticks to be correctly recognized as the closing code fence instead of PHP interpolated string punctuation, but we will need a solution so that the waysyntecthandlesembedactions will work 100% correctly in all situations.Altering the
PHP Sourcesyntax definition can help to work around it, which (although obviously not a solution,) suggests that, for some reason, thewith_prototype(rewritten from theembedin the Markdown syntax definition) is not being considered at this point (in both Sublime andsyntect):