Never fuse atoms across a font-style change

[kostub]

Plan: docs/plans/2026-08-04-mathit-text-italic-routing.md (PR 1 of 3, items 1-2)
LLD: docs/lld/2026-07-27-mathit-text-italic-routing.md

Goal

Both atom-merge sites decline to fuse atoms whose fontStyle differs, so a style set on part of a run survives to the typesetter.

fuse: keeps the first atom's properties, including its fontStyle, so an unguarded merge either drops a style or spreads it onto an unstyled atom. Neither site checked style, and they cover disjoint inputs: site A merges only Number atoms, site B merges Ordinary atoms after preprocessMathList: has rewritten variables and numbers. 1\mathit{2} is fixed only by A, f\mathit{x} only by B.

This independently fixes a live rendering bug: 1\mathbf{2} renders its 2 non-bold today, because site A's merge discards the bold before getBold can map the digit to U+1D7CE (LLD §2.6). Geometrically neutral for uniform-style input.

Commits

1. [item 1] Guard digit fusion on fontStyle in MTMathList.finalized
2. [item 2] Guard Rule-14 fusion on fontStyle in preprocessMathList

Full swift test green at each commit (451 tests).

Summary by CodeRabbit

• Bug Fixes
  • Fixed math rendering so adjacent numbers and symbols with different font styles remain separate.
  • Prevented italic, bold, and mixed-style content from being incorrectly merged.
  • Continued combining adjacent content when it uses the same font style, preserving existing formatting behavior.
  • Improved consistency of font-style handling during math layout and rendering.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro Plus

Run ID: 8e846eb6-645f-46bd-8a06-9759e0fbbc18

📥 Commits

Reviewing files that changed from the base of the PR and between 6d0cb6e and fdd18ab.

📒 Files selected for processing (4)

• iosMath/lib/MTMathList.m
• iosMath/render/internal/MTTypesetter.m
• iosMathTests/MTMathListTest.m
• iosMathTests/MTTypesetterTest.m

---

📝 Walkthrough

Walkthrough

The change prevents number and ordinary-atom fusion across different font styles. Uniformly styled atoms continue to fuse. Regression tests cover mixed, default, italic, bold, and unstyled inputs.

Changes

Font-style-aware atom fusion

Layer / File(s)	Summary
Number finalization preserves font styles iosMath/lib/MTMathList.m, iosMathTests/MTMathListTest.m	Number fusion now requires matching fontStyle values. Tests verify mixed-style runs remain separate and uniform-style runs still fuse.
Typesetter preprocessing preserves font styles iosMath/render/internal/MTTypesetter.m, iosMathTests/MTTypesetterTest.m	Ordinary-atom fusion now requires matching fontStyle values. Tests cover style boundaries, bold digit conversion, uniform fusion, and unstyled lists.

Estimated code review effort: 2 (Simple) | ~10 minutes

Possibly related PRs

• kostub/iosMath#267: Both PRs modify atom fusion and preprocessing, but address different behaviors.

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly and concisely describes the main change: preventing atom fusion across font-style changes.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch feature/mathit-routing-pr1

---

_{Comment @coderabbitai help to get the list of available commands.}

[gemini-code-assist]

Caution

The consumer version of Gemini Code Assist on GitHub has been sunset. All code review activity has officially ceased.

[kostub]

Code review — PR 1 of 3 (fontStyle guard at both fusion sites)

No blocking findings. The diff is two extra conjuncts and their tests; I could not find a defect, and there is no machinery here to collapse.

What I verified

• Both guards are load-bearing; neither is redundant. Site A fuses before changeFont runs, so with only the site B guard 1\mathbf{2} still arrives at preprocessMathList: as a single Default Number and the bold 2 is lost. With only the site A guard, the two resulting Ordinary atoms re-fuse at site B. Reverting either source hunk fails testPreprocessDoesNotFuseAcrossFontStyles.
• The new tests genuinely fail without the change. Reverted MTMathList.m + MTTypesetter.m to master on top of the PR's test files: testFinalizedDoesNotFuseDigitsAcrossFontStyles and testPreprocessDoesNotFuseAcrossFontStyles fail; the three "still fuses" / "unchanged" methods pass, so they work as regression pins.
• grep confirms fuse: has exactly two call sites (MTMathList.m:1712, MTTypesetter.m:557) — no third merge path is left unguarded.
• Geometrically neutral, including for the newly-split mixed-style case. Splitting an Ord run into two Ord atoms adds nothing: getInterElementSpaces()[0][0] is kMTSpaceNone, so createDisplayAtoms adds no kern, and both nuclei are appended to the same _currentLine before a single addDisplayLine, so the CTLine string is byte-identical to the fused one. The superscript path is also unaffected — the italic correction is taken from the last character of the nucleus, which is the same character either way, and !prevNode.subScript && !prevNode.superScript already blocked fusing into a scripted atom.
• _currentAtoms / index-range bookkeeping is unchanged, because a fused atom already contributed its _fusedAtoms copies rather than itself.
• The \mathbf digit fix does not hit a font-coverage hole. U+1D7CE/U+1D7D0 are present in all eight bundled .otfs (checked via fontTools cmap), so the newly-preserved bold digit resolves in every bundled font.
• Full swift test at fdd18ab: 451 tests, 0 failures.

Notes, not requests

• The test-only @interface MTTypesetter (FusionTesting) category is the right call over adding an MTTypesetter+Internal.h: it exposes nothing outside the test binary and costs three lines.
• \mathsf digits map to U+1D7E2, which firamath.otf lacks (the only bundled font that does). This PR widens the reach of that pre-existing gap from \mathsf{2} to 1\mathsf{2}, but the gap and its CoreText-substitution behaviour are unchanged by this diff — flagging only so it isn't a surprise later.

Not approving — leaving the merge call to you.

[kostub]

@coderabbitai review

[coderabbitai]

✅ Action performed

Review finished.

Note: CodeRabbit is an incremental review system and does not re-review already reviewed commits. This command is applicable only when automatic reviews are paused.