fix: preserve grapheme clusters (combining marks) in cell emission - #114
Open
natemoo-re wants to merge 3 commits into
Open
fix: preserve grapheme clusters (combining marks) in cell emission#114natemoo-re wants to merge 3 commits into
natemoo-re wants to merge 3 commits into
Conversation
natemoo-re
commented
Aug 23, 2026
Comment on lines
+8
to
+10
| /* Maximum combining marks stored per cell. Marks beyond this limit are | ||
| * silently truncated from the end; the first CELL_MAX_COMBINING are kept. */ | ||
| #define CELL_MAX_COMBINING 8 |
Member
Author
There was a problem hiding this comment.
this gives us plenty of headroom for the common case and is unlikely to be a problem in the short-term, but I have a slight concern that it will be a long-term issue if any terminal protocols introduce complex metadata via ZWJ that requires >8 codepoints
an alternative design would be bumping this cieling and keeping a dynamic map of combining size per cell rather than reserving a flat 8 per cell.
natemoo-re
marked this pull request as ready for review
August 23, 2026 11:49
commit: |
|
Size Increased — +1.0 KB 101.1 KB unpacked |
Contributor
natemoo-re
force-pushed
the
nm/cell-grapheme
branch
from
August 28, 2026 00:17
07a84e9 to
d7c2d69
Compare
The Cell struct stored only a single uint32_t codepoint, so render_text silently dropped every combining mark (wcwidth ≤ 0): the emitted ANSI stream contained bare base codepoints with no continuations. Root cause: kitty-graphics Unicode placeholder cells require two combining diacritics (row/col index from the rowcolumn-diacritics table) to follow the base U+10EEEE codepoint. Because those marks were dropped, every placeholder cell emitted as row 0 — producing the N-repeated-top-band banding artifact in multi-row placements. The same drop affected any combining-mark or ZWJ content: accented characters (e + U+0301), flag pairs, skin-tone modifiers. Fix: - Cell gains `uint32_t combining[8]` (zero-terminated). Marks beyond 8 are silently truncated from the end; the first marks always survive. - cells_fill uses a designated-init template so combining[] is zeroed on every back-buffer reset, and setcell clears it on every base write. - render_text tracks the last-written column; zero-width codepoints go to append_combining() instead of being discarded. - present_cups / present_lines emit combining[] bytes immediately after the base character, before any cursor repositioning. - OUT_BYTES_PER_CELL bumped 64→128 to cover base + 8 combining marks. - cell_cmp checks combining[] so a changed mark triggers a diff/rediff. - Spec updated: §8.3.3 normative preservation requirement, §13 cluster cell representation with truncation semantics, §13 measurement note. Four new tests: kitty placeholder (U+10EEEE + 2 marks), combining accent (e + U+0301), ZWJ family emoji (ZWJ preserved per-cell; following emoji start new cells — inherent cell-model constraint, documented), and truncation-from-end pinned at 8 marks.
natemoo-re
force-pushed
the
nm/cell-grapheme
branch
from
August 28, 2026 02:21
d7c2d69 to
1a1f83d
Compare
cowboyd
reviewed
Sep 1, 2026
Comment on lines
+1035
to
+1037
| combining marks stores zero in every slot. When a text string produces more than | ||
| 8 combining marks for a single base codepoint, the excess marks are silently | ||
| truncated from the end (marks 1–8 are kept, marks 9+ are discarded), ensuring |
Collaborator
There was a problem hiding this comment.
Should this result in a warning being issued? It seems like if it was, then it should be issued only once.
cowboyd
reviewed
Sep 1, 2026
| cw = 1; | ||
| if (cw > 0) { | ||
| setcell(ct, x, y0, cp, fg, bg); | ||
| last_x = x; |
Collaborator
There was a problem hiding this comment.
We should not unconditionally set the last_x because setcell() could be a no-op in the event that it is outside the clip region:
if (ct->clipping) {
if (x < ct->clipx || x >= ct->clipx + ct->clipw)
return;
if (y < ct->clipy || y >= ct->clipy + ct->cliph)
return;
}
cowboyd
reviewed
Sep 1, 2026
Collaborator
There was a problem hiding this comment.
Can we make the tests visual as in other test cases? It would involve enhancing our print() helper (yet again!)
expect(trim(print(ansi, 12, 4))).toEqual(`
┌──────────┐
│café │
│ │
└──────────┘`.trim());We really ought to see if we can use libghostty for this (not now ofc, but sometime in the unspecified future)
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
cells with combining marks (zero-width codepoints) were silently dropped when emitted because
render_textskipped codepoints withcw == 0. the cell model stored only a singleuint32_t, meaning combining accents, ZWJ emoji, skin-tone modifiers, flag variation emoji, and kitty graphics were broken.the fix here is to update
Cell's model to trackuint32_t combining[8]and haverender_textappend zero-width codepoints to the current column before cursor output. this requires updatingOUT_BYTES_PER_CELLfrom64to128to cover worst-case scenario (base char + 8 combining marks, silently dropping marks beyond 8).spec §8.3.3 has been updated with a grapheme-cluster preservation requirement. §13 now describes cells as "a grapheme cluster" rather than "a Unicode codepoint" and defines the capacity and truncation behavior.
semi-related to #84