summaryrefslogtreecommitdiff
path: root/docs
diff options
context:
space:
mode:
Diffstat (limited to 'docs')
-rw-r--r--docs/effects.md285
-rw-r--r--docs/next-anim.md1107
-rw-r--r--docs/next-pdfrender.md2
-rw-r--r--docs/render-pipeline.md949
-rw-r--r--docs/typ/building.typ4
5 files changed, 47 insertions, 2300 deletions
diff --git a/docs/effects.md b/docs/effects.md
deleted file mode 100644
index a9b27a6b..00000000
--- a/docs/effects.md
+++ /dev/null
@@ -1,285 +0,0 @@
-# Effects: feel reviews
-
-**The rule above all the others:** no effect may alter or cover a focus
-indicator (the focused pane's tag colour, its grip, the cursor, the
-selection) or reduce its contrast. The effects are sugar; the indicators are
-how a person knows where they are. Every G stage is reviewed against this.
-Lift, for one, falls only on pane bodies and rails, never on a tag, a grip or
-a header, and its strength is capped so text and the selection keep min(their
-contrast, 4.5) (tests in src/gui/gui.zig). The rule binds effects, not a
-theme's structure: the rules between panes (2px, `rule_px`) are the theme's
-borders, and the one over a tag band stays in the band's slack, never over
-its text. A dim touches only the unfocused
-panes, with the same floor, and the focused pane's text is never at less
-contrast than theirs (tests in src/draw.zig).
-
-**Density.** A pixel shell draws every size given in pixels in logical
-pixels, `dp()` in gui.zig (SDL's window display scale, rounded, at least
-one; `PARDES_TEST_SCALE` forces one in a harness, and a test window is 1x
-otherwise): the rules (`rule_px`), the tag rule, the grip's ring and its
-dirty gap and mode bars, the rail's gap to the text and its minimum, the
-thumb's inset, the rim, a bar cursor's minimum width, underline and
-strikethrough thickness, the rule-grab margin, and the Crt's scanline and
-mask periods (`pardesScale`). Left as they are: what is measured in cells
-or the tagline (the rail's width, `rail_px` scaled with the tagline; Lift's
-blur and offsets; every motion, in cells), Shadertoy's iResolution,
-fragCoord and iMouse (the display's pixels, as ghostty's), and the pet
-overlay.
-
-Each effect of docs/render-pipeline.md §9 lands behind its own switch, off,
-and is kept only after a feel review (§8.4): a frame series on the virtual
-clock, a recording, and a verdict — keep, polish or drop — with one line of
-why. The verdict is the user's.
-
-| effect | switch | state | review material | verdict |
-|---|---|---|---|---|
-| Crt (bundled post pass) | `Crt 0..3` | kept, rewritten (no barrel) | fx-compare stills, live window | kept at the user's live look |
-| Ripple, Glitch | — | removed | live window | dropped by the user after a live look |
-| G1 lift | `Lift shadow\|rim\|auto` (off), `InactiveDim <percent>` | opt-in until the focus-lift default is decided | lift-shots-2 (off, shadow, rim, auto on acme, dusk, forge); frame series + mp4 per Motion flavour | shadow on acme: keep as is. Round 1 dropped glow and surface (surface lowered the focused text's contrast) and made rim a hairline just above the focused tag (light on a dark page, shade on a light one). auto on a dark page recedes the others (InactiveDim 30) |
-| G3 cursor | follows `Motion`; `CursorBlink on\|off` (on) | glide on by default through the default flavour (smooth); blink on (decision 6) | .scratch/render/cursor/cursor-flavours-grid.mp4 (and quarter speed), raw/<flavour>/<motion>/ | pending |
-| G4 bloom, vignette, grain | `Bloom`, `Vignette`, `Grain` `0..3` (off) | bundled post passes; none touches a tag, grip, notice or the cursor (pardesSpare); each capped per theme (pardesCap) | .scratch/render/{bloom,vignette,grain}/ (levels 0-3 on forge, dusk, acme; crops; mp4 stepping through the levels) | pending |
-| G5 lapis | `Theme lapis` (its `decor`) | a native theme with ornament: tag plaques with hard shadows inside their bands (the focused one striped), dots, rail checker, title shadow | .scratch/render/lapis/ (whole window at 1x and 2x, crops, a slide mp4) | pending |
-| G6 smooth scroll | follows `Motion` (on through the default smooth; `Motion off` lands at once) | a mouse wheel's notch glides, a touchpad follows the finger | .scratch/render/scroll/ (glide-<flavour>.mp4, quarter speed, strips) | pending |
-| G7 panel transitions | `PanelSlide`, `PanelZoom`, `PanelVertical`, `PanelDissolve` (off) with `Motion` | a moving pane casts a shadow; a retargeted move keeps its speed; the GUI's dissolve is a crossfade | .scratch/render/transitions/ (slide smooth, bouncy, retargeted; dissolve) | pending |
-| G8 notice chips | follows `Motion` and `Lift` | a notice floats in tier 4 on a lift (shadow) while Lift is on, drops in easing out and dissolves easing in, quicker than it came (150 ms against 180) | .scratch/render/notices/ (acme, forge) | pending |
-| G9 selection glow | `SelectionGlow on\|off` (off) | a soft halo of the selection's colour round each block of it in a body, fading in over 100 ms | .scratch/render/selection/ (forge, acme, lapis; lapis-glow-zoom.png) | pending |
-| G10 theme crossfade | always (a theme change) | the chrome's colours fade through OKLab, easing in and out (out when retargeted) | .scratch/render/crossfade/ (forge, lapis, acme cut short by dusk) | pending |
-| G11 look-hover glow | `HoverGlow on\|off` (off) | a soft underline of the theme's accent under the word a look-hover would open, fading in over 80 ms | .scratch/render/hover/ (forge, acme, lapis; zooms) | pending |
-| G12 ambient occlusion | `Occlusion on\|off` (off) | pane bodies darken toward their edges, at most 2% (§9.1: 1 to 2%) | .scratch/render/occlusion/ (on/off on forge, acme, lapis; acme-diff-x40.png) | pending |
-| G13 parallax | `Parallax on\|off` (off) | a theme's page pattern under the bodies (lapis's dots) moves with the text at a quarter of its speed, glides included | .scratch/render/parallax/ (lapis, notches down and up) | pending |
-| G14 afterglow | — | removed | live lapis showcase | dropped by the user: "one of the animations is like a flashing on the whole pane or something, that is really bad, just remove it" (a pane whose whole text changed, a file opened, a +Search refilled, a terminal's output, lit up at once) |
-
-## Motion flavours
-
-`Motion off|crisp|smooth|bouncy|playful` (default smooth, tuned so arrivals
-land inside §8.1's 220 ms; bouncy and playful run longer, opt-in) is one parameter set
-(animation.Motion) every fx animation reads, so a flavour is data, not a
-branch in each effect. Input is never blocked, and a new target always
-retargets from the current position and velocity.
-
-| flavour | timing (ω, length) | follow-through (ζ) | anticipation | squash/stretch | exaggeration (gain) | secondary lag | arcs |
-|---|---|---|---|---|---|---|---|
-| off | instant | — | — | — | — | — | — |
-| crisp | 60, 0.6× (leaves at full speed) | 1 (none) | — | — | 1 | in step | — |
-| smooth | 34, 1.1× (slow in, slow out; ≤220 ms) | 1 (none) | — | — | 1 | 0.85 | 0.04 |
-| bouncy | 24, 1.4× | 0.35 (overshoots ~30%, settles twice) | — | 0.15 | 1.25 | 0.8 | 0.08 |
-| playful | 18, 1.7× | 0.28 (overshoots ~40%) | 0.15 of the move | 0.35 | 1.4 | 0.55 | 0.15 |
-
-The flavours are distinct by design, not tuning: crisp leaves at full speed
-and never passes its mark; smooth takes twice as long and eases in and out; bouncy clearly passes its mark and settles back once or twice;
-playful winds up the other way first, flies well past, and stretches along
-its path and squashes as it lands, its text with it (the user's choice); a
-pointer inverts the same stretched box, and landed, a pane is exactly its
-target. A motion only a few pixels big (Lift) is
-exaggerated by the flavour's gain so its character shows. The flavours drive
-the Lift spring (time-based), the notice drop and the pane moves (PanelSlide,
-PanelZoom, PanelVertical opening or moving, whose length scales with the
-flavour); a closing pane keeps its effect's own exit. Chart and side-by-side
-video: .scratch/render/motion/compare/.
-
-Which principles each motion uses:
-
-- **Lift** (G1): timing, slow in / slow out (the spring), follow-through
- (bouncy and playful lift past full and settle back), anticipation (playful
- dips before it rises; below zero nothing is drawn, so it reads as a beat
- before the lift), staging (only the focused pane lifts; nothing else moves
- with a focus change), exaggeration (the flavour's gain). Squash, stretch,
- arcs and secondary lag have nothing to act on in a lift.
-- **Notice drop and pane moves**: timing and follow-through from the flavour's
- spring over the move's length; anticipation (playful). A pane past its mark
- never takes a neighbour's click, a rising one stays in its box, and a notice
- past its row stays in its pane's body.
-- **Cursor** (G3): timing (the flavour's spring, 2.4x its pace: it follows
- input), follow-through (bouncy and playful pass the target), squash and
- stretch and secondary action (the trailing corners lag by the flavour's
- follow, so a jump smears and collapses), arcs (a jump over 8 cells bows by
- the flavour's arc), anticipation (playful's wind-up). A step of a cell or
- less, or any move in insert mode, lands at once, so typing never trails;
- a scroll carries the cursor with the text. Only the focused cursor moves.
- A pane that moves carries it too: the springs keep its place in the pane
- and the quad is drawn through the pane's presented box, so through a
- slide, a zoom or a drag it rides with the text, and a layout change with
- no transition lands it with the text. A focus change (to another pane, or
- into a column's or the workspace's tag) glides from where the cursor is
- drawn to the new one, free of the old pane's edges while it crosses
- (.scratch/render/cursor/move/: ride-*, cross-*).
-
-
-Notes on G1: the lift and the dim run on one focus spring per pane, at the
-Motion flavour's pace, sampled at each frame's own time; while it moves the
-GUI draws at the display's rate, and a grid snaps. A notice floats on a lift
-of 1 while a shadow or rim is on. InactiveDim under `Lift auto` defaults to
-30. The fade is in linear light, so on a dark page it is gentle: at 30
-forge's text goes from 15.5:1 to 11.2:1 and dusk's from 6.4:1 to 4.8:1,
-plainly quieter and still easy reading; at 10 forge barely moves (14.1:1).
-On a light page the same percent bites far harder (acme: 7.0:1 at 10, the
-4.5 floor at 30), which is why `auto` shades there instead of dimming.
-
-Notes on G4: the three are post passes, so each is a frame of cost only
-when it is in the chain. None moves (grain is still, by choice: moving
-grain would redraw every frame for a texture), so none asks for frames.
-The focus rule is kept by construction: the shell hands every pass the
-rectangles of the tags (with their grips), the column and workspace tags,
-the notices and the cursor, and a bundled pass leaves them as they are; its
-strength is capped per theme so the text on the page and the selection's
-text keep min(their contrast, 4.5) (Post.capsOf, tested over the native
-themes). Bloom takes only what is brighter than the page, so a light page
-does not bloom at all; its glow is a dual Kawase blur in float textures at
-half size and down (3 to 5 steps by level), within §8.2's 5%. The vignette
-is round, starts halfway to a corner and reaches 10, 18 and 28%. Grain is
-1.5, 3 and 5 of 255 on page-coloured pixels only, a logical pixel big.
-Found on the way: SDL GPU binds at most 4 KiB of a uniform block, so every
-Shadertoy colour after the palette (iBackgroundColor and on) read zero; the
-block is now cut in three with ghostty's names and order kept.
-
-Notes on G5: lapis (src/themes/lapis.zig) is 0x4200.cafe's lapis.css in
-pardes's chrome, drawn inside it (decision 1). The tags are the site's
-trail and title bars: every tag a plaque in its own band, framed in gold
-(2 px, the trail's border) with the trail's hard vermilion shadow (5 px)
-down and right, the frame and the shadow both inside the band's rows and
-columns, so no shadow reaches a body, a grip or the next pane; down, the
-shadow gets what the band's slack under the text leaves (a pixel or three
-at a tagline under the body's size, more at a larger one), to the right
-its full offset. The focused pane's plaque is the code window's title bar,
-gold and lapis stripes of 2 px, its words on plates of the focused ground
-(the words' cells stay opaque, the blanks between let the stripes show),
-so the focus indicator is plainer to see than any theme's tint, never
-less; the unfocused ones are plain. Column and workspace tags take the
-quieter object-label language: a 1 px frame and a 2 px shadow of vermilion
-toward the page. The plaques are decor drawn under their cells (a group's
-under range) with the cells blended premultiplied over them; blank tag
-cells on the tag's ground are see-through there. They need an opaque
-window: under WindowOpacity 100 the tags are plain. Also from the site:
-the colours (lapis-deep page, vellum text, gold names and rules,
-vermilion rubrics and tag rule, gold selection in lapis ink), the page's
-gold dot grid (22 px, #e8c46a22) on page-coloured cells only, fixed to the
-window; the dither frame's checker in the scroll track, anchored to its
-rail (the thumb solid gold); the titles' vermilion offset shadow under a
-file name's letters. Left out: Crimson Pro and Departure Mono, the
-ornaments, a decor crossfade (it snaps while the colours fade). The tty
-shows the colours alone: a tag row abuts its body and its neighbour, so
-there is no cell for a shadow row or column.
-
-Notes on G6: a mouse wheel's notch (SDL's whole-number step) scrolls the
-core by its lines at once, so keys and clicks act where the text is, and
-the picture follows on the Motion flavour's spring at half again its pace,
-critically damped whatever the flavour (text never passes its mark; §8.1),
-from its current speed, so notches in a row carry on without a stall. How
-far the rows moved is measured on the next frame against the last (a
-wrapped line's rows, a terminal's output), within a couple of rows' slack
-for the cursor's line and its number moving with the scroll; the rows that
-left are kept and drawn in the gap the lag opens, and a lag is never more
-than the body's height (a flick past that jumps). A touchpad's fractions
-keep moving the picture under the finger, as before; there is no momentum
-(SDL reports no finger lift on a wheel). The glide runs on the core's
-continuous path (Pardes.shell_continuous), so the shell draws every loop
-while it moves and a virtual clock moves with it; idle, it costs nothing.
-A PDF page scrolls as before. The test feed takes wheel events
-(ESC]777;mouse;wheel;<amount>;<x>;<y>BEL).
