Renderer: align ordered list items on the widest marker of their list

[rosekanari]

Ordered list items each hang under their own marker, so their text starts at a different x. This aligns every item of a list on the widest marker of that list.

The problem

renderListItem calls applyListParagraphStyle(_:hangingUnder:) with that item’s own prefix, which delegates to applyHangingIndent(_:indent:hangingUnder:):

style.firstLineHeadIndent = indent
style.headIndent = indent + width(of: prefix)

The body font is not monospaced, so prefixes of different items have different widths. Measured at a 21 pt body:

prefix | width | vs "1. " -- | -- | -- "1. " | 18.47 pt | — "9. " | 21.95 pt | +3.49 "10. " | 31.00 pt | +12.53 "100. " | 43.68 pt | +25.21

A ten-item list is enough to see it: the text of item 10 starts 12.53 pt further right than that of item 1. Nothing overflows — no width is imposed — but nothing lines up either.

The change

render pre-scans the joined lines and, for each contiguous run of ordered items at one indent level, computes the widest prefix; every item of the run hangs under it. The run is fully known before any item is rendered, so the widest prefix is too.

renderLine and renderListItem take it as an optional parameter defaulting to nil, so unordered items, task items and every other caller are untouched.

Two files, 69 insertions, 6 deletions.

Tests

testOrderedItemsOfOneListShareTheirHangingIndent is added. It renders a twelve-item list and asserts that all items share one headIndent. It fails on unpatched main and passes with this change. The full suite passes.

testWiderOrderedMarkerHangsFurther keeps passing, and for the right reason: it renders "1. alpha" and "10. alpha" as two separate one-item documents, so each still hangs under its own — and only — marker. Its stated intent, “a wider marker has to hang further, or 10. would overlap its own text”, is preserved: within a list, every item now hangs under the widest one, so the widest never overlaps its text.

Notes

Verified against 28d86fd. Built and tested with:

xcodebuild -project QuickMD.xcodeproj -scheme QuickMD \
           -destination 'platform=macOS' test CODE_SIGNING_ALLOWED=NO

Written with Claude Code.

Happy to split, rename, or drop the added test if you would rather keep the suite as it is

[b451c]

Merged locally with --no-ff (8049a64) and pushed to main - thank you, the run detection and the failing-then-passing test were exactly the right shape.

One follow-up on top of it (7abb18c), because the shared headIndent only aligns the wrapped lines: the first line of an item starts at firstLineHeadIndent plus its own marker glyphs, so "1. alpha" and "10. kappa" still began their text 12.5 pt apart, and the wrapped lines of "1." now hung 12.5 pt to the right of its own first line. The follow-up pushes the narrower markers right by (widest - own) via firstLineHeadIndent, so the markers are right-aligned under the widest one and every item's text shares one edge on the first line and on wrapped lines alike. It also measures the widest marker in the body font instead of counting characters ("9. " is wider than "1. "), and keeps a list together across blank lines and nested content. Tests read the x of the first text glyph per item through TextKit, so they check the visual result rather than a single indent field.

Will ship in the next release with credit in the CHANGELOG.

[b451c]

Shipped in v1.10.0: https://github.com/b451c/quickmd/releases/tag/v1.10.0 - credited in the CHANGELOG and release notes. Thank you!