P0-17: Add HTML tag folding support

[hybridmachine]

Summary

Code folding does not currently recognize HTML tags. Users cannot fold/collapse HTML elements like <div>...</div>, <section>...</section>, etc. The editor should support folding at HTML tag boundaries, consistent with how Notepad++ handles HTML folding.

Expected Behavior

• HTML tag pairs (e.g., <div>...</div>) should be foldable
• Nested tags should create nested fold regions
• Self-closing tags (e.g., <br/>, <img/>) should not create fold points

Current Behavior

Folding in HTML files does not recognize tag structure. Fold markers are missing or incorrect for HTML content.

[hybridmachine]

Implementation Plan

Overview

HTML tag folding does not work because the macOS port's applyLanguageToView() sets only generic fold properties (fold, fold.compact, fold.comment, fold.preprocessor). The HTML lexer (LexHTML.cxx) requires the fold.html property to be explicitly set to "1" for tag-based folding to activate. The Windows version handles this in ScintillaEditView::setXmlLexer() (line 1000), but the macOS port has no equivalent language-specific property configuration.

Root Cause

In LexHTML.cxx:1263, the fold logic checks:

const bool foldHTML = options.foldHTML;  // ← driven by "fold.html" property
const bool fold = foldHTML && options.fold;

Without fold.html=1, foldHTML is false, which makes fold false, disabling all HTML tag folding.

Step-by-Step Implementation

File to modify: macos/platform/lexer_styles.mm — applyLanguageToView() (lines 323–347)

1. After setting the generic fold properties (line 340), add a block that checks if the current lexer is "hypertext" or "xml" and sets the appropriate HTML/XML-specific fold properties:

// HTML/XML-specific fold properties
if (strcmp(lang.lexerName, "hypertext") == 0 || strcmp(lang.lexerName, "xml") == 0)
{
    ScintillaBridge_sendMessage(sci, SCI_SETPROPERTY, (uintptr_t)"fold.html", (intptr_t)"1");
    ScintillaBridge_sendMessage(sci, SCI_SETPROPERTY, (uintptr_t)"fold.hypertext.comment", (intptr_t)"1");
    ScintillaBridge_sendMessage(sci, SCI_SETPROPERTY, (uintptr_t)"fold.html.preprocessor", (intptr_t)"1");
}

These match the properties set by the Windows version in ScintillaEditView::setXmlLexer() (lines 998–1002).

Design Decisions

Decision	Choice	Rationale
Where to add the check	Inside applyLanguageToView()	Keeps all lexer property configuration in one place
Match on lexer name	"hypertext" and "xml"	These are the two lexer names that use LexHTML.cxx
Include fold.html.preprocessor	Yes, set to "1"	Enables folding for embedded <script> / <style> blocks, matching Windows behavior
Include fold.hypertext.comment	Yes, set to "1"	Enables folding for HTML comments, matching Windows behavior

Edge Cases

Scenario	Expected Behavior
Self-closing tags (<br/>, <img/>)	No fold point — handled by LexHTML.cxx already
Nested tags (<div><p>...</p></div>)	Nested fold regions — handled by LexHTML.cxx fold level tracking
Embedded <script> blocks	Foldable — enabled by fold.html.preprocessor
HTML comments (<!-- ... -->)	Foldable — enabled by fold.hypertext.comment
XML files	Also get tag folding — same lexer family
Switching from HTML to another language	Generic properties still apply; HTML-specific ones are harmless to other lexers

Testing Plan

• Open an HTML file — verify fold markers appear in the margin next to tag pairs
• Fold a <div>...</div> block — content should collapse
• Verify nested tags create nested fold regions
• Verify <br/> / <img/> self-closing tags do NOT create fold points
• Open an XML file — verify tag folding also works
• Open a C++ file — verify brace folding still works (no regression)
• Open a Python file — verify indent folding still works (no regression)
• Verify <script> and <style> blocks are foldable in HTML
• Verify HTML comments are foldable

File Change Summary

File	Action	Description
macos/platform/lexer_styles.mm	MODIFY	Add HTML/XML-specific fold properties in applyLanguageToView()

Estimated Size

Small — ~6 lines of code added to one file.

[hybridmachine]

Good root-cause read overall — the missing fold.html in macos/platform/lexer_styles.mm::applyLanguageToView() does look like the main blocker. A few corrections would tighten the plan before implementation:

1. fold.html.preprocessor is not part of the Windows parity path you cited. ScintillaEditView::setXmlLexer() sets fold, fold.compact, fold.html, and fold.hypertext.comment, but not fold.html.preprocessor. In lexilla/lexers/LexHTML.cxx, foldHTMLPreprocessor already defaults to true, so setting it is optional clarity, not required for this bug.
2. fold.hypertext.comment does not enable plain HTML comment folding. In LexHTML.cxx it is used for comments inside embedded script blocks; <!-- --> / CDATA folding already goes through the HTML/SGML path once fold.html is enabled.
3. The lexer-family scope is wider than hypertext + xml. This repo also defines PHP as phpscript, and LexHTML.cxx exports a phpscript lexer. If the goal is parity with upstream setXmlLexer() or consistent behavior across HTML-family files, either include phpscript or explicitly say PHP-family files are out of scope for P0-17: Add HTML tag folding support #97.
4. If XML parity is in scope, upstream also sets lexer.xml.allow.scripts=0 for the XML case. The macOS plan currently omits that, so it does not fully mirror the Windows setup for XML.
5. The test plan should avoid treating <script>/<style> or HTML comments as proof that fold.html.preprocessor / fold.hypertext.comment worked — those exercise different code paths. The core acceptance checks here are still tag pairs, nesting, and self-closing tags.