-
-Notes on G7: the geometry already followed the Motion flavour's spring
-(sxrrwplk); a move retargeted mid-way now carries its speed on
-(Track.launch, from the old move's speed along the new one's longest
-side), so a second Newcol during a slide bends the motion instead of
-stopping it dead. A pane that slides, zooms or rises stands off the page
-while it moves: a soft shadow as Lift's, at sin(pi x progress) of its
-strength (none as it sets out or lands), on the other panes' bodies and
-rails alone and under the same contrast ceiling, moving with the pane. The
-GUI's PanelDissolve is a crossfade: a changed cell's new form drawn whole,
-its old one over it at 1 - progress; the blend is in sRGB, not linear
-light (a two-colour mix in linear light needs both in one shader; open
-point 3). The tty keeps its per-cell swap. GUI golden 18-mid-transition
-changes (the slide's shadow at frame 6).
-
-Notes on G8: most of it was there: the drop eases out on the Motion
-flavour's spring (sxrrwplk), the dissolve eases in, and a notice floats in
-tier 4 with Lift's shadow at its fade while Lift is on, on the bodies
-alone. What G8 changed: leaving was slower than arriving (a 300 ms
-dissolve against a 180 ms fall); §8.1 wants leaving the quicker, in 100 to
-160 ms, so MessageDissolve now defaults to 150 (a test holds both bands).
-
-Notes on G9: the halo is Lift's soft kind in the selection's own colour,
-cast by each block of selected rows in a pane body (the rows that touch),
-with the block as its caster, so the selection itself is never touched;
-a cursor beside a block joins the caster, so it is never glowed over
-either. It stays on bodies (never a tag or a grip), is at most 22% (§8.2:
-coloured glows at most 25%) and less where the text round it would drop
-below min(its contrast, 4.5); it comes up over 100 ms with an ease-out
-(micro-feedback), on the core's continuous path, and is not drawn while
-panes move. The blocks are read from the cells painted in the theme's
-selection colour, a pixel shell's inference (§1.5, smell 4) until
-selections are regions.
-
-Notes on G10: a theme change already faded the chrome (tags, rules, grips,
-rails) over 10 frames (160 ms, in §8.1's 150 to 250 for a state change),
-mixing sRGB bytes linearly in time. Now each colour goes through OKLab, so
-a fade between two hues keeps its lightness instead of dipping dark and
-muddy, on a smooth ease-in-out; a change made mid-fade sets off from where
-it shows and eases out, so it moves at once instead of stalling. The page,
-its text and the syntax colours still change at once with the theme, as
-before: fading them would re-resolve every cell each frame.
-
-Notes on G11: the word a look-hover would open is already lit (the
-affordance's tint); HoverGlow adds a strip three logical pixels tall along
-its foot in the theme's accent, its edges blurred by one, at most 22% and
-under the text's contrast floor, fading in over 80 ms (micro-feedback). It
-stays on bodies and is split round a cursor, so it never touches one. The
-word is found by the tint's colour, as G9 finds the selection.
-Found while reviewing: with the machine loaded, test/gui_golden.py's
-two-captures-agree settle can take a scene's frame before the scene's
-change lands, and every later hash moves up a scene; a rerun on a quiet
-machine matches.
-
-Notes on G12: four soft strips cast from just outside each body inward,
-clipped to it and round a cursor, black in linear light, 2% at most and
-under the lift's contrast ceiling, in the pane's own group so it moves
-with it. At §9.1's strength it is all but invisible on a dark page (a
-level of 255 on forge) and faint on a light one (acme's page steps one
-level toward the edges); acme-diff-x40.png shows where it falls.
-
-Notes on G13: a body region now carries its first line (Region.line: a
-file's scroll, a terminal's grid offset; on the wire, a u32 more a
-region). Under Parallax the dots under each body are offset by a quarter
-of its line times the row height, plus G6's glide offset while one runs,
-so they drift with the text as it glides and settle with it (measured: 4
-lines at 35 px move the 22 px grid 13 px, 35 mod 22). The cell shader
-reads the offset from the instance's serial slot, free since G7 took the
-dissolve's per-cell noise out. Without a pattern, nothing happens.
-
-Notes on G14: removed with its builtin, setting and code after the user
-saw it in the lapis showcase: "one of the animations is like a flashing on
-the whole pane or something, that is really bad, just remove it". Every
-changed cell glowing meant a pane whose whole text changed flashed whole.
-
-## Terminal track (§9.2)
-
-Audit, after decision 2 (keep every whimsical effect and polish it, never
-remove it): the six core-composed character transitions (ascii, edges,
-fall, wave, scramble, typewriter) and curtain stay as they are, each
-bounded to its pane's final rectangle and each cell's last sample its
-canonical one; slide and vertical are the tty compositor's and follow the
-Motion flavour (sxrrwplk); zoom stays (nearest-cell scaling, its look is
-the effect); dissolve swaps each changed cell once at its threshold;
-message fades and the theme's chrome fade are the core's (the latter in
-OKLab since G10), and InactiveDim (T5's unfocused dim) works in a grid as
-it does in a pixel shell.
-
-| # | effect | switch | state | review material | verdict |
-|---|---|---|---|---|---|
-| T2 | cursor jump trail | `JumpTrail on\|off` (off) | a jump of three cells or more leaves 3 to 5 cells of trail behind the cursor, their ground drawn toward the text's colour in OKLab, most beside it, fading in 120 ms; truecolor terminals (vaxis's answer or COLORTERM); never the cursor's cell, a selection or outside the focused body; off under Motion off; a few cells a frame | .scratch/render/tty/trail/ (pyte replay of a real tui session at 60 fps: trail-forge.mp4, zooms) | pending |
-| T1 chip shadow | `ChipShadow on\|off` (off) | a notice chip casts a shadow a cell right and down in its body: half blocks (▌ ▀ ▘) in a darker ground on blank cells, the ground 35% darker (OKLab) under text; never the cursor, a selection, a tag or another chip; truecolor only | .scratch/render/tty/shadow/ (forge, acme; zooms) | pending |
-| T4 ordered swap | `PanelDissolve` (off) | the terminal's dissolve swaps each changed cell once at its place in a 4x4 Bayer matrix (shifted per pane), an even patterned screen instead of noise; hit-testing follows the same order | .scratch/render/tty/dissolve/ (ordered-swap-sim: the matrix applied to two real frames; see note) | pending |
-| T7 thumb flash | `ThumbFlash on\|off` (off) | a pane whose view moves has its scroll thumb drawn toward the text's colour, 60% at once and back over 250 ms (quadratic); the thumb only, never a focus indicator; truecolor only; a few cells a frame while it fades | .scratch/render/tty/thumb-flash.mp4, thumb-flash-strip.png (pyte replay, forge) | pending |
-| T3 wipe | `PanelCurtain` (off) | the curtain becomes a wipe, left to right: new grid behind the head, old ahead, and a soft edge of two columns whose ink and ground sit 2/3 and 1/3 of the way to the old ground (OKLab; a palette colour steps at the half); only the head's columns change a frame; the head's cells are not hit-testable until settled | .scratch/render/tty/wipe/ (curtain-gui.mp4, curtain-edge-zoom.png) | pending |
-
-Note on T4's material: a pane opening in a tui session recorded through
-pyte did not animate (PanelSlide did not either there), so the review
-material applies the same matrix to two real frames of that session; the
-unit test holds the order (a 4x4 block reveals exactly t x 16 cells).
-
-Not built: T5 (unfocused dim) is InactiveDim, which already works in a
-grid. T6 (a truecolor ramp on the active tag band) would change the
-focused tag's colour, a focus indicator no effect may alter, so it stays
-out. The §9.2 byte budget (deferring a sample after an oversized frame,
-DECRQM for sync 2026) is not built either: the terminal effects here change
-a few cells a frame, and only PanelSlide rewrites a whole pane per sample.
diff --git a/docs/next-anim.md b/docs/next-anim.md
deleted file mode 100644
index 3d95dfe6..00000000
--- a/docs/next-anim.md
+++ /dev/null
@@ -1,1107 +0,0 @@
-# anim: the game animation pass, as built
-
-The author's words were the spec: scrolling whose timing is consistent so
-it never feels choppy; every scroll animated, a long one too; smooth
-while keeping the responsiveness; the thin insert-mode caret animated like
-the block; small motions that read as smooth; the game's habits for
-implementation and testing, frame timings measured as a game measures
-them. Later rounds: input latency is a hard constraint; p90/p99 against
-the mean is the smoothness measure; throughput on a big file; every file
-kind, pane kind and way a view moves, the tty, long lines walked along,
-picture and PDF first frames; ponytail for all of it; recordings.
-
-Positions, not a verdict: argue with them.
-
-## The change stack (workspace `pardes-n-anim`, above converge `oqolqlko`)
-
-| change | what |
-|---|---|
-| anim 1 `vtmwqzpt` | `src/FrameStats.zig`, the frame probe in `Pardes.frames`, stamped by `pump`; the Debug box's timing lines; `PARDES_FRAMES` dump at exit |
-| anim 1b `sxxnrzqz` | input-to-present latency, work zones, the view shown, the tty's bytes; the timedemos (`PARDES_FRAMES_DEMO=scroll\|stress\|jumps\|kinds\|panes\|longlines\|one\|clip-*`); `PARDES_FRAMES_CAPTURE` (every presented frame's pixels, for the samples) |
-| anim 2 `yqwqutws` | the GUI's paced frame loop: a tick grid, each frame sampled at its tick, an input's frame at once |
-| anim 3 `wkxvyoqy` | every scroll glides, whatever moved the view |
-| anim 4 `qztrptlo` | the thin caret glides; a step eases; the caret rides a scroll's glide |
-| anim 5 `ntnlpots` | colours painted ahead on a worker (`Stream.paintAhead`) |
-| anim 6 `xykrnoql` | the tty paced too, its acks and output waited through |
-| anim 7 `powlyrxv` | what the probe found: the spin, glide measure and emission, far-jump and long-line paints, the caret's sideways scroll, no join on the loop, pictures decoded off the loop |
-| anim 8 | this note |
-| anim 9 `nqznmskw` | the view's motion is one system: a stream of inputs moves it at its own rate, the spring only settles it |
-| anim 10 `umnwnrvn` | the caret is its own body: drawn from its own springs, never carried by the text's glide |
-| anim 11 `knzzzknu` | the integrator is `animation.Follow`; a PDF's strip moves in it, in pixels; a slow stream eases onto its mark |
-| anim 12 `nqmoxkwz` | the tty runs the same glide (`src/glide.zig`), drawn at whole rows |
-
-Tests (one check each, `-Dtest-filter`): "frame stats: gaps into a motion
-count as jank ...", "every scroll glides on one law ...", "a cursor eases
-a cell at a time, keeps up with typing, and glides a jump ...", "the
-focused cursor glides a jump, eases a step, the thin caret too ...",
-"Stream: a scroll's colours are painted ahead on a worker ...", "a long
-line is painted only around the caret", "a caret deep in a long line
-scrolls the view sideways to it, measured from the view's edge". Builds:
-Debug and ReleaseSafe, gui and tty. Failing: only the two baseline
-failures NEXT.md names. Not re-baselined: the GUI golden `16-debug` (the
-Debug box is three rows taller; it shows `frame virtual clock` under the
-golden's clock so it stays still).
-
-## Method
-
-The probe (anim 1, 1b): a fixed ring of 16384 frames in `Pardes` (1.3 MB,
-nothing allocated per frame, three clock reads a frame, always on). Per
-drawn frame: update, render, present µs; the gap since the last present;
-whether something was moving into it (a gap over 1.5 periods then is
-jank); the input-to-present latency of the first input it answers (SDL's
-event timestamp, moved onto the monotonic clock; the tty stamps a key as
-its loop takes it); the view and the picture's offset (throughput is rows
-actually presented); zones (the core's body rows and colours; the GUI's
-instance build, shaping and rasterizing in it; a soft present's GPU wait,
-upload and flip); the tty's bytes. `Debug` shows the last second.
-
-The timedemos are a game's: fixed scripts fed through the shell's own
-input paths, named phases, the session ending itself. The tty is driven
-through a pty I own (`.scratch/tools/ttydemo.py`, typing the same
-scenarios as bytes, answering the frame acks). The corpus
-(`.scratch/tools/corpus.sh`) links 13 big real files from the genizah
-(a generated C parser, helix's commands.rs, a Python table, jquery, a
-minified bundle, a grammar JSON, CHANGELOG.md, an HTML spec page, the
-simdutf amalgamation, a TypeScript parser, a long shell test, a Go table,
-guide.typ) and generates five (a 600 KB one-line JSON, CJK/emoji/
-combining/RTL/tab text, a 32 MB log, 400k short lines, six 360 KB lines).
-The stress file is layout.zig x23 (38k lines, 1.6 MB).
-
-**Every before/after below is a pair run back to back** (`batch3.sh`:
-base then after per demo), the before binary being main+probe (anim 1/1b
-alone, `.scratch/base-bin`). The first batch, before and after an hour
-apart, was useless: this i7-11700 under intel_pstate moves its clock 2-5x
-with load, which moved every number more than any change did.
-
-## What the probe found, and what came of it
-
-1. **Only a wheel notch animated.** Ctrl-d/u/f/b, PageDown, G, ge, gg, n,
- a Look: zero animated frames; a wheel flick past a body less three rows
- snapped. Now every move glides (below).
-2. **The loop ran uncapped while anything moved**, ~230 frames a second
- on a 60 Hz display (the core asks for a 1 ms wait; the Wayland soft
- present has no vsync), so the compositor showed whichever frame was
- newest at each refresh: motion sampled at a random phase, the choppy
- the author felt. Moving gaps before: p50 4.3 ms, stdev 2.2. Now one
- frame a tick, sampled at the tick: p50 16.6, p99 17.2-18.0 ms.
-3. **Pacing asleep made every frame 2-5x slower** (the same body-rows
- work 3 -> 16 µs, an input's first frame 2.6 -> 7.5 ms p50 RS): a core
- that sleeps 80% of each 16 ms is clocked down. Fixed by spinning while
- something moves, as a game's frame limiter does (A/B below).
-4. **Tree-sitter parsing on the frame** (perf: 36% of the Debug stress
- run, most of it the DebugAllocator capturing a stack trace for every
- tree-sitter allocation): a scroll past the painted stretch parsed the
- next one on the frame, 6-58 ms in Debug, a hitch every two pages. Now
- painted ahead on a worker; Debug held PageDown render p90 24 -> 1 ms.
-5. **A key typed in the middle of a 500 KB line repainted the line**:
- 172-213 ms a key in ReleaseSafe (minified JS, one-line JSON). Now 16 KB
- round the caret.
-6. **The caret's sideways scroll walked a long line from its start**
- (no-wrap, `ensureCursorVisible`, `rawAtDisplay` measures the whole
- line): now measured from the view's edge (`lineDisplayOffset`, which
- already existed for selections).
-7. **A picture decoded on its first frame**: 80-94 ms RS, 285-296 ms
- Debug for the docs' screenshot. Now on a thread; the pane appears at
- once and the picture lands a frame or so after the decode.
