Promote characters to wide if they have space available after

[emil-e]

This PR improves glyph rendering by allowing characters to dynamically expand into adjacent empty space, rather than always being squashed to fit their allocated cell width.

What changed

Previously, a glyph was scaled to fit either 1 or 2 cells depending on whether it was marked as wide. Now, if a glyph is wider (in aspect ratio) than its allocated slot, the renderer checks if the following characters are spaces. If so, it shrinks those spaces to zero-width via Emacs display properties, effectively claiming their cell width for the glyph while maintaining monospace alignment.

Key details

• Lines are right-trimmed of trailing spaces, so positions at and past the newline represent empty space that can be freely claimed.
• On a genuinely full line, the column check prevents expansion before reaching the row boundary.
• Uses Emacs display properties (specifically (space :width 0)) to reclaim space without mutating buffer text.
• The aspect-ratio comparison ensures we only expand when the glyph is proportionally wider than the available slot.

Files changed

• src/Renderer.zig — added dynamic width expansion logic and tracking of column position in CellInfo
• src/emacs.zig — added char-after, space, and :width to interned symbols

[michaelneuper]

When trying to build I get some compiler errors: https://pastebin.com/fp0VtvDJ

[dakra]

@michaelneuper you have to use zig 15.2.

as I see you use homebrew:

brew uninstall zig
brew install zig@0.15

[emil-e]

@claude Review this

[claude]

Claude finished @emil-e's task in 8m 40s —— View job

---

Code review for #286

• Read CLAUDE.md files for repo guidelines
• Read changed files (Renderer.zig, emacs.zig, ghostel-test.el)
• Verify diff against origin/main (used merge ref pull/286/merge)
• Identify findings (BUG/QUESTION/NIT)
• Post inline comments and summary

Summary

The patch is a nice generalization — handing each adjustable glyph off to its own function, hoisting selected-window out of the loop, and using (space :width 0) to reclaim cells is clean. The putTextProperty rework into a comptime symbol lookup is a real improvement: every callsite is now compile-time checked against the intern table.

The native test suite is red on all four test-native jobs (Linux/macOS × 29.4/snapshot). The other jobs are green.

Findings

BUG — adjustGlyph over-claims and claims-following-space fails (src/Renderer.zig:747-772). The break in the claim loop is if (glyph_aspect < cell_aspect) break; (strict <), and the past-EOL path is continue rather than break. Together those two let char_width step past the exact aspect-ratio match and keep growing as long as there's room left in the row. The failing test traces to char_width=4 (min-width=(4)) where the test expects (2). Inline comment has the per-iteration trace and a question about intended semantics — depending on the answer, either the break logic changes or the test expectation changes. Either way the four native jobs need to go green before merge.

QUESTION — char-after guards (src/Renderer.zig:761). The claim_pos >= row_end - 1 check above is the only thing that keeps the char-after call inside a valid range, and the else break is the only thing that stops (space :width 0) being splashed onto non-space buffer text. Both are correct today, but they're tightly coupled with the past-EOL continue, so changing one should re-check the other.

NIT — putTextProperty signature surfaces an implicit constraint (src/emacs.zig:307-314). Resolving prop via @field(sym, prop) gives compile-time checking (great), but the requirement that prop exists in interned_symbols isn't visible from the signature. A one-line doc comment pointing at the intern table would save the next contributor a head-scratch.

NIT — tab vs space indentation in the new tests (test/ghostel-test.el:15283-15391). The new glyph-adjust-* blocks are indented with tabs while the rest of the file uses spaces. untabify / indent-region would fix it.

Not findings

• col tracking through SPACER_TAIL / SPACER_HEAD is correct because Zig runs the while … : (col += 1) continue-expression on continue, so a wide cell + its tail still advances col by 2.
• The min_width / height display spec shape (((min-width (N)) (height S))) round-trips correctly with (assq 'min-width …) in the tests.
• The bold-color tests are not new — they were moved earlier in the file to sit next to the new glyph tests, no behavior change.
• Compile errors in @michaelneuper's pastebin are from Zig 0.14 vs 0.15 and unrelated to this PR.
  • Branch: emil-e/promote-wide