[andiwand]

🤖 Generated with Claude Code

Refs #764.

Summary

HtmlConfig::lazy_layout (default on) turns level 1 of docs/design/lazy-loading.md off. Then no view writes a chunk or the style that hides one. The browser lays out everything, and its own find and a screen reader reach every line, row and page. A host can choose that for small files.

• Text and sheets write all their lines or rows as one block or table. pdf pages get no chunk class.
• chunk_classes(config, near) gives a chunk's classes, empty when the switch is off, and write_chunk_style writes nothing then.
• The text styles move from .odr-chunk to .odr-text-lines, so a text page without chunks keeps its line numbers.
• The apple (lazyLayout), jni (lazyLayout) and python (lazy_layout) bindings carry it. wasm does not mirror HtmlConfig.
• The chunk size stays internal, because the measurements were flat around it. The design doc says so.

The reference output changes:

• 5 txt pages, for the new class
• 1,158 pdf pages, because the chunk classes now follow the size classes
• text.css and text.js, where main's text.js had changed in the merge of feat(html): lay out only the lines of a text file near the viewport #1247

After the class change is reverted and the class order is ignored, every changed page is identical to the reference. Pins: public 1ff0147, private 07c9ef2.

Test

• html.lazy_layout_off_writes_no_chunk: a text file, a sheet and a pdf with the switch off have no chunk class and no hiding style. All 2,530 tests pass.
• The jni ctest and the python test_html.py (12) pass with the round trip of the field. Apple is left to CI.
• test/browser/plaintext: 54 checks, 0 failed.

Review fixes

• CHANGELOG.md had no entry for the lazy layout of feat(html): lay out only the lines of a text file near the viewport #1247, feat(html): lay out only the pdf pages near the viewport #1248 and feat(html): lay out only the chunks of a sheet near the viewport #1250, or for this switch. A consumer notices both: the find of the browser does not reach a far part, and HtmlConfig has a new field. Two entries under HTML output now state them.
• The jni test checks the round trip with assertFalse(readBack.lazyLayout), not assertTrue(!…).
• Checked again: 964 C++ tests pass. The public txt and pdf pages are byte-identical to 1ff0147. test/browser/plaintext: 54 checks, 0 failed. I did not run the jni test after the change of its assert.