-8. **The ReleaseSafe tail is the soft present**: a GPU fence wait (p50 0.6
- ms, p99 6.4) and a 6.7 MB upload (p50 0.9-1.7, p99 4.8) every frame.
- With the GPU swapchain (X11 here) the scroll is 0.34 ms of work p50.
- Not changed: see the positions.
-9. Smaller, fixed: the glide's shift search on a page's first frame (6 ms
- Debug: the core's line count is tried first), its emission of every
- body row however far it lagged (twice the instances), a far jump
- painting three views on the frame, a paint asked at every keystroke, a
- paint of a 600 KB line joined on the loop (half a second), the tty's
- loop running at the terminal's round trip (its acks woke it: one frame
- every 0.4-2 ms while anything moved).
-
-## The design as built
-
-### Pacing (anim 2, 6, 7)
-
-While anything moves, the GUI draws one frame a display refresh on a grid
-of ticks (`waitInput`, `Gui.pace_ns`), and the core is advanced to the
-tick (`guiNow` -> `Shell.paced_ns`), not to whenever the loop woke, so each
-frame moves the motion by exactly a period. A late frame skips the ticks
-it missed. An input ends the wait at once and its frame is drawn then;
-the grid restarts from it. Idle, the loop sleeps as before. Patterns: a
-fixed presentation cadence over variable-step evaluation (the springs are
-closed form, evaluated at the tick; the core's stepped animations keep
-their 16 ms steps), a frame limiter, input acted on at once and never
-gated (only the picture is paced).
-
-The tick is waited out polling, not asleep (anim 7). Same binary, same
-script, interleaved, ReleaseSafe, the scroll demo (`.scratch/runs/ab*`):
-
-| loop | input-to-present p50/p90 ms (notches) | moving gap p50/p90/p99 ms | CPU s per 18 s run |
-|---|---|---|---|
-| before: uncapped | 3.2/15.2 | 4.4/9.3/10.1 | 1.9 |
-| paced, asleep | 6.8-8.2/11-14 | 16.4/19.0/23.6 | 2.9-3.1 |
-| paced, spinning | 1.8-1.9/7.7-15 | 16.6/16.8/17.4 | 7.0-7.3 |
-
-The tty is paced asleep (a terminal has no vsync and no sub-cell motion,
-and power matters more there), and its wait now runs through the frame
-acks and pty output to the tick (`waitOnce`): its moving frames went from
-every 0.4-2 ms (each a write) to one a 16 ms tick.
-
-### Every scroll glides on one law (anim 3, 7)
-
-`stepGlide` keeps each pane's view as the last loop saw it (`Glide.seen`)
-and finds the one that moved, whatever moved it. The body as the last
-frame drew it is still in `core.surface` then, so it is copied there (no
-per-frame copy). The next frame measures the rows moved (the core's line
-count tried first; the shift search only when wrapped rows disagree),
-keeps the rows that left for the gap, and kicks the spring.
-
-- The law: the Motion flavour's critically damped spring (the one the
- author liked on a notch) at 1.5x its pace; a move during a glide adds
- its rows and keeps the speed: retarget, never queue. A notch settles in
- 12 frames, a half page 15, a page or a 30 000-line jump 16.
-- Past what two frames share, the last frame's body is laid just past the
- new one and slides one body (`.scratch/ge-montage.png`: line 55 runs
- into 9448). The lag may reach two bodies, so a jump mid-glide carries on.
-- Readable: at most a body in four frames (`Glide.sample`).
-- Not gliding: a new pane (appears at once), a pane moved or resized in
- the same step, a terminal's output, a touchpad (follows the finger),
- generated text rewritten as its view moved, Motion off (all land; the
- test checks).
-- The caret and the glide quad ride the picture (`glideLift`).
-
-### The view's motion is one system (anim 9)
-
-The author watched sample 3 and saw a bounce a line under a held j: every
-repeat restarted the spring's ease to its new target, so the picture sped
-up and slowed down once an input. Now the picture's motion is one system
-with its own state (`Glide`: its lag behind the text, its speed, and the
-inputs that have been feeding it), integrated in one place every frame
-(`Glide.advance`), and every scroll is an input into it (`Glide.kick`),
-absorbed by what the system is doing:
-
-| the system is | an input comes | it does |
-|---|---|---|
-| at rest | any move (a j, a notch, a page, a jump) | the spring starts an ease from rest |
-| easing | the same way, within 150 ms | the inputs are a stream now: it moves at their rate (rows an input over their spacing), a lag of an input or so behind, never easing to each |
-| fed by a stream | another of the stream | the target moves on; the speed stays the stream's |
-| fed by a stream | the inputs stop (none for 1.5 spacings) | the spring settles it, from the speed it has, never past its mark: the only ease-out |
-| moving | an input the other way | the stream ends; the spring turns it through zero from its speed |
-| moving | a jump past what two frames share | its rows are added (the stitched slide); the speed carries on |
-
-The caret is its own body (anim 10): its state is its own springs
-(`CursorGlide`, screen cells as floats), integrated on the same tick, and
-it is drawn from that state only, never from the text's glide (the ride
-that lifted it with the glide's fractional lag is deleted: every motion
-of the text had leaked into it). Its only coupling to the view is its
-target: its cell in the view as it will be once settled (the core's view
-is that one).
-
-| an input | the caret's target | it does |
-|---|---|---|
-| a page or half page, a scroll that keeps its screen row (Ctrl-f/b/d/u, PageDown/Up, a j or k at the edge row) | unchanged | not a pixel's move while the text slides under it |
-| a move in a still view (h/j/k/l, w, a click, a typed character) | the new cell | the short ease |
-| a stream of such moves (a held key, typing) | each new cell | lands on each, at the keys' rate |
-| a scroll that carries its line (the wheel, Ctrl-o to a far place in view) | where its line will be | moves there whole, at the view's pace, and lands exactly |
-| a jump that moves both (G, a Look) | the final cell | one ease, no trip with the old page |
-| a terminal's output | its cell | lands with the output (which does not glide) |
-| Motion off | the cell | there at once |
-
-While a glide moves the text the caret is drawn as its quad (from its
-springs), not as a cell of the moving rows. Invariants, checked over every
-frame of the scripted scroll and held-key runs (`GlideRun.checkCaret`):
-while its cell stays and it is on it, its drawn row is identical frame to
-frame; it only closes on its cell; it lands within half a second of its
-last change. The blink keys off the last key or click, never a frame.
-
-So a held j, a spun wheel and a held Ctrl-d are the same case (a stream),
-a single j or notch the same ease as before. Every input, a repeat too,
-is still drawn at once: in a stream the picture is on the stream's
-straight line wherever a frame samples it (tried: holding repeats to the
-next tick cost a held key 7 ms of latency for no evenness one could see).
-
-### A PDF's strip, the same system (anim 11)
-
-The integrator is `animation.Follow` now (the Glide's state and rules,
-moved out; the text's Glide holds one, in rows). A PDF pane's strip holds
-another, in pixels: the core's position is `document_scroll_y` as every
-input leaves it, and the strip is drawn `follow.offset` from it
-(`pdf_view.stripAt`, each frame at the core's clock). Whatever moved the
-core's position since the last frame kicks it, so every way a PDF moves is
-one input to the same rules, the text's table above unchanged:
-
-| input | the strip |
-|---|---|
-| a wheel notch (a line's pixels), a spin | eases; a spin is a stream, at its rate |
-| PageDown/Up, Ctrl-d/u/f/b, held | a stream, at its rate (to the 15 bodies a second cap) |
-| ge, gg, a Look at `file.pdf:12`, Back, a section in the outline | a jump: past a body, one body's slide, as a text's far jump |
-| a click on the rail, a tick | eases |
-| a touchpad's fraction (`pdf_scroll.glide` false) | follows, drawn where the core is the same frame |
-| a resize, a fit or zoom change (a relayout) | there at once |
-| a search's reveal (it moves the strip as it draws) | there at once |
-| Motion off | there at once |
-
-Rendering stays on the loop (MuPDF; its first-frame and context work is
-the open item it was). So that no frame of a glide waits on it, the pages
-a glide crosses to its mark are rendered in its input's frame, where the
-instant move rendered its destination; a jump past a body (`far`) renders
-only where it comes to rest, and its slide comes in over what is there
-(the page above the mark is not drawn for a slide of a tenth of a second:
-that page was the input frame's whole extra cost). Drawn at the strip's
-position, rounded only by the draw (`pixel_offset_y` keeps the
-fraction). Pictures are fit to their body and neither scroll nor pan:
-nothing there to move.
-
-**Why the slow PDF scroll read as jumps**: a wheel notch moved the strip a
-whole line's pixels (25 here) in the one frame after it, and nothing in
-between: at a slow turn, one 25 px jump a notch. The touchpad path was
-never quantized (fractions all the way to the placement). Now a notch is
-a kick, and the strip eases over ~6 frames.
-
-### The tty, the same system (anim 12)
-
-The glide is shared now (`src/glide.zig`: the Glide, and its step that
-finds the view that moved, snapshots the body and measures the move); the
-GUI keeps only its wheel notches and its pixel draw. The tty steps the same
-glide each loop (`pollFrame`), integrates it on its 16 ms ticks, and draws
-it at whole rows (`panel_compositor.glideRows`): the body's rows moved by the offset
-rounded, the position never the step, the rows that left in the gap. So a
-flick's fast part moves several rows a tick as the body dictates and its
-tail is one-row steps further and further apart: a deceleration in the
-timing, not in the step. The end is the instant version's. The caret: one
-whose line rode the scroll (a wheel's) is carried with its rows; one that
-kept its row (a page, a held j at the bottom) stays on its cell, still.
-
-Checks: "the tty draws a glide at whole rows" (a flick and a spin: once
-slowing, steps a tick never grow, one-row steps never closer together, to
-where the core is; then j held on the bottom row: the caret's row is the
-same every tick). The timedemos typed into a pty (160x50, `ttydemo.py`,
-RS): past each glide's fastest tick, steps never grew in the decelerating
-part of any of 24 (scroll) and 186 (stress) glides, and the ticks between
-one-row steps never shrank; the only growth is at a stream's constant
-speed, where the rounding alternates (5, 4, 5, 4 rows a tick at the 4.5 a
-tick cap). Bytes: a tick costs a body repaint (p50 7.4-7.9 KB, max
-13.4-15.3 KB, the before's max), so a glide costs its row steps' worth:
-the scroll demo wrote 383 -> 990 KB, the stress demo 3.8 -> 7.1 MB. Terminal
-scroll regions would make a one-row step a few bytes (position 8).
-
-| tty | | latency ms p50/p90/p99/max | frame work ms | moving gap ms | bytes/frame p50/max |
-|---|---|---|---|---|---|
-| RS scroll | before | 2.3/3.0/11.5/11.5 | 2.2/2.7/8.5/11.0 | 15.9/16.7/18.8/18.8 | 120/15264 |
-| | after | 2.5/3.1/11.5/11.5 | 2.3/2.8/4.4/11.9 | 15.7/16.7/18.4/18.7 | 0/15264 |
-| RS stress | before | 1.4/2.9/5.9/15.6 | 1.1/2.7/4.9/14.1 | 15.6/16.6/18.4/18.4 | 140/13383 |
-| | after | 1.5/2.8/5.5/14.0 | 1.9/2.9/3.8/12.5 | 14.3/16.5/18.1/19.0 | 7363/13383 |
-| Debug scroll | before | 9.4/14.5/28.7/28.7 | 6.0/13.0/21.2/36.2 | 15.5/19.0/21.7/21.7 | 128/15264 |
-| | after | 6.0/13.8/44.0/44.0 | 6.1/9.9/17.0/27.3 | 15.5/19.6/24.9/26.0 | 0/15264 |
-| Debug stress | before | 2.4/13.0/92.9/148.4 | 2.2/10.8/68.2/148.1 | 15.6/18.7/25.4/25.4 | 6858/13383 |
-| | after | 2.8/10.3/90.5/180.3 | 2.7/7.7/56.7/180.2 | 12.8/17.6/65.0/180.4 | 7270/13383 |
-
-Found on the way (sample 1): at a spin's end the measure found no shift
-under which the rows agreed (the colours painted ahead arrive as it
-stops) and gave the glide up, a 1.6-row snap on the last frame, in the
-merge head as in anim 9-11. When the core moved a few lines and the rows
-only disagree, the lines are the rows' guess now: it glides them, and the
-spin's settle is whole (33, 4, 1.8 ... 0.03 px, never growing). Re-measured
-on the head (stress, RS): the caret under a held j moves 0.0 rows frame to
-frame over the 144 frames the view moves; every settle's steps never grow
-after its fastest frame, the last at most 0.05 px. The snapshot harness's
-virtual clock moves a frame a wake while the glide moves (it only moved to
-the next due wait, so a tty glide stood still there): all 110 scripts pass,
-no golden changed.
-
-(The Debug maxima are single frames of highlight paint on the 1.6 MB
-file, as before; ponytail: a terminal's scrollback under the tty's wheel
-lands, it reaches the core as a mouse event the GUI's notch count does not
-see.)
-
-### Coming to rest (anim 9, 11)
-
-Found while measuring the PDF: a slow stream (notches 150 ms apart, just
-inside the stream window) ran its constant rate into the mark, stopping
-from 2.6 px a frame, and its last speed, never zeroed, then carried it
-0.04 row past the mark and back. A stream is now never faster than the
-spring's own pace from where it is (omega a unit of lag a second): it eases
-onto each mark. The check: "a follow comes to rest without a snap" (a
-spin, a page, and notches 144 ms apart: toward the mark only, onto it from
-under a third of a pixel, past its fastest frame each step no longer than
-the last); it fails on the old rule.
-
-Measured on every settle the runs have (frames after an input until the
-view is still, speeds normalized by the frame's own time): before anim 9
-the old spring's tails were already monotonic; the anim 9+10 stream rule
-had one rising tail in the scroll demo's notches (the stop from 2.6 px)
-and a last step of 0.88 px; now, over the scroll, stress and PDF runs
-(ReleaseSafe and Debug), no tail rises after its fastest frame and the
-last step before rest is at most 0.06 px (text) and 0.05 px (PDF).
-
-### The cursor (anim 4)
-
-The thin caret glides as the block does. A step of a cell or less eases
-instead of landing: its own spring 1.5x stiffer, the four corners
-together, never overshooting, ~60 ms; at a key's repeat it trails half a
-cell and never smears a caret into a block. A jump still smears and bows.
-
-### Colours painted ahead (anim 5, 7)
-
-`Stream.paintAhead` asks the disk engine's worker (its own generation
-bit) to paint the stretch two views ahead the way the view last went,
-with its own parser and cursor (`syntax.highlightApart`; markdown and
-typst, whose injections share parsers, stay on the loop).
