perf(render): elide redundant SGR escapes in the map render

[mparrett]

What

Adds a small render-only "pen" to the map render that skips re-emitting a set-fg/set-bg whose color already matches the last one emitted. move-cursor and write are untouched, and iteration order is unchanged, so the output is the same — same glyphs, same colors, fewer escape sequences.

Why

term's set-fg/set-bg write an SGR escape on every call, and the map loop calls them for every cell unconditionally. SGR attributes persist across cursor moves, so any cell whose color equals the previous one's is re-sending bytes for no visual change. On native that's wasted output; on the WASM build each call is a crossing into the JS host and xterm.js, so the waste is larger there.

How it works

• Eliding across cursor moves is safe. SGR state is sticky: moving the cursor doesn't reset the pen color, so skipping a redundant set-fg is correct as long as nothing reset the terminal's color in between.
• Reset discipline. The pen mirrors the terminal's current color, valid only while every styled write goes through it. Panels, vfx overlays, and the full-clear emit raw and leave it stale, so each contiguous penned batch resets the pen before its first cell.

Where it helps

Elision fires where same-color cells are already adjacent in emit order, so it does most of its work on the full-screen repaints — floor change, camera pan, resize — which walk the map row-major and are the biggest single render spikes. The per-move FOV-delta path iterates in set order and elides little; that's left as-is on purpose. Sorting it into runs would raise the elision rate slightly, but fire/water/lava tiles draw their color from the global rand stream, so reordering iteration would change which cell consumes which draw. Harmless in itself (that stream is unseeded, so the shimmer already varies run to run) but a needless behavior change for ~1% fewer escapes, so base iteration order stands.

Scope

Modest on its own — most frames are per-move FOV-delta renders, where there's little to elide. Its value is cutting the full-repaint spikes, and serving as the foundation for a frame-to-frame appearance diff (a follow-up PR stacked on this one), which is where the bulk of the render savings come from.

Safety

Rendering is a pure side effect — it doesn't touch the seed, action log, or world state — so this cannot affect determinism or replay. Static terrain and glyphs render identically before and after (verified by comparing the displayed cells, colors included, over a fixed input sequence); animated tiles already vary run to run via the unseeded rand stream, unaffected either way. Existing test suite passes unchanged.

[github-actions]

PR Preview Action v1.8.1
Preview removed because the pull request was closed.
2026-07-20 18:36 UTC

[mparrett]

Reviews complete:

• [#145](perf(render): elide redundant SGR escapes in the map render #145): Looks sound. Pen state is reset at the relevant raw-render boundaries, including VFX restoration and dirty-map batches.

• [#146](perf(render): skip frame-to-frame unchanged cells (appearance diff) #146): Looks sound atop perf(render): elide redundant SGR escapes in the map render #145. The appearance cache is correctly cleared for full/camera redraws and invalidated for entity movement and VFX restoration.

• [#149](perf(render): synchronized output (DEC 2026) across all frame renders #149): Needs a fix before I’d merge it.

  It claims to synchronize every full-screen frame, but only updates xsofy.render. The animated title (xsofy.title/draw-frame), animated descent screen, and xsofy.screenfx/clear-screen still emit large frames with plain term/flush, so they remain prone to the tearing this PR targets. The sync helpers should be shared (likely via screenfx) and used for those paths too.

All three PR heads are mergeable and their CI test runs are green. The stack wiring is correct: #145 → #146 → #149. After merging each lower PR, retarget/rebase the next PR onto main before merging it.