Motivation
geometry.col/opts.col is currently accepted by show()'s API but silently unused for placement — virt_lines always render starting at the window's text-area left edge regardless of the extmark's column, so there is no way to position an image horizontally within a line today. This is explicitly documented as "reserved for a future version" in docs/spec/renderer-placement.md's "Screen coordinate conversion" section, and was confirmed/tightened in a prior review pass (the LuaLS annotations, doc/blit.txt, and the spec memo were all made to consistently describe it as unused rather than silently broken).
Proposed Solution
Investigate a horizontal-positioning mechanism that doesn't fight virt_lines' left-edge-anchoring behavior — e.g. inline virtual text padding before the image's virtual line content, or a different extmark strategy entirely for this case. This needs its own design discussion (per AGENTS.md's "when a spec is ambiguous... STOP and ask" — there may be more than one valid approach with real trade-offs) and a docs/spec/renderer-placement.md update before implementation.
Alternatives Considered
- Status quo (column always ignored): acceptable for the common "image below a line" use case (e.g. markdown image previews), which is why this was deferred rather than blocking v0.x.
Additional Context
Low priority relative to the other v1.0-milestone issues — no known caller need yet, tracked so the col field's silent no-op isn't forgotten.
Checklist
Motivation
geometry.col/opts.colis currently accepted byshow()'s API but silently unused for placement —virt_linesalways render starting at the window's text-area left edge regardless of the extmark's column, so there is no way to position an image horizontally within a line today. This is explicitly documented as "reserved for a future version" indocs/spec/renderer-placement.md's "Screen coordinate conversion" section, and was confirmed/tightened in a prior review pass (the LuaLS annotations,doc/blit.txt, and the spec memo were all made to consistently describe it as unused rather than silently broken).Proposed Solution
Investigate a horizontal-positioning mechanism that doesn't fight
virt_lines' left-edge-anchoring behavior — e.g. inline virtual text padding before the image's virtual line content, or a different extmark strategy entirely for this case. This needs its own design discussion (perAGENTS.md's "when a spec is ambiguous... STOP and ask" — there may be more than one valid approach with real trade-offs) and adocs/spec/renderer-placement.mdupdate before implementation.Alternatives Considered
Additional Context
Low priority relative to the other v1.0-milestone issues — no known caller need yet, tracked so the
colfield's silent no-op isn't forgotten.Checklist