-`refreshHighlights` takes it on arrival (`takeAhead`). A paint still
-reaching into what is asked is let finish; one that does not is told to
-stop and left to end on its own, never joined on the loop. Not asked
-while the view stands still, nor for stretches over 512 KB.
-
-## Per scenario: glides, snaps or follows
-
-| scenario | | why |
-|---|---|---|
-| wheel notch, a spin | glides, each notch retargeting | discrete steps, piled onto one glide |
-| touchpad | follows | direct manipulation: smoothing is lag |
-| Ctrl-d/u/f/b, PageDown/Up, held | glide, retargeting | discrete jumps |
-| G, 12G, 300G, 9000G, gg, ge | glide; past a body, one body's slide | a jump shorter than a body glides its rows; longer ones on the same law |
-| n/N, `*`, a query's Ctrl-n/p moving its owner | glide (the owner's view) | the eye follows where the place went |
-| Look `:40`, `:7000`, `file:line` in the same pane | glide | same |
-| a Look opening a new pane | appears at once | nothing to glide from |
-| jumplist Ctrl-o/Ctrl-i, SPC j o/i | glide | jumps |
-| rail B1/B2/B3, a tick | glide | pardes's rail is acme's: a click scrolls by the row clicked; there is no thumb to drag, so nothing to follow (a thumb drag, if added, must be skipped in `stepGlide` as the touchpad is) |
-| wheel over the gutter | glides | the wheel's |
-| typing, arrows, w/e/l on long lines | the caret eases | small steps |
-| Enter at the bottom of the view | a one-row glide each | the text rolls up |
-| a 200-line paste | one body's slide | a long jump |
-| terminal output | snaps | a stream would always lag |
-| terminal under the wheel | glides | the wheel's; keys in its normal mode snap (position 4) |
-| PDF | glides (its strip, in pixels, anim 11); a touchpad follows | the same rules as text |
-| no-wrap sideways scroll | snaps | horizontal glides were not asked; position 6 |
-
-## Ponytail pass (on the stack, amended in place)
-
-Deleted: the cursor-quad columns of the probe (unused by any analysis:
-about 5 lines in anim 1 and 3 in 1b), the `PARDES_PACE_SLEEP` toggle and
-the sleeping branch it kept (about 12 in anim 7; the A/B above is its
-record); nothing else in the stack had a cheaper rung.
-Shortcuts marked `ponytail:` with their ceilings: the ring's size, the
-tick grid's phase (ours, not the display's), one glide at a time, one
-paint at a time, the spin's cost, a thread a picture, the 16 KB long-line
-paint. Measured behaviour unchanged by the pass (the toggle's default was
-what was measured).
-
-## Numbers
-
-Every row is a pair run back to back (`batch3.sh`), the final binaries,
-ms p50/p90/p99/max; *latency* is input-to-present of the frames that answer
-an input; *frame work* is update + render + present of every drawn frame;
-*present gap* only over frames something was moving into; *late* counts
-moving gaps over 25 ms. The soft present (Wayland) in both. Small phases
-(under ten inputs) move a couple of ms round to round in both directions;
-the long phases are the ones to read.
-
-### Headline: every phase of a demo pooled, before -> after
-
-| build | demo | latency | frame work | present gap (moving) |
-|---|---|---|---|---|
-| rs | scroll | 4.2/8.2/11.4/17.4 → 2.4/7.8/14.1/14.1 | 3.6/8.4/11.2/17.0 → 3.0/6.6/11.5/14.0 | 4.5/9.4/11.9/12.6 → 16.6/16.8/16.9/17.8 |
-| rs | stress | 6.4/12.4/20.7/29.5 → 2.1/6.5/12.2/14.5 | 3.6/10.4/19.3/29.5 → 2.1/6.3/10.0/14.4 | 3.3/8.7/9.6/12.4 → 16.5/16.8/17.1/22.2 |
-| rs | jumps | 6.5/10.1/15.9/21.5 → 3.8/11.2/12.8/20.5 | 4.0/8.8/13.5/21.1 → 1.9/7.1/11.6/20.0 | 8.4/17.1/21.2/23.8 → 16.6/16.8/18.0/22.7 |
-| rs | kinds | 6.3/10.9/21.0/220.1 → 3.3/8.1/18.4/74.4 | 4.1/8.7/13.7/218.1 → 2.4/7.1/11.7/67.5 | 4.2/9.1/11.4/18.6 → 16.6/16.8/17.6/39.0 |
-| rs | panes | 6.8/9.8/20.8/84.3 → 5.4/9.4/17.9/39.2 | 5.2/8.7/13.8/84.0 → 3.6/8.1/14.6/39.1 | 4.3/9.1/14.8/25.7 → 16.6/16.8/18.0/43.2 |
-| rs | longlines | 8.7/17.6/189.4/218.7 → 7.7/17.9/26.5/37.7 | 4.7/9.1/13.7/210.2 → 3.6/9.3/19.9/36.7 | 4.3/9.2/11.3/22.0 → 16.5/16.8/17.3/22.1 |
-| dbg | scroll | 8.2/15.6/33.3/40.2 → 4.4/15.6/23.7/29.7 | 3.7/8.9/15.5/32.9 → 3.1/7.5/17.6/23.2 | 4.4/8.9/11.8/15.0 → 16.6/16.8/17.3/18.0 |
-| dbg | stress | 4.1/25.0/51.6/65.8 → 4.0/8.6/39.0/111.1 | 3.9/17.4/43.0/63.1 → 3.6/8.3/22.6/111.1 | 4.0/9.3/16.7/33.7 → 16.4/16.8/23.1/111.2 |
-| dbg | jumps | 12.7/23.2/41.8/247.9 → 11.5/21.2/35.9/280.9 | 6.9/15.0/33.3/247.4 → 3.5/11.5/22.9/280.4 | 8.1/16.9/30.0/38.9 → 16.6/16.9/29.6/73.9 |
-| dbg | kinds | 8.1/18.7/53.0/1979.3 → 4.6/15.2/56.8/438.8 | 4.4/12.3/28.9/1963.1 → 3.6/9.0/24.4/188.1 | 4.3/8.8/12.3/72.5 → 16.5/16.9/24.5/115.3 |
-| dbg | panes | 7.9/16.2/65.6/304.3 → 6.2/16.1/64.1/266.2 | 5.3/13.1/18.7/301.1 → 4.1/12.9/23.7/262.6 | 4.6/9.2/27.9/108.3 → 16.6/16.8/37.1/146.3 |
-| dbg | longlines | 16.6/33.1/35.8/2801.3 → 16.0/33.0/130.0/319.3 | 10.1/15.7/25.3/1963.0 → 9.2/16.5/80.0/180.3 | 8.6/16.8/27.4/34.5 → 16.6/16.8/18.2/167.2 |
-
-CPU (user s) per run, before -> after: base-rs-scroll 1.968, after-rs-scroll 7.072, base-rs-stress 4.059, after-rs-stress 14.329, base-rs-jumps 17.935, after-rs-jumps 48.597, base-rs-kinds 34.805, after-rs-kinds 94.319, base-rs-panes 11.838, after-rs-panes 24.085, base-rs-longlines 37.191, after-rs-longlines 70.976, base-dbg-scroll 2.524, after-dbg-scroll 7.842, base-dbg-stress 6.108, after-dbg-stress 18.640, base-dbg-jumps 31.800, after-dbg-jumps 60.583, base-dbg-kinds 56.343, after-dbg-kinds 109.000, base-dbg-panes 19.374, after-dbg-panes 29.634, base-dbg-longlines 77.796, after-dbg-longlines 90.633.
-The spin is most of the difference (a core while something moves).
-
-Read in Debug: the before's long-line spikes were 2 s (a key in a 500 KB
-line repainting it); after, a key there is 80-180 ms in Debug (16 KB of
-tree-sitter through the DebugAllocator: position 2) and 18-22 ms in
-ReleaseSafe; the pooled Debug p99 rises because those keys are many and
-the 2 s ones were few, the max falls 2801 -> 319.
-
-### Per scenario
-
-#### GUI ReleaseSafe: `scroll`
-
-| scenario | | latency ms p50/p90/p99/max | frame work ms | present gap ms (moving) | rows/s | late |
-|---|---|---|---|---|---|---|
-| notches | before | 3.8/6.9/6.9/6.9 | 3.4/8.4/10.8/11.5 | 4.5/9.5/11.9/12.6 | 7 | 0 |
-| | after | 1.9/8.4/8.4/8.4 | 1.8/3.2/8.3/8.3 | 16.7/16.9/17.8/17.8 | 7 | 0 |
-| spin | before | 2.9/8.0/9.2/9.2 | 3.2/8.2/9.2/10.0 | 4.0/9.0/10.2/11.1 | 35 | 0 |
-| | after | 2.4/7.1/11.8/11.8 | 2.3/6.4/11.7/11.7 | 8.8/16.7/17.0/17.0 | 34 | 0 |
-| pages | before | 6.8/7.2/17.4/17.4 | 6.7/7.1/17.0/17.0 | - | 3788 | 0 |
-| | after | 7.6/14.1/14.1/14.1 | 4.1/7.6/13.6/14.0 | 16.6/16.8/16.9/17.0 | 3598 | 0 |
-| retarget | before | 8.4/11.2/11.2/11.2 | 8.4/11.1/11.1/11.1 | - | 150 | 0 |
-| | after | 5.0/11.4/11.4/11.4 | 5.0/5.2/11.3/11.3 | 16.6/16.8/16.9/16.9 | 92 | 0 |
-| cursor | before | 6.9/8.8/11.4/11.4 | 6.1/8.8/11.3/11.5 | 5.5/9.8/12.4/12.5 | 0 | 0 |
-| | after | 2.4/8.2/10.6/10.6 | 4.0/6.7/10.4/11.5 | 16.6/16.7/16.9/17.7 | 0 | 0 |
-
-#### GUI ReleaseSafe: `stress`
-
-| scenario | | latency ms p50/p90/p99/max | frame work ms | present gap ms (moving) | rows/s | late |
-|---|---|---|---|---|---|---|
-| spin | before | 2.8/8.0/9.9/11.3 | 2.8/7.9/9.7/11.3 | 3.3/8.7/9.6/12.4 | 189 | 0 |
-| | after | 2.1/6.0/7.6/10.6 | 2.2/6.5/7.5/10.6 | 8.2/13.2/16.8/16.9 | 175 | 0 |
-| held-ctrl-d | before | 7.6/15.1/20.0/20.0 | 7.5/15.0/20.0/20.0 | - | 539 | 0 |
-| | after | 2.2/6.5/14.5/14.5 | 2.1/6.1/9.3/14.4 | 16.6/16.9/17.3/18.2 | 508 | 0 |
-| held-j | before | 6.8/11.1/15.3/15.3 | 6.7/11.0/15.3/15.3 | - | 35 | 0 |
-| | after | 2.0/6.3/11.5/11.5 | 2.2/6.0/7.6/11.4 | 16.6/16.9/17.1/17.8 | 33 | 0 |
-| held-pagedown | before | 11.2/19.4/29.5/29.5 | 11.1/19.3/29.5/29.5 | - | 758 | 0 |
-| | after | 2.1/5.7/7.8/7.8 | 2.2/6.3/7.7/10.3 | 16.6/16.8/17.3/22.2 | 703 | 0 |
-| far | before | 8.1/17.2/17.6/17.6 | 8.1/15.8/16.2/16.2 | - | 64670 | 0 |
-| | after | 12.2/12.4/12.5/12.5 | 1.8/8.6/12.3/12.4 | 16.6/16.8/17.0/17.0 | 60219 | 0 |
-
-#### GUI ReleaseSafe: `jumps`
-
-| scenario | | latency ms p50/p90/p99/max | frame work ms | present gap ms (moving) | rows/s | late |
-|---|---|---|---|---|---|---|
-| counted-G-big | before | 6.9/12.0/16.5/16.5 | 6.5/9.6/16.1/16.1 | 4.7/9.4/10.7/10.7 | 9605 | 0 |
-| | after | 8.6/12.5/12.8/12.8 | 5.2/8.5/12.1/12.3 | 16.6/16.8/22.5/22.5 | 8998 | 0 |
-| pages-up-big | before | 8.2/14.2/21.2/21.2 | 8.1/14.1/21.1/21.1 | - | 1784 | 0 |