My recommendation would be to keep the fix centered on enabling fold.html, then add the other properties only if you want explicit parity for embedded-script comments or XML/PHP behavior.

[hybridmachine]

Updated Implementation Plan (v2)

Incorporating feedback from code review. Key corrections:

• Removed fold.html.preprocessor — it defaults to true in LexHTML.cxx, and the Windows setXmlLexer() doesn't set it either.
• Corrected fold.hypertext.comment description — it controls comments inside embedded script blocks, not plain <!-- --> comments. Plain HTML comment folding is already handled by the HTML/SGML path once fold.html is enabled. Including it for parity with Windows but with accurate documentation.
• Added phpscript lexer — the macOS port defines PHP with the phpscript lexer (language_defs.mm:143), which is also exported by LexHTML.cxx. Including it for consistent HTML-family behavior.
• Added lexer.xml.allow.scripts=0 for XML — matches the Windows setXmlLexer() behavior for pure XML files (line 987).
• Tightened test plan — focused on tag pairs, nesting, and self-closing tags as the core acceptance criteria.

---

Root Cause

Unchanged. In LexHTML.cxx:1263:

const bool foldHTML = options.foldHTML;  // driven by "fold.html" property
const bool fold = foldHTML && options.fold;

Without fold.html=1, all HTML tag folding is disabled.

Step-by-Step Implementation

File to modify: macos/platform/lexer_styles.mm — applyLanguageToView() (lines 323–347)

After the generic fold properties (line 340), add:

// HTML-family fold properties (hypertext, phpscript, xml all use LexHTML.cxx)
if (strcmp(lang.lexerName, "hypertext") == 0 ||
    strcmp(lang.lexerName, "phpscript") == 0 ||
    strcmp(lang.lexerName, "xml") == 0)
{
    ScintillaBridge_sendMessage(sci, SCI_SETPROPERTY, (uintptr_t)"fold.html", (intptr_t)"1");
    ScintillaBridge_sendMessage(sci, SCI_SETPROPERTY, (uintptr_t)"fold.hypertext.comment", (intptr_t)"1");
}

// XML: disable embedded scripts (matches upstream setXmlLexer behavior)
if (strcmp(lang.lexerName, "xml") == 0)
{
    ScintillaBridge_sendMessage(sci, SCI_SETPROPERTY, (uintptr_t)"lexer.xml.allow.scripts", (intptr_t)"0");
}

Design Decisions

Decision	Choice	Rationale
Drop fold.html.preprocessor	Omit	Defaults to true in LexHTML.cxx; Windows doesn't set it either
Keep fold.hypertext.comment	Include, set to "1"	Parity with Windows setXmlLexer() — enables folding for comments in embedded scripts
Include phpscript lexer	Yes	Uses LexHTML.cxx; Windows includes PHP in the same setXmlLexer() path
Add lexer.xml.allow.scripts=0 for XML	Yes	Matches Windows behavior — XML files shouldn't parse embedded scripts

Edge Cases

Scenario	Expected Behavior
Self-closing tags (<br/>, <img/>)	No fold point — handled by LexHTML.cxx
Nested tags (<div><p>...</p></div>)	Nested fold regions — handled by LexHTML.cxx fold level tracking
<!-- --> HTML comments	Foldable once fold.html=1 is set (HTML/SGML code path)
Comments inside <script> blocks	Foldable via fold.hypertext.comment
XML files	Tag folding works; embedded scripts disabled
PHP files	Tag folding works via phpscript lexer
Switching to non-HTML language	HTML-specific properties are ignored by other lexers

Testing Plan

Core acceptance (tag folding):

• Open an HTML file — verify fold markers appear next to tag pairs (<div>, <section>, etc.)
• Click a fold marker — content between open/close tags should collapse
• Verify nested tags (<div><p>...</p></div>) create nested, independently foldable regions
• Verify self-closing tags (<br/>, <img/>) do NOT create fold points

HTML-family coverage:

• Open an XML file — verify tag folding works
• Open a PHP file — verify tag folding works

Regression:

• Open a C++ file — verify brace folding still works
• Open a Python file — verify indent folding still works

File Change Summary

File	Action	Description
macos/platform/lexer_styles.mm	MODIFY	Add HTML-family fold properties in applyLanguageToView()

Estimated Size

Small — ~10 lines of code added to one file.