[yeroo]

What and why

HTML import (docxcore::import::html) used to keep an <a>'s text and drop its target. Now an
<a href> becomes an Inline::Hyperlink:

• An absolute http, https, ftp or mailto URL becomes the link's target. href="#name"
  becomes an anchor link (w:anchor, percent-decoded); # alone makes no link.
• Every other href stays plain text, as before. This covers javascript:, vbscript:, data:,
  file:, any other scheme, relative and empty hrefs.
  • link_of reads an href the way a browser's URL parser does. It trims controls at both ends
    and removes tab, LF and CR inside, so java	script: is caught. Then it checks an
    allow-list, so obfuscations fail closed.
  • Space, ", <, > and ` are percent-encoded.
  • The limit is 4096 bytes. An href longer than twice that is refused unread.
• How the link is carried. The link is part of each element's computed look. So the cascade,
  restyle, mis-nesting and the formatting-element reopen carry it exactly as they carry
  formatting. A link that crosses <p> boundaries becomes a link in each paragraph. An <a>
  closes an open <a> first, as a browser's adoption agency does.
• Look. Links get a browser's look (underline and Word's Hyperlink blue 0563C1) as direct
  formatting, and the page's CSS overrides it. color: inherit / currentcolor now gives back
  the parent's colour, over a link's blue or a <font color>. That makes Gmail's
  a{color:inherit;text-decoration:none} work.
• Breaks, tabs and pictures. A break, tab or picture inside a link is written outside it, and
  the link goes on after it. An unresolved image's alt text inside a link is the link's text.
• Relationships. None are made at import. save_package already mints an external
  relationship for every target-only link. A test shows the targets survive HtmlImport::package
  → save → load. The suite's converting child saves the same way, so an .htm opened in the suite
  shows its links.
• Attribute values. A named character reference without ; that is followed by = or an
  alphanumeric is kept as written, as HTML does. So ?id=3&copy=2 stays a correct target
  (review r1).

Notes are in docs/claude/html-import.md under "HTML import links (#1879)", with the known
limits.

Closes #1879

Tests

• html.rs link_tests (16):
  • formatting and the link look;
  • CSS overrides, including inherit;
  • anchors;
  • 22 unsafe or relative hrefs, including the obfuscations;
  • limits and encoding;
  • spaces inside and around links;
  • paragraphs, breaks and tabs;
  • pictures and alt text;
  • an empty link, and list markers;
  • nested <a>;
  • table cells;
  • &name= in attributes.
• docxcore/tests/html_links.rs over a new newsletter-style fixture links.html: links in
  order, and package save/load keeping targets (TargetMode="External") and w:anchor.
• html_css.rs: the Gmail fixture's mailto: link, and links.html in the every-prefix
  loop.
• Suite: convert_tests::a_web_pages_links_open_as_hyperlinks.
• Whole run on ae7e6b24:
  • It passed: cargo fmt, the workspace cargo clippy --all-targets,
    cargo test --workspace --locked, suite fmt / clippy / xvfb-run cargo test, the scripts'
    unittests and check-claude-md. 11954 tests passed, 0 failed.
  • Not run on this machine (left to CI): the html-export clippy/test configuration (no
    wasm32-unknown-unknown target here) and the UI sweep. In an earlier run of the sweep on
    2cc88c4, 139 of 144 passed and 1 was an expected failure. Of the 4 failures:
    • 3 hit the 218-character workbook path limit, because of this checkout's long sandbox path;
    • 1 was project-columns, in Project table code this PR does not touch.

Review record

• Plan: two rounds with the implementer before agreement (v2). That added the link look,
  inherit restoring the parent's colour, wider percent-encoding, links counted by their text for
  empty paragraphs, and alt text inside links.
• revmux round 1: no Major. Both minors were fixed in ae7e6b24:
  • Color::Keep's doc was stale;
  • a ;-less &name= was decoded in attribute values.
• Deferred to a follow-up: the Markdown export writes link destinations unescaped. That
  code predates this PR and lives in markdown.rs, which the PR does not touch.
• The plan's out-of-scope items, filed as follow-ups in Leftovers from #1879: HTML import: <a href> links become hyperlinks #1912:
  • bookmarks for <a name> / id;
  • the Hyperlink character style;
  • <base href> and relative links;
  • link title / target and hyperlinked pictures.
• Follow-ups: Leftovers from #1879: HTML import: <a href> links become hyperlinks #1912 (leftovers checklist).
• review stopped: round 1 had no Major; 1 minor finding(s) in Leftovers from #1879: HTML import: <a href> links become hyperlinks #1912

[codecov-commenter]

⚠️ Please install the to ensure uploads and comments are reliably processed by Codecov.

Codecov Report

❌ Patch coverage is 98.09725% with 9 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
docxcore/src/import/html.rs	98.09%	8 Missing ⚠️
docxcore/src/import/mod.rs	98.07%	1 Missing ⚠️

📢 Thoughts on this report? Let us know!

[yeroo]

Merged automatically (auto-merge is on for this checkout): review stopped: round 1 had no Major; 1 minor finding(s) in #1912; merge rounds: none; whole suite green on ae7e6b2 (11954 tests; html-export and the UI sweep ran in CI, 14 checks passed); merge-check ok.

Follow-ups, all in the leftovers issue #1912 (none are duplicates):

• Markdown export: escape hyperlink destinations (spaces, parentheses). This is the review r1 minor, deferred because it predates this PR and lives in markdown.rs, which the PR does not touch.
• Four out-of-scope items from the plan:
  • bookmarks for <a name> / id anchor targets;
  • the Hyperlink character style on imported links;
  • resolving relative links against <base href>;
  • link title / target and hyperlinked pictures.

No finding was disputed. Both other r1 minors were fixed in ae7e6b2.