-| | after | 8.4/8.8/15.4/15.4 | 4.1/6.0/12.7/15.3 | 16.6/16.8/18.0/18.1 | 1706 | 0 |
-| search-n-big | before | 7.3/9.1/10.1/10.1 | 4.3/8.9/10.7/12.4 | 5.0/10.0/21.6/23.8 | 1255 | 0 |
-| | after | 11.4/11.8/11.8/11.8 | 1.8/7.0/11.7/11.7 | 16.6/16.8/16.9/18.0 | 1255 | 0 |
-| rail-big | before | 6.2/15.6/15.6/15.6 | 4.5/8.9/15.5/15.5 | 5.3/9.9/12.5/14.9 | 3028 | 0 |
-| | after | 8.5/12.6/17.4/17.4 | 1.9/6.2/12.5/17.2 | 16.6/16.8/17.0/18.0 | 3030 | 0 |
-| query-ctrl-n-big | before | 7.8/15.9/16.1/16.1 | 4.2/10.3/15.8/16.0 | 4.7/10.2/16.1/16.1 | 624 | 0 |
-| | after | 6.3/8.8/9.2/9.2 | 1.8/6.0/9.1/11.7 | 16.6/16.8/17.9/18.2 | 724 | 0 |
-| star-big | before | 6.9/10.1/10.1/10.1 | 3.5/8.6/11.2/11.2 | 4.5/10.6/16.0/16.0 | 853 | 0 |
-| | after | 11.5/11.7/11.7/11.7 | 1.9/11.4/11.6/11.6 | 16.6/16.8/16.9/16.9 | 853 | 0 |
-| jumplist-big | before | 12.5/15.6/19.1/19.1 | 6.8/12.6/19.0/19.0 | 16.9/18.9/22.7/22.7 | 4506 | 0 |
-| | after | 11.6/11.8/11.9/11.9 | 1.8/7.1/11.7/11.8 | 16.6/16.9/18.7/21.9 | 7227 | 0 |
-| insert-big | before | 3.3/7.4/12.6/21.5 | 2.4/7.0/12.3/21.0 | 15.9/16.8/22.0/22.4 | 42 | 0 |
-| | after | 2.5/6.2/8.6/20.5 | 1.9/6.0/7.5/20.0 | 16.6/16.8/16.9/17.9 | 58 | 0 |
-| counted-G-small | before | 9.1/12.4/12.4/12.4 | 9.0/12.3/12.3/12.3 | - | 384 | 0 |
-| | after | 8.2/11.4/11.4/11.4 | 2.5/8.6/11.3/12.0 | 16.6/16.8/18.2/18.2 | 363 | 0 |
-| pages-up-small | before | 7.6/15.6/15.7/15.7 | 7.5/15.4/15.6/15.6 | - | 78 | 0 |
-| | after | 11.6/11.8/12.6/12.6 | 1.8/6.8/11.6/12.5 | 16.6/16.8/17.0/17.5 | 74 | 0 |
-| search-n-small | before | 8.9/10.1/12.0/12.0 | 3.3/8.8/10.3/11.7 | 4.0/9.3/12.2/14.1 | 44 | 0 |
-| | after | 7.2/10.7/11.4/11.4 | 5.6/7.1/11.3/11.3 | 16.6/16.8/18.1/18.1 | 45 | 0 |
-| rail-small | before | 8.8/9.8/12.3/12.3 | 8.8/9.7/12.3/12.3 | - | 71 | 0 |
-| | after | 7.3/7.8/8.7/8.7 | 7.1/7.5/8.5/8.5 | 16.6/16.8/16.9/16.9 | 72 | 0 |
-| query-ctrl-n-small | before | 8.7/9.0/12.5/12.5 | 4.6/8.7/10.3/12.3 | 5.1/10.0/16.2/16.2 | 91 | 0 |
-| | after | 7.1/7.6/10.8/10.8 | 1.8/7.0/10.5/10.5 | 16.6/16.8/17.5/17.5 | 94 | 0 |
-| star-small | before | 9.3/10.4/10.4/10.4 | 6.8/9.5/11.0/11.0 | 6.9/9.7/12.4/12.4 | 166 | 0 |
-| | after | 7.8/8.4/8.4/8.4 | 1.8/7.7/9.2/9.2 | 16.6/16.9/22.7/22.7 | 155 | 0 |
-| look-small | before | 7.5/7.5/7.5/7.5 | 7.4/7.4/7.4/7.4 | - | 0 | 0 |
-| | after | 8.3/9.6/9.6/9.6 | 5.9/9.4/12.2/12.2 | 16.6/16.8/18.0/18.0 | 80 | 0 |
-| jumplist-small | before | 9.4/10.2/12.2/12.2 | 8.7/10.1/12.1/12.1 | 16.6/17.2/22.0/22.0 | 88 | 0 |
-| | after | 8.2/9.5/11.2/11.2 | 6.6/8.1/12.3/12.3 | 16.6/17.2/22.4/22.4 | 87 | 0 |
-| insert-small | before | 4.6/8.7/12.1/12.3 | 4.1/8.2/12.2/13.1 | 18.1/20.4/23.2/23.5 | 27 | 0 |
-| | after | 2.0/6.2/8.0/12.9 | 1.8/5.7/7.6/12.8 | 16.6/16.8/17.4/17.9 | 38 | 0 |
-| look-new-pane | before | 7.8/7.8/7.8/7.8 | 4.5/7.0/9.7/9.7 | 16.6/17.4/18.8/18.8 | 0 | 0 |
-| | after | 14.8/14.8/14.8/14.8 | 14.0/14.0/14.0/14.0 | - | 0 | 0 |
-
-#### GUI ReleaseSafe: `kinds`
-
-| scenario | | latency ms p50/p90/p99/max | frame work ms | present gap ms (moving) | rows/s | late |
-|---|---|---|---|---|---|---|
-| generated-parser.c | before | 4.4/10.4/15.9/25.6 | 3.3/8.3/13.4/23.7 | 4.0/8.9/10.7/12.3 | 3794 | 0 |
-| | after | 2.1/6.7/11.6/22.9 | 2.1/6.9/12.2/21.4 | 16.6/16.8/17.2/18.0 | 3788 | 0 |
-| commands.rs | before | 6.8/11.8/22.4/62.2 | 4.1/9.0/18.2/60.6 | 4.3/9.0/10.8/11.8 | 1696 | 0 |
-| | after | 2.0/7.0/11.7/16.2 | 1.9/6.5/11.5/14.7 | 16.5/16.8/18.1/22.6 | 1694 | 0 |
-| tables.py | before | 6.9/12.7/16.3/25.2 | 3.7/8.7/13.8/23.4 | 4.2/9.1/11.4/12.3 | 2447 | 0 |
-| | after | 2.0/6.4/9.5/11.4 | 1.9/5.7/11.3/12.6 | 16.6/16.8/17.9/22.6 | 2447 | 0 |
-| jquery.js | before | 6.8/10.2/14.5/199.5 | 4.2/9.1/13.2/197.7 | 4.1/9.1/10.9/11.5 | 2154 | 0 |
-| | after | 1.9/7.0/11.8/31.1 | 1.9/6.7/11.3/29.3 | 16.5/16.8/17.4/22.7 | 2147 | 0 |
-| minified.js | before | 7.0/16.8/21.4/21.9 | 5.7/9.0/11.5/12.5 | 7.5/10.0/11.8/11.8 | 0 | 0 |
-| | after | 7.0/16.5/22.5/22.6 | 4.8/8.3/12.0/12.5 | 16.6/16.8/16.9/16.9 | 0 | 0 |
-| grammar.json | before | 6.2/9.9/13.1/205.0 | 3.2/8.1/12.1/204.2 | 4.0/8.6/9.7/10.5 | 2286 | 0 |
-| | after | 1.8/7.0/11.5/22.4 | 1.8/6.9/11.1/20.4 | 16.6/16.8/16.9/18.0 | 2282 | 0 |
-| oneline.json | before | 6.7/8.9/13.1/25.3 | 6.2/8.9/12.5/24.0 | 5.0/10.0/11.5/11.5 | 4 | 0 |
-| | after | 6.8/10.1/35.0/47.4 | 6.4/9.1/32.0/45.8 | 16.6/16.8/16.8/16.8 | 4 | 0 |
-| CHANGELOG.md | before | 4.1/16.5/26.4/28.2 | 3.3/8.7/21.8/28.1 | 4.0/9.0/11.6/13.6 | 715 | 0 |
-| | after | 2.9/7.6/13.7/25.4 | 2.4/7.1/11.4/25.2 | 16.6/16.8/22.7/25.5 | 712 | 2 |
-| cascade.html | before | 5.2/9.4/16.3/87.4 | 3.3/8.3/12.1/79.9 | 3.9/8.9/10.4/11.1 | 645 | 0 |
-| | after | 2.2/6.7/10.8/74.4 | 1.9/6.5/11.1/67.5 | 16.6/16.8/17.0/17.6 | 645 | 0 |
-| simdutf.cpp | before | 5.1/12.5/23.0/24.5 | 4.0/9.1/22.5/24.4 | 4.4/8.9/10.8/12.8 | 9874 | 0 |
-| | after | 2.0/6.9/11.0/11.0 | 1.9/6.2/10.8/11.0 | 16.6/16.8/17.2/17.6 | 9872 | 0 |
-| parser.ts | before | 5.4/9.5/13.0/33.0 | 4.5/8.9/12.3/32.0 | 4.5/9.6/11.4/12.0 | 1779 | 0 |
-| | after | 2.0/7.0/10.6/25.8 | 2.0/6.8/11.9/24.8 | 16.5/16.8/17.5/18.3 | 1778 | 0 |
-| fcnal-test.sh | before | 4.1/10.4/14.0/57.5 | 3.3/8.6/12.2/55.3 | 4.0/8.9/10.6/11.6 | 1037 | 0 |
-| | after | 1.9/7.0/10.8/58.8 | 1.9/6.6/10.7/56.6 | 16.6/16.8/17.5/18.2 | 1036 | 0 |
-| data.go | before | 6.9/9.2/14.3/28.0 | 3.9/8.6/12.3/27.8 | 4.3/9.1/11.2/12.1 | 3898 | 0 |
-| | after | 2.1/6.8/10.6/38.6 | 2.0/6.2/11.7/38.1 | 16.6/16.8/17.1/18.2 | 3892 | 0 |
-| guide.typ | before | 6.8/13.2/17.7/17.9 | 4.3/9.0/15.6/17.7 | 4.4/9.2/11.3/13.3 | 119 | 0 |
-| | after | 2.6/7.9/13.3/19.3 | 2.0/7.1/12.5/15.6 | 16.5/16.8/18.3/23.7 | 119 | 0 |
-| unicode.txt | before | 6.9/9.8/15.6/53.9 | 4.1/8.7/11.7/21.0 | 4.4/9.3/11.6/13.1 | 4642 | 0 |
-| | after | 2.1/6.9/17.3/55.2 | 1.9/6.3/11.2/22.0 | 16.6/16.8/17.0/18.4 | 4635 | 0 |
-| big.log | before | 3.8/11.0/15.8/29.8 | 3.3/8.8/12.8/13.9 | 4.1/8.9/10.7/12.4 | 69343 | 0 |
-| | after | 2.4/8.7/23.4/24.8 | 2.4/7.2/12.3/13.9 | 16.5/16.7/17.0/17.8 | 69269 | 0 |
-| words.txt | before | 6.4/8.6/12.4/15.7 | 3.6/8.5/11.4/12.3 | 4.1/9.3/11.2/12.5 | 92633 | 0 |
-| | after | 1.9/6.6/13.8/14.3 | 2.2/7.0/11.9/12.0 | 16.6/16.8/17.3/22.2 | 92628 | 0 |
-| hugelines.txt | before | 6.7/10.0/16.5/177.6 | 6.0/9.3/12.3/177.1 | 7.0/15.7/18.6/18.6 | 12 | 0 |
-| | after | 6.8/10.6/15.2/29.9 | 6.3/10.2/12.1/28.1 | 16.6/17.0/39.0/39.0 | 12 | 1 |
-| minified.js! | before | 7.2/16.1/21.3/220.1 | 5.3/9.0/12.4/218.1 | 6.4/12.2/17.0/17.0 | 0 | 0 |
-| | after | 7.2/17.3/22.3/27.3 | 5.0/8.7/12.4/25.4 | 16.6/16.8/16.9/16.9 | 0 | 0 |
-| oneline.json! | before | 6.7/10.5/13.0/17.4 | 5.8/9.4/12.3/12.8 | 4.9/9.5/11.0/11.1 | 0 | 0 |
-| | after | 6.4/10.2/20.0/38.1 | 5.3/8.9/12.7/37.6 | 16.6/16.8/16.9/16.9 | 0 | 0 |
-| hugelines.txt! | before | 6.4/9.6/16.0/18.2 | 5.4/8.7/11.7/13.7 | 5.0/9.7/12.0/12.3 | 2 | 0 |
-| | after | 5.6/9.4/13.6/16.5 | 3.6/8.1/11.5/13.4 | 16.6/16.8/17.0/17.2 | 2 | 0 |
-
-#### GUI ReleaseSafe: `panes`
-
-| scenario | | latency ms p50/p90/p99/max | frame work ms | present gap ms (moving) | rows/s | late |
-|---|---|---|---|---|---|---|
-| command-cat | before | 7.1/9.1/14.5/31.5 | 5.9/8.2/12.9/25.9 | 4.5/9.0/14.8/14.8 | 4668 | 0 |
-| | after | 6.7/11.1/13.7/24.6 | 6.2/10.2/12.3/12.7 | 16.6/16.9/22.8/22.8 | 4616 | 0 |
-| command-yes | before | 6.7/9.6/13.0/22.6 | 6.0/8.9/12.6/20.6 | 5.7/9.8/11.9/11.9 | 4855 | 0 |
-| | after | 6.1/9.1/12.7/23.5 | 4.6/8.5/12.5/21.4 | 16.6/16.8/17.4/17.4 | 4924 | 0 |
-| pager | before | 7.0/9.2/12.7/12.9 | 6.1/8.7/12.3/12.6 | 4.4/9.0/13.0/13.0 | 0 | 0 |
-| | after | 6.9/9.7/14.3/14.6 | 5.5/9.2/12.4/14.6 | 16.6/16.8/17.3/17.3 | 0 | 0 |
-| directory | before | 6.4/9.3/12.2/12.3 | 4.4/8.6/12.0/12.2 | 4.4/9.1/11.4/12.2 | 2 | 0 |
-| | after | 5.6/8.7/12.4/12.5 | 3.5/7.6/11.9/12.4 | 16.6/16.8/17.7/18.0 | 2 | 0 |
-| pdf | before | 6.9/16.1/44.7/44.9 | 6.4/13.3/31.5/44.8 | 3.9/9.1/9.5/9.5 | 0 | 0 |
-| | after | 7.0/16.6/38.5/39.2 | 6.4/14.6/34.5/39.1 | 16.6/16.8/17.8/17.8 | 0 | 0 |
-| image | before | 6.9/10.1/24.1/84.3 | 5.9/9.1/12.4/84.0 | 4.6/9.1/10.7/10.7 | 0 | 0 |
-| | after | 6.9/9.0/12.7/14.1 | 6.4/8.9/12.4/28.5 | 16.6/16.8/43.2/43.2 | 0 | 1 |
-| help | before | 6.8/8.8/12.1/12.6 | 3.5/8.2/11.5/12.5 | 4.2/8.8/11.5/12.4 | 69 | 0 |
-| | after | 2.6/7.1/8.0/14.4 | 1.9/6.8/8.0/14.0 | 16.6/16.8/17.1/18.0 | 69 | 0 |
-| changelog | before | 6.3/9.6/12.9/13.0 | 4.0/8.6/12.0/12.9 | 4.2/9.2/11.7/12.7 | 220 | 0 |
-| | after | 1.8/6.8/13.7/14.3 | 1.8/6.4/11.2/14.3 | 16.6/16.8/17.3/18.3 | 220 | 0 |
-| query-typing | before | 7.5/10.6/13.8/13.8 | 1.7/11.7/13.8/14.1 | 2.8/15.4/23.2/25.7 | 0 | 1 |
-| | after | 7.3/9.3/12.4/12.4 | 1.9/7.9/12.3/12.6 | 16.6/17.0/22.9/22.9 | 17 | 0 |
-
-#### GUI ReleaseSafe: `longlines`
-
-| scenario | | latency ms p50/p90/p99/max | frame work ms | present gap ms (moving) | rows/s | late |
-|---|---|---|---|---|---|---|
-| w-held-minified.js-wrap | before | 3.0/7.9/9.4/9.4 | 3.2/8.1/9.4/9.9 | 4.1/9.1/10.5/11.0 | 0 | 0 |
-| | after | 1.8/5.9/7.7/7.7 | 1.7/5.7/7.3/7.6 | 16.6/16.7/16.9/17.3 | 0 | 0 |
-| e-held-minified.js-wrap | before | 3.1/8.3/9.3/9.3 | 3.1/8.0/9.2/11.3 | 4.1/9.0/10.0/12.4 | 0 | 0 |
-| | after | 1.7/5.9/7.2/7.2 | 1.7/5.7/7.2/7.2 | 16.6/16.8/17.7/18.2 | 0 | 0 |
-| l-held-minified.js-wrap | before | 6.6/10.1/11.7/11.7 | 3.3/8.5/11.3/11.6 | 4.2/9.2/10.3/12.0 | 0 | 0 |
-| | after | 1.9/5.9/7.3/7.3 | 1.9/5.9/7.3/7.3 | 16.6/16.8/17.1/17.4 | 0 | 0 |
-| dollar-zero-minified.js-wrap | before | 7.5/8.7/8.7/8.7 | 7.3/8.6/8.6/8.6 | - | 0 | 0 |
-| | after | 7.5/7.7/7.7/7.7 | 3.2/7.7/11.4/11.4 | 16.6/16.9/16.9/16.9 | 0 | 0 |
-| f-t-minified.js-wrap | before | 7.3/7.6/7.8/7.8 | 7.2/7.5/7.7/7.7 | - | 0 | 0 |
-| | after | 7.2/7.5/9.0/9.0 | 4.6/11.2/11.6/12.0 | 14.5/16.7/17.9/18.0 | 0 | 0 |
-| gl-gh-minified.js-wrap | before | 7.4/7.6/7.7/7.7 | 7.3/7.5/7.6/7.6 | - | 0 | 0 |
-| | after | 7.4/8.1/8.4/8.4 | 5.0/11.0/11.4/11.4 | 13.9/16.7/18.7/18.7 | 0 | 0 |
-| insert-end-minified.js-wrap | before | 7.2/11.0/12.7/12.7 | 7.0/10.7/12.6/12.6 | 5.7/9.8/10.6/10.6 | 0 | 0 |
-| | after | 1.8/6.2/8.0/8.0 | 1.8/6.1/10.6/10.9 | 16.6/16.8/17.3/18.1 | 0 | 0 |
-| insert-middle-minified.js-wrap | before | 171.7/177.5/185.4/185.4 | 7.3/174.3/184.4/184.4 | 5.2/9.8/10.3/10.3 | 0 | 0 |
-| | after | 16.8/26.5/28.1/28.1 | 12.7/24.9/27.3/27.3 | 15.9/16.7/16.9/16.9 | 0 | 0 |
-| sweep-L-W-minified.js-wrap | before | 7.1/11.5/212.2/212.2 | 6.9/10.2/14.1/210.2 | 5.5/9.4/12.9/12.9 | 0 | 0 |
-| | after | 7.3/12.1/13.5/13.5 | 6.8/11.2/12.6/12.9 | 16.6/16.9/20.9/20.9 | 0 | 0 |
-| w-held-oneline.json-wrap | before | 3.9/10.1/11.3/11.3 | 4.1/8.6/10.9/11.5 | 4.8/9.6/11.4/12.0 | 0 | 0 |
-| | after | 1.8/6.0/13.3/13.3 | 1.8/5.7/7.3/13.2 | 16.6/16.8/17.0/17.6 | 0 | 0 |
-| e-held-oneline.json-wrap | before | 3.2/7.8/9.0/9.0 | 3.2/8.2/9.4/11.1 | 4.1/9.2/10.6/12.2 | 0 | 0 |
-| | after | 1.8/5.6/7.7/7.7 | 1.8/5.6/7.6/7.6 | 16.6/16.8/16.9/17.0 | 0 | 0 |
-| l-held-oneline.json-wrap | before | 4.4/10.4/13.3/13.3 | 3.2/8.4/11.9/13.2 | 4.0/9.1/11.0/13.4 | 0 | 0 |
-| | after | 1.9/6.1/7.5/7.5 | 1.8/6.0/7.4/7.5 | 16.6/16.8/17.2/17.4 | 0 | 0 |
-| dollar-zero-oneline.json-wrap | before | 7.6/8.0/8.0/8.0 | 7.4/7.7/7.7/7.7 | - | 0 | 0 |
-| | after | 9.0/9.4/9.4/9.4 | 2.3/8.9/10.8/10.8 | 16.5/16.9/17.0/17.0 | 0 | 0 |
-| f-t-oneline.json-wrap | before | 7.4/7.6/9.2/9.2 | 7.3/7.5/9.1/9.1 | - | 0 | 0 |
-| | after | 7.3/8.8/9.3/9.3 | 4.6/10.4/11.7/12.3 | 14.7/16.7/17.0/17.4 | 0 | 0 |
-| gl-gh-oneline.json-wrap | before | 7.5/7.6/7.6/7.6 | 7.3/7.4/7.5/7.5 | - | 0 | 0 |
-| | after | 7.5/7.7/8.6/8.6 | 4.0/10.6/11.5/11.5 | 14.5/16.6/16.9/16.9 | 0 | 0 |
-| insert-end-oneline.json-wrap | before | 7.4/10.8/12.8/12.8 | 7.2/10.6/12.9/12.9 | 7.8/10.3/14.0/14.0 | 0 | 0 |
-| | after | 1.8/6.0/7.4/7.4 | 1.8/5.9/11.1/11.3 | 16.6/16.8/17.0/17.0 | 0 | 0 |
-| insert-middle-oneline.json-wrap | before | 193.1/199.8/207.9/207.9 | 5.2/195.5/206.4/206.4 | 4.0/9.3/13.6/13.6 | 0 | 0 |
-| | after | 21.3/29.0/37.7/37.7 | 18.1/25.4/36.7/36.7 | 16.6/16.9/17.9/17.9 | 0 | 0 |
-| sweep-L-W-oneline.json-wrap | before | 7.2/11.3/31.8/31.8 | 6.8/9.3/12.2/12.5 | 4.4/9.3/10.1/10.5 | 6 | 0 |
-| | after | 7.7/12.1/21.5/21.5 | 6.9/11.0/12.6/21.5 | 16.6/16.8/16.9/16.9 | 6 | 0 |
-| w-held-hugelines.txt-wrap | before | 3.2/8.5/11.8/11.8 | 3.2/8.2/10.2/12.2 | 4.2/9.1/10.5/13.3 | 0 | 0 |
-| | after | 1.8/5.8/10.7/10.7 | 1.8/5.5/7.4/10.6 | 16.6/16.8/17.3/17.9 | 0 | 0 |
-| e-held-hugelines.txt-wrap | before | 5.1/7.3/11.7/11.7 | 4.1/8.5/11.3/11.9 | 4.5/9.5/12.1/12.9 | 0 | 0 |
-| | after | 3.7/7.3/12.6/12.6 | 2.4/7.1/12.5/12.5 | 16.6/16.8/18.1/18.1 | 0 | 0 |
-| l-held-hugelines.txt-wrap | before | 7.3/11.6/13.0/13.0 | 7.1/11.4/12.9/12.9 | - | 0 | 0 |
-| | after | 7.2/11.4/13.4/13.4 | 7.1/11.2/13.2/13.2 | - | 0 | 0 |
-| dollar-zero-hugelines.txt-wrap | before | 7.3/7.4/7.4/7.4 | 7.1/7.2/7.2/7.2 | - | 0 | 0 |
-| | after | 7.3/7.3/7.3/7.3 | 6.0/10.4/11.0/11.0 | 15.6/16.6/16.7/16.7 | 0 | 0 |
-| f-t-hugelines.txt-wrap | before | 7.2/7.5/7.6/7.6 | 7.1/7.3/7.5/7.5 | - | 0 | 0 |
-| | after | 7.2/7.4/9.2/9.2 | 4.4/10.9/11.7/11.9 | 14.6/16.7/16.9/16.9 | 0 | 0 |
-| gl-gh-hugelines.txt-wrap | before | 7.4/7.6/7.6/7.6 | 7.2/7.4/7.5/7.5 | - | 0 | 0 |
-| | after | 7.4/7.9/10.8/10.8 | 4.6/10.6/11.2/11.2 | 14.8/16.8/17.7/17.7 | 0 | 0 |
-| insert-end-hugelines.txt-wrap | before | 7.4/10.6/12.1/12.1 | 7.2/10.5/12.0/12.0 | - | 0 | 0 |
-| | after | 2.1/7.4/8.0/8.0 | 2.2/7.3/10.7/10.8 | 16.6/16.8/16.9/17.0 | 0 | 0 |
-| insert-middle-hugelines.txt-wrap | before | 7.9/10.7/13.0/13.0 | 7.2/10.2/12.1/12.1 | - | 0 | 0 |
-| | after | 7.9/12.9/13.5/13.5 | 7.2/12.2/12.4/12.4 | - | 0 | 0 |
-| sweep-L-W-hugelines.txt-wrap | before | 7.8/12.3/14.9/14.9 | 7.0/11.5/13.8/14.8 | 5.8/10.8/19.9/22.0 | 0 | 0 |
-| | after | 7.7/11.9/13.1/13.1 | 7.1/12.1/13.6/14.6 | 16.7/16.9/20.1/20.1 | 0 | 0 |
-| w-held-minified.js-nowrap | before | 7.1/16.3/28.5/28.5 | 3.5/8.2/10.2/13.7 | 4.5/9.2/10.9/11.3 | 0 | 0 |
-| | after | 2.7/15.9/22.6/22.6 | 2.1/3.9/12.7/12.7 | 16.6/16.8/17.9/17.9 | 0 | 0 |
-| e-held-minified.js-nowrap | before | 15.5/19.6/24.7/24.7 | 4.4/8.4/10.1/10.1 | - | 0 | 0 |
-| | after | 15.2/18.8/23.0/23.0 | 4.1/8.2/9.4/9.4 | - | 0 | 0 |
-| l-held-minified.js-nowrap | before | 15.7/20.7/27.7/27.7 | 4.3/8.4/11.3/11.3 | - | 0 | 0 |
-| | after | 15.3/19.6/24.2/24.2 | 3.9/8.9/9.9/9.9 | - | 0 | 0 |
-| dollar-zero-minified.js-nowrap | before | 9.2/10.1/10.1/10.1 | 9.0/9.9/9.9/9.9 | - | 0 | 0 |
-| | after | 9.3/10.4/10.4/10.4 | 3.7/9.1/10.2/10.2 | 16.6/16.8/16.9/16.9 | 0 | 0 |
-| f-t-minified.js-nowrap | before | 8.9/9.1/9.9/9.9 | 8.7/9.0/9.8/9.8 | - | 0 | 0 |
-| | after | 8.7/9.0/9.7/9.7 | 3.3/8.8/10.7/11.0 | 16.0/16.7/17.9/18.3 | 0 | 0 |
-| gl-gh-minified.js-nowrap | before | 9.1/9.2/10.4/10.4 | 8.9/9.1/10.3/10.3 | - | 0 | 0 |
-| | after | 8.9/9.2/9.3/9.3 | 2.8/8.9/9.7/9.7 | 16.1/16.8/18.2/18.2 | 0 | 0 |
-| insert-end-minified.js-nowrap | before | 8.8/12.1/13.8/13.8 | 8.5/11.0/13.7/13.7 | 4.5/8.9/11.9/11.9 | 0 | 0 |
-| | after | 2.1/7.3/9.0/9.0 | 2.2/7.1/9.1/9.4 | 16.6/16.8/16.9/17.1 | 0 | 0 |
-| insert-middle-minified.js-nowrap | before | 185.1/192.3/201.4/201.4 | 3.4/172.0/182.2/182.2 | 4.0/8.8/9.4/9.4 | 0 | 0 |
-| | after | 16.1/26.7/29.1/29.1 | 8.5/15.2/18.4/18.4 | 16.6/16.9/17.2/17.2 | 0 | 0 |
-| sweep-L-W-minified.js-nowrap | before | 11.2/19.6/26.4/26.4 | 7.0/10.2/13.9/14.2 | 5.6/9.6/11.6/12.2 | 0 | 0 |
-| | after | 11.9/20.0/26.9/26.9 | 3.9/10.9/13.9/26.2 | 16.6/16.8/17.1/17.1 | 1 | 0 |
-| w-held-oneline.json-nowrap | before | 8.8/16.9/21.6/21.6 | 3.3/7.9/8.9/13.2 | 4.2/8.9/9.9/10.6 | 0 | 0 |
-| | after | 8.1/20.2/23.4/23.4 | 2.3/5.8/9.3/9.3 | 16.5/16.7/19.7/19.7 | 0 | 0 |
-| e-held-oneline.json-nowrap | before | 15.8/20.7/24.4/24.4 | 4.5/8.0/9.4/9.4 | - | 0 | 0 |
-| | after | 15.7/20.7/24.5/24.5 | 3.8/6.3/8.5/8.5 | - | 0 | 0 |
-| l-held-oneline.json-nowrap | before | 15.7/19.1/29.3/29.3 | 3.6/7.4/9.1/9.1 | - | 0 | 0 |
-| | after | 15.8/20.5/25.1/25.1 | 3.8/7.4/10.1/10.1 | - | 0 | 0 |
-| dollar-zero-oneline.json-nowrap | before | 9.9/9.9/9.9/9.9 | 9.6/9.7/9.7/9.7 | - | 0 | 0 |
-| | after | 9.8/10.2/10.2/10.2 | 4.5/9.6/12.6/12.6 | 16.4/16.8/17.8/17.8 | 0 | 0 |
-| f-t-oneline.json-nowrap | before | 9.7/9.8/10.0/10.0 | 9.6/9.7/9.9/9.9 | - | 0 | 0 |
-| | after | 9.4/9.6/10.1/10.1 | 3.2/9.4/13.1/13.2 | 16.1/16.8/18.2/18.6 | 0 | 0 |
-| gl-gh-oneline.json-nowrap | before | 9.6/9.9/10.0/10.0 | 9.5/9.7/9.8/9.8 | - | 0 | 0 |
-| | after | 9.6/10.3/10.4/10.4 | 2.7/9.4/10.2/10.2 | 16.5/16.8/22.1/22.1 | 0 | 0 |
-| insert-end-oneline.json-nowrap | before | 9.5/13.3/15.2/15.2 | 9.2/11.7/15.1/15.1 | 5.2/9.5/12.7/12.7 | 0 | 0 |
-| | after | 2.3/7.3/10.9/10.9 | 2.3/7.0/9.9/10.8 | 16.6/16.8/17.0/17.8 | 0 | 0 |
-| insert-middle-oneline.json-nowrap | before | 209.8/216.8/218.7/218.7 | 3.0/195.5/196.9/196.9 | 4.0/9.0/9.4/9.4 | 0 | 0 |
-| | after | 23.1/29.0/35.4/35.4 | 10.3/21.1/28.7/28.7 | 16.6/16.8/17.9/17.9 | 0 | 0 |
-| sweep-L-W-oneline.json-nowrap | before | 10.3/14.9/24.9/24.9 | 6.6/11.0/14.2/14.8 | 4.4/9.3/14.3/14.4 | 0 | 0 |
-| | after | 10.0/15.6/21.7/21.7 | 4.9/11.6/15.0/15.5 | 16.6/16.8/18.2/18.2 | 0 | 0 |
-| w-held-hugelines.txt-nowrap | before | 12.8/18.8/22.8/22.8 | 4.5/8.1/11.5/12.7 | 4.2/8.9/9.9/11.2 | 0 | 0 |
-| | after | 9.8/18.4/20.3/20.3 | 2.4/7.9/11.2/11.2 | 16.5/16.8/18.2/18.2 | 0 | 0 |
-| e-held-hugelines.txt-nowrap | before | 15.7/20.0/20.9/20.9 | 6.0/9.7/11.9/11.9 | - | 0 | 0 |
-| | after | 14.9/19.2/20.8/20.8 | 5.6/9.7/10.7/10.7 | - | 0 | 0 |
-| l-held-hugelines.txt-nowrap | before | 15.4/20.1/25.1/25.1 | 5.3/10.0/13.7/13.7 | - | 0 | 0 |
-| | after | 15.6/19.4/21.8/21.8 | 5.2/9.2/11.3/11.3 | - | 0 | 0 |
-| dollar-zero-hugelines.txt-nowrap | before | 9.7/9.8/9.8/9.8 | 9.5/9.7/9.7/9.7 | - | 0 | 0 |
-| | after | 9.4/9.4/9.4/9.4 | 3.1/8.5/9.3/9.3 | 16.6/16.8/16.9/16.9 | 0 | 0 |
-| f-t-hugelines.txt-nowrap | before | 9.4/10.4/11.4/11.4 | 9.3/10.3/11.2/11.2 | - | 0 | 0 |
-| | after | 9.2/9.4/9.9/9.9 | 2.8/9.0/9.4/9.8 | 16.3/16.7/16.8/17.7 | 0 | 0 |
-| gl-gh-hugelines.txt-nowrap | before | 9.4/9.9/10.1/10.1 | 9.3/9.7/9.9/9.9 | - | 0 | 0 |
-| | after | 9.3/9.9/10.9/10.9 | 3.0/9.2/10.7/10.7 | 16.3/16.7/17.0/17.0 | 0 | 0 |
-| insert-end-hugelines.txt-nowrap | before | 9.6/13.7/15.1/15.1 | 9.0/12.6/14.9/14.9 | 6.0/9.3/10.3/10.3 | 0 | 0 |
-| | after | 2.3/7.1/9.4/9.4 | 2.3/6.8/9.2/9.2 | 16.6/16.7/16.8/16.8 | 0 | 0 |
-| insert-middle-hugelines.txt-nowrap | before | 18.4/22.5/26.1/26.1 | 4.7/8.8/10.6/10.6 | 5.0/9.2/10.0/10.0 | 0 | 0 |
-| | after | 17.0/19.9/23.0/23.0 | 5.9/8.6/12.4/12.4 | 16.6/16.9/17.1/17.1 | 0 | 0 |
-| sweep-L-W-hugelines.txt-nowrap | before | 11.9/18.7/24.0/24.0 | 6.2/9.1/14.4/14.5 | 4.2/9.0/9.8/9.8 | 0 | 0 |
-| | after | 10.2/18.1/22.5/22.5 | 7.0/11.1/14.1/14.5 | 16.6/16.9/17.1/17.1 | 0 | 0 |
-
-#### GUI Debug: `scroll`
-
-| scenario | | latency ms p50/p90/p99/max | frame work ms | present gap ms (moving) | rows/s | late |
-|---|---|---|---|---|---|---|
-| notches | before | 8.2/14.4/14.4/14.4 | 3.4/7.8/9.3/14.3 | 4.4/8.6/10.1/10.4 | 7 | 0 |
-| | after | 3.5/19.8/19.8/19.8 | 3.0/3.5/19.8/19.8 | 16.7/17.0/17.6/17.6 | 7 | 0 |
-| spin | before | 3.6/8.8/15.8/15.8 | 3.5/8.6/14.0/15.7 | 4.2/8.9/12.7/14.1 | 35 | 0 |
-| | after | 4.3/8.4/20.4/20.4 | 4.3/7.7/20.4/20.4 | 10.1/16.7/17.1/17.1 | 34 | 0 |
-| pages | before | 15.4/25.7/40.2/40.2 | 15.0/25.2/32.8/32.8 | - | 3787 | 0 |
-| | after | 15.0/22.0/29.7/29.7 | 3.0/7.1/17.6/21.3 | 16.6/16.8/17.3/18.0 | 3607 | 0 |
-| retarget | before | 16.7/33.3/33.3/33.3 | 16.3/32.9/32.9/32.9 | - | 150 | 0 |
-| | after | 9.6/23.7/23.7/23.7 | 3.0/6.2/23.2/23.2 | 16.6/16.8/16.8/16.8 | 92 | 0 |
-| cursor | before | 12.2/15.6/16.2/16.2 | 6.5/11.9/15.4/15.5 | 5.6/9.7/12.5/15.0 | 0 | 0 |
-| | after | 3.8/14.9/18.4/18.4 | 3.6/6.3/17.1/17.7 | 16.6/16.8/17.2/18.0 | 0 | 0 |
-
-#### GUI Debug: `stress`
-
-| scenario | | latency ms p50/p90/p99/max | frame work ms | present gap ms (moving) | rows/s | late |
-|---|---|---|---|---|---|---|
-| spin | before | 3.6/8.9/25.4/33.7 | 3.6/8.8/17.2/33.7 | 4.0/9.3/16.7/33.7 | 188 | 3 |
-| | after | 3.7/7.9/13.1/23.3 | 3.6/7.3/11.9/23.2 | 8.2/13.1/16.7/16.8 | 175 | 0 |
-| held-ctrl-d | before | 8.3/33.2/45.9/45.9 | 8.1/32.8/45.5/45.5 | - | 539 | 0 |
-| | after | 4.3/8.5/24.4/24.4 | 3.7/7.5/12.6/24.0 | 16.6/17.2/18.4/23.1 | 508 | 0 |
-| held-j | before | 11.9/15.6/30.7/30.7 | 11.5/15.2/30.3/30.3 | - | 35 | 0 |
-| | after | 3.6/8.3/25.0/25.0 | 3.3/7.3/12.2/24.7 | 16.6/16.8/17.9/22.4 | 33 | 0 |
-| held-pagedown | before | 9.7/38.3/63.4/63.4 | 9.7/38.0/63.1/63.1 | - | 758 | 0 |
-| | after | 4.8/13.1/111.1/111.1 | 4.6/8.7/82.3/111.1 | 16.6/17.5/83.3/111.2 | 703 | 8 |
-| far | before | 15.2/56.1/65.8/65.8 | 14.8/32.4/35.9/35.9 | - | 64673 | 0 |
-| | after | 15.3/30.0/39.0/39.0 | 3.2/8.7/22.6/23.3 | 16.6/16.8/17.1/17.1 | 60139 | 0 |
-
-#### GUI Debug: `jumps`
-
-| scenario | | latency ms p50/p90/p99/max | frame work ms | present gap ms (moving) | rows/s | late |
-|---|---|---|---|---|---|---|
-| counted-G-big | before | 13.0/41.0/46.9/46.9 | 6.9/20.5/40.0/40.0 | 4.6/9.2/12.0/12.0 | 9607 | 0 |
-| | after | 12.4/26.7/32.5/32.5 | 3.0/11.1/19.3/24.9 | 16.6/16.8/17.0/17.0 | 9013 | 0 |
-| pages-up-big | before | 18.7/43.6/56.7/56.7 | 18.3/43.2/56.3/56.3 | - | 1784 | 0 |
-| | after | 20.1/28.3/46.7/46.7 | 5.0/8.7/27.9/46.2 | 16.6/16.8/22.3/22.4 | 1709 | 0 |
-| search-n-big | before | 12.7/14.8/33.3/33.3 | 4.3/9.3/14.2/27.8 | 4.6/9.3/18.2/19.2 | 1254 | 0 |
-| | after | 11.9/14.5/24.7/24.7 | 4.3/9.5/14.0/14.0 | 16.6/16.8/22.7/24.4 | 1255 | 0 |
-| rail-big | before | 8.2/37.9/50.1/50.1 | 3.5/7.9/37.0/49.3 | 4.4/8.7/10.4/10.9 | 3032 | 0 |
-| | after | 9.0/35.6/38.7/38.7 | 3.2/7.1/34.9/38.0 | 16.6/16.8/22.3/22.6 | 3030 | 0 |
-| query-ctrl-n-big | before | 22.2/39.4/40.3/40.3 | 7.2/21.8/38.3/38.9 | 7.1/15.5/31.1/31.1 | 628 | 2 |
-| | after | 7.0/22.4/28.3/28.3 | 3.6/8.8/18.9/19.4 | 16.6/16.8/26.4/27.4 | 727 | 3 |
-| star-big | before | 12.8/36.9/36.9/36.9 | 7.2/12.3/19.6/19.6 | 8.2/10.1/17.2/17.2 | 853 | 0 |
-| | after | 11.6/44.7/44.7/44.7 | 3.1/11.0/21.8/21.8 | 16.6/16.9/17.0/17.0 | 853 | 0 |
-| jumplist-big | before | 33.4/41.0/41.4/41.4 | 8.1/32.9/41.0/41.0 | 15.9/20.3/21.2/21.2 | 4759 | 0 |
-| | after | 17.9/23.4/23.5/23.5 | 3.0/12.4/23.0/23.0 | 16.6/16.8/18.4/20.1 | 7394 | 0 |
-| insert-big | before | 9.3/14.9/29.9/247.9 | 6.5/13.5/23.3/247.4 | 11.8/23.0/31.6/35.5 | 42 | 12 |
-| | after | 10.3/17.0/30.5/280.9 | 3.4/11.4/22.9/280.4 | 16.6/19.1/26.5/33.4 | 58 | 13 |
-| counted-G-small | before | 13.4/26.4/26.7/26.7 | 12.6/25.0/26.1/26.1 | - | 384 | 0 |
-| | after | 11.4/22.9/29.3/29.3 | 3.9/9.7/27.8/27.8 | 16.6/16.8/18.3/18.3 | 363 | 0 |
-| pages-up-small | before | 21.8/34.4/36.4/36.4 | 21.1/33.6/35.6/35.6 | - | 78 | 0 |
-| | after | 23.0/35.3/35.9/35.9 | 3.1/7.7/34.6/35.2 | 16.6/16.8/17.4/17.9 | 75 | 0 |
-| search-n-small | before | 15.2/20.5/25.3/25.3 | 3.2/14.4/20.2/24.5 | 4.0/9.0/31.3/31.3 | 44 | 1 |
-| | after | 16.2/17.5/21.1/21.1 | 3.0/16.0/21.2/21.2 | 16.6/16.9/31.3/31.3 | 45 | 1 |
-| rail-small | before | 14.3/16.1/18.2/18.2 | 14.1/15.1/18.2/18.2 | - | 71 | 0 |
-| | after | 14.5/17.5/21.8/21.8 | 6.8/16.0/20.3/20.3 | 16.6/16.8/17.0/17.0 | 72 | 0 |
-| query-ctrl-n-small | before | 15.0/17.0/19.5/19.5 | 4.2/14.3/17.4/17.9 | 4.8/9.1/19.0/19.0 | 91 | 0 |
-| | after | 12.8/17.4/17.4/17.4 | 3.4/13.6/16.7/16.7 | 16.6/16.8/25.3/25.3 | 94 | 1 |
-| star-small | before | 15.9/28.3/28.3/28.3 | 6.5/14.1/27.5/27.5 | 4.7/8.8/13.5/13.5 | 166 | 0 |
-| | after | 17.3/19.5/19.5/19.5 | 3.0/12.1/18.5/18.5 | 16.6/16.8/17.1/17.1 | 155 | 0 |
-| look-small | before | 26.1/26.1/26.1/26.1 | 24.7/24.7/24.7/24.7 | - | 0 | 0 |
-| | after | 15.4/20.6/20.6/20.6 | 7.7/14.3/20.4/20.4 | 16.7/16.8/16.9/16.9 | 81 | 0 |
-| jumplist-small | before | 16.3/28.9/29.2/29.2 | 9.8/25.2/28.3/28.3 | 16.7/18.8/21.2/21.2 | 81 | 0 |
-| | after | 17.1/21.5/27.1/27.1 | 4.7/12.1/20.8/26.3 | 16.6/17.0/18.8/19.7 | 87 | 0 |
-| insert-small | before | 13.1/19.4/32.8/41.8 | 7.7/17.0/29.2/41.5 | 13.0/24.1/34.0/38.9 | 27 | 13 |
-| | after | 9.6/20.6/28.4/61.6 | 4.9/14.0/23.4/60.6 | 16.4/18.5/32.4/73.9 | 37 | 35 |
-| look-new-pane | before | 14.9/14.9/14.9/14.9 | 8.2/8.7/8.8/8.8 | 16.8/17.3/17.9/17.9 | 0 | 0 |
-| | after | 24.9/24.9/24.9/24.9 | 18.4/18.4/18.4/18.4 | - | 0 | 0 |
-
-#### GUI Debug: `kinds`
-
-| scenario | | latency ms p50/p90/p99/max | frame work ms | present gap ms (moving) | rows/s | late |
-|---|---|---|---|---|---|---|
-| generated-parser.c | before | 7.3/22.4/29.5/103.2 | 3.6/11.7/24.9/86.5 | 4.2/8.7/10.1/13.0 | 3787 | 0 |
-| | after | 3.9/7.9/26.6/39.3 | 3.2/6.8/17.6/31.5 | 16.5/16.9/18.4/18.8 | 3789 | 0 |
-| commands.rs | before | 7.4/21.6/32.5/33.8 | 3.8/11.8/30.0/33.4 | 4.4/8.9/10.8/15.1 | 1694 | 0 |
-| | after | 4.3/8.7/23.0/35.4 | 3.8/8.0/12.8/18.8 | 16.5/17.2/18.5/23.3 | 1690 | 0 |
-| tables.py | before | 6.8/25.3/32.2/77.5 | 3.6/12.0/29.5/69.8 | 4.2/8.7/14.5/16.6 | 2444 | 0 |
-| | after | 3.9/8.9/25.5/26.7 | 3.2/7.8/17.2/24.8 | 16.5/16.9/17.9/23.1 | 2442 | 0 |
-| jquery.js | before | 7.6/18.8/28.0/1979.3 | 4.3/12.1/23.3/1958.4 | 4.2/8.7/10.6/12.6 | 2217 | 0 |
-| | after | 3.8/8.4/27.9/119.7 | 3.2/7.1/14.6/98.3 | 16.5/16.9/17.6/17.8 | 2149 | 0 |
-| minified.js | before | 8.4/18.7/29.4/34.0 | 4.1/12.4/17.2/17.5 | 4.4/8.4/9.1/9.1 | 0 | 0 |
-| | after | 7.7/19.3/28.9/32.5 | 4.3/13.1/16.6/18.4 | 16.6/16.8/17.2/17.2 | 0 | 0 |
-| grammar.json | before | 7.8/17.3/23.9/1870.4 | 3.9/12.2/21.5/1850.4 | 4.4/8.7/10.1/17.2 | 2351 | 0 |
-| | after | 3.5/8.4/26.5/84.4 | 3.2/7.3/14.2/61.4 | 16.5/16.8/17.4/18.4 | 2279 | 0 |
-| oneline.json | before | 7.2/16.2/45.4/92.1 | 6.0/13.0/27.4/76.5 | 4.5/9.1/45.4/45.4 | 4 | 1 |
-| | after | 7.8/14.1/180.6/189.0 | 6.5/13.1/78.9/188.1 | 16.6/16.8/16.9/16.9 | 4 | 0 |
-| CHANGELOG.md | before | 10.7/58.2/72.2/72.5 | 4.7/12.8/66.5/72.5 | 4.4/8.9/58.8/72.5 | 715 | 7 |
-| | after | 7.2/60.5/74.0/75.3 | 4.7/17.0/70.5/75.1 | 16.6/18.1/73.5/75.4 | 713 | 31 |
-| cascade.html | before | 8.5/17.8/25.9/119.6 | 4.0/13.8/19.6/83.9 | 4.2/8.8/10.6/11.3 | 643 | 0 |
-| | after | 3.9/8.6/19.7/72.3 | 3.3/7.0/18.1/36.4 | 16.5/16.8/18.8/22.8 | 642 | 0 |
-| simdutf.cpp | before | 6.4/29.0/58.8/65.4 | 3.3/11.2/52.7/64.7 | 4.2/8.6/9.6/12.8 | 9852 | 0 |
-| | after | 4.0/12.9/113.9/115.2 | 3.8/7.9/55.9/114.8 | 16.5/16.8/56.5/115.3 | 9862 | 9 |
-| parser.ts | before | 8.6/17.7/25.6/27.0 | 3.7/13.4/19.8/26.0 | 4.2/8.8/10.4/13.8 | 1776 | 0 |
-| | after | 3.7/8.8/21.8/22.2 | 3.1/7.0/16.2/21.1 | 16.5/16.9/18.9/22.6 | 1775 | 0 |
-| fcnal-test.sh | before | 8.1/17.7/27.8/30.8 | 4.0/11.9/21.7/27.1 | 4.2/8.8/11.1/14.0 | 1034 | 0 |
-| | after | 4.0/8.7/18.6/82.3 | 3.5/7.3/14.4/62.3 | 16.6/16.8/18.1/23.2 | 1034 | 0 |
-| data.go | before | 8.6/18.7/24.3/42.6 | 3.8/13.6/21.5/40.0 | 4.2/8.8/12.8/19.1 | 3897 | 0 |
-| | after | 4.2/9.7/19.1/42.0 | 3.4/7.4/17.0/39.4 | 16.5/16.9/22.8/24.1 | 3898 | 0 |
-| guide.typ | before | 7.8/31.1/43.8/52.2 | 4.0/11.4/37.3/50.9 | 4.4/9.1/33.2/37.4 | 117 | 6 |
-| | after | 4.9/24.8/46.6/54.8 | 3.9/11.3/37.1/53.5 | 16.5/16.8/40.5/46.7 | 117 | 21 |
-| unicode.txt | before | 8.7/19.2/92.4/435.3 | 4.2/10.3/17.4/146.9 | 4.4/8.9/10.2/11.5 | 4510 | 0 |
-| | after | 4.0/9.2/92.9/438.8 | 3.2/6.8/16.3/146.3 | 16.5/16.8/17.7/18.2 | 4505 | 0 |
-| big.log | before | 8.3/17.6/61.1/188.2 | 4.2/12.7/21.5/25.7 | 4.3/8.7/9.9/14.9 | 69103 | 0 |
-| | after | 6.6/14.8/73.4/201.2 | 4.8/11.2/15.8/24.4 | 16.5/17.0/24.5/28.9 | 69008 | 3 |
-| words.txt | before | 7.9/15.7/22.7/27.6 | 3.3/11.4/16.4/17.7 | 4.1/8.7/10.2/16.5 | 92552 | 0 |
-| | after | 3.3/8.5/25.1/47.8 | 3.0/7.1/15.1/17.6 | 16.5/16.8/17.7/19.0 | 92274 | 0 |
-| hugelines.txt | before | 8.9/15.3/34.3/1973.2 | 7.0/13.3/16.5/1963.1 | 5.3/16.7/25.9/25.9 | 12 | 1 |
-| | after | 7.9/15.8/33.0/96.5 | 6.2/13.8/17.4/90.4 | 16.7/18.9/22.1/22.1 | 12 | 0 |
-| minified.js! | before | 9.1/21.0/43.3/1875.3 | 5.8/15.0/24.5/1852.3 | 5.0/11.7/17.4/17.4 | 0 | 0 |
-| | after | 8.4/21.5/46.2/87.9 | 5.0/14.6/21.4/64.6 | 16.6/16.8/17.0/17.0 | 0 | 0 |
-| oneline.json! | before | 7.8/18.2/47.5/48.0 | 6.6/12.7/16.8/48.0 | 7.5/9.3/48.1/48.1 | 0 | 1 |
-| | after | 7.6/17.1/72.2/193.8 | 7.0/12.9/21.4/182.3 | 16.6/16.8/22.5/22.5 | 0 | 0 |
-| hugelines.txt! | before | 8.1/17.8/33.0/41.8 | 6.1/11.4/16.6/18.4 | 6.3/10.2/11.8/12.7 | 2 | 0 |
-| | after | 6.6/18.7/32.3/46.7 | 5.6/11.3/14.9/22.2 | 16.6/16.9/23.9/26.1 | 2 | 1 |
-
-#### GUI Debug: `panes`
-
-| scenario | | latency ms p50/p90/p99/max | frame work ms | present gap ms (moving) | rows/s | late |
-|---|---|---|---|---|---|---|
-| command-cat | before | 8.9/19.7/34.8/35.0 | 4.6/9.1/15.7/27.5 | 7.0/37.5/58.1/58.1 | 4509 | 5 |
-| | after | 7.8/19.3/36.5/45.4 | 4.1/9.2/17.9/30.8 | 16.6/37.9/53.8/53.8 | 4501 | 6 |
-| command-yes | before | 8.0/17.0/27.5/34.5 | 6.6/14.5/17.5/24.1 | 4.4/34.9/44.9/44.9 | 4616 | 7 |
-| | after | 7.6/15.7/34.5/35.0 | 6.4/13.6/17.8/22.8 | 16.6/34.5/58.8/58.8 | 4545 | 7 |
-| pager | before | 7.5/14.6/18.6/20.2 | 6.5/13.7/17.1/18.0 | 4.3/8.7/46.2/46.2 | 0 | 1 |
-| | after | 8.6/16.2/19.7/19.8 | 7.6/14.8/19.0/19.2 | 16.6/16.9/47.3/47.3 | 0 | 1 |
-| directory | before | 7.6/15.3/18.7/19.2 | 5.0/12.6/17.7/18.6 | 4.7/9.2/14.5/16.6 | 2 | 0 |
-| | after | 5.4/14.4/18.6/24.6 | 3.9/12.9/17.3/21.6 | 16.6/16.8/17.1/17.1 | 2 | 0 |
-| pdf | before | 7.7/57.3/138.0/190.3 | 6.7/13.8/113.8/174.4 | 4.3/8.9/13.5/13.5 | 0 | 0 |
-| | after | 8.0/59.7/155.2/161.8 | 8.1/41.0/134.9/161.5 | 16.6/16.8/16.9/16.9 | 0 | 0 |
-| image | before | 7.5/16.5/128.7/304.3 | 6.8/15.2/18.7/301.1 | 5.4/9.1/10.4/10.4 | 0 | 0 |
-| | after | 7.9/14.2/17.2/18.3 | 6.8/13.9/17.0/133.0 | 16.7/17.0/146.3/146.3 | 0 | 1 |
-| help | before | 7.8/14.4/18.8/19.8 | 5.2/12.4/17.3/18.2 | 4.7/8.8/12.2/16.3 | 69 | 0 |
-| | after | 3.3/12.7/17.2/17.4 | 3.0/7.9/16.1/17.2 | 16.5/16.8/17.5/18.7 | 69 | 0 |
-| changelog | before | 8.1/16.7/19.3/20.2 | 3.8/12.9/18.2/19.5 | 4.3/8.9/10.6/14.2 | 220 | 0 |
-| | after | 3.9/8.7/16.2/25.1 | 3.7/7.7/14.6/23.7 | 16.6/16.8/17.4/22.5 | 220 | 0 |
-| query-typing | before | 13.1/16.8/41.6/41.6 | 6.5/12.4/38.2/102.2 | 7.6/17.3/27.9/108.3 | 0 | 2 |
-| | after | 13.4/50.3/266.2/266.2 | 3.1/13.0/262.6/262.6 | 16.6/17.0/62.1/62.1 | 17 | 2 |
-
-#### GUI Debug: `longlines`
-
-| scenario | | latency ms p50/p90/p99/max | frame work ms | present gap ms (moving) | rows/s | late |
-|---|---|---|---|---|---|---|
-| w-held-minified.js-wrap | before | 3.6/8.6/19.8/19.8 | 3.3/8.2/9.1/19.0 | 4.2/9.0/9.7/10.4 | 0 | 0 |
-| | after | 3.5/7.4/12.3/12.3 | 3.2/7.2/10.8/11.6 | 16.6/16.9/17.4/17.4 | 0 | 0 |
-| e-held-minified.js-wrap | before | 3.7/11.5/15.0/15.0 | 3.3/8.1/13.9/14.4 | 4.2/9.0/10.3/11.4 | 0 | 0 |
-| | after | 3.3/7.3/8.9/8.9 | 3.1/6.9/8.5/8.8 | 16.6/16.8/17.0/17.1 | 0 | 0 |
-| l-held-minified.js-wrap | before | 8.1/17.0/19.1/19.1 | 4.6/10.3/17.6/18.4 | 4.7/9.2/11.6/13.4 | 0 | 0 |
-| | after | 3.2/7.4/8.6/8.6 | 3.1/7.2/8.5/8.5 | 16.6/16.7/17.5/17.7 | 0 | 0 |
-| dollar-zero-minified.js-wrap | before | 7.3/7.4/7.4/7.4 | 6.5/6.6/6.6/6.6 | - | 0 | 0 |
-| | after | 14.6/14.7/14.7/14.7 | 2.8/9.0/13.7/13.7 | 16.6/16.8/17.4/17.4 | 0 | 0 |
-| f-t-minified.js-wrap | before | 14.3/16.7/16.9/16.9 | 13.6/16.1/16.2/16.2 | - | 0 | 0 |
-| | after | 14.4/20.7/20.7/20.7 | 3.0/13.8/20.1/20.1 | 16.4/16.8/17.6/17.7 | 0 | 0 |
-| gl-gh-minified.js-wrap | before | 14.0/16.4/16.5/16.5 | 13.3/15.6/15.9/15.9 | - | 0 | 0 |
-| | after | 13.8/20.2/20.4/20.4 | 3.2/13.1/19.7/19.7 | 16.5/16.7/17.3/17.3 | 0 | 0 |
-| insert-end-minified.js-wrap | before | 14.9/18.5/20.1/20.1 | 12.1/17.1/19.5/19.5 | 7.1/9.8/10.9/10.9 | 0 | 0 |
-| | after | 3.3/8.6/14.6/14.6 | 3.1/8.3/13.6/13.8 | 16.5/16.8/17.3/18.2 | 0 | 0 |
-| insert-middle-minified.js-wrap | before | 14.1/1994.4/1994.4/1994.4 | 6.9/13.5/1963.0/1963.0 | 6.7/11.1/11.5/11.5 | 0 | 0 |
-| | after | 82.9/83.5/95.1/95.1 | 79.0/81.3/84.2/84.2 | 16.4/16.8/17.0/17.0 | 0 | 0 |
-| sweep-L-W-minified.js-wrap | before | 13.0/19.3/1858.0/1858.0 | 3.6/12.4/18.7/1848.9 | 4.4/8.7/10.0/10.0 | 0 | 0 |
-| | after | 13.4/17.0/71.2/71.2 | 12.4/15.5/60.7/84.4 | 16.6/16.8/84.5/84.5 | 0 | 1 |
-| w-held-oneline.json-wrap | before | 3.8/8.3/9.3/9.3 | 3.3/7.8/8.6/9.1 | 4.1/8.8/9.3/9.8 | 0 | 0 |
-| | after | 3.7/8.1/19.4/19.4 | 3.4/7.4/12.6/18.4 | 16.6/16.9/17.5/22.5 | 0 | 0 |
-| e-held-oneline.json-wrap | before | 3.7/8.1/11.7/11.7 | 3.3/7.8/8.4/11.3 | 4.2/8.8/9.3/9.8 | 0 | 0 |
-| | after | 3.6/7.2/9.1/9.1 | 3.4/7.0/8.5/8.9 | 16.6/16.9/17.1/17.9 | 0 | 0 |
-| l-held-oneline.json-wrap | before | 8.1/18.0/21.3/21.3 | 3.7/8.6/18.9/20.5 | 4.3/8.9/10.8/13.0 | 0 | 0 |
-| | after | 3.7/7.9/9.2/9.2 | 3.4/7.4/9.0/9.0 | 16.7/16.9/17.6/22.5 | 0 | 0 |
-| dollar-zero-oneline.json-wrap | before | 7.6/12.9/12.9/12.9 | 6.8/12.0/12.0/12.0 | - | 0 | 0 |
-| | after | 18.1/18.7/18.7/18.7 | 2.9/16.3/17.7/17.7 | 16.7/16.8/16.9/16.9 | 0 | 0 |
-| f-t-oneline.json-wrap | before | 14.2/17.3/17.6/17.6 | 13.4/16.4/16.7/16.7 | - | 0 | 0 |
-| | after | 18.8/20.1/20.3/20.3 | 5.1/18.0/19.4/19.5 | 15.5/16.8/16.9/17.0 | 0 | 0 |
-| gl-gh-oneline.json-wrap | before | 13.8/16.8/17.0/17.0 | 12.9/15.9/16.1/16.1 | - | 0 | 0 |
-| | after | 19.8/20.4/20.6/20.6 | 6.7/18.1/19.7/19.7 | 15.4/16.8/17.2/17.2 | 0 | 0 |
-| insert-end-oneline.json-wrap | before | 15.4/18.1/21.5/21.5 | 13.4/17.2/20.7/20.7 | 8.6/14.1/20.7/20.7 | 0 | 0 |
-| | after | 3.3/8.2/20.3/20.3 | 3.1/8.1/17.5/19.4 | 16.6/16.8/18.0/18.2 | 0 | 0 |
-| insert-middle-oneline.json-wrap | before | 15.1/1887.6/1887.6/1887.6 | 7.1/18.2/1844.0/1844.0 | 7.8/14.6/17.4/17.4 | 0 | 0 |
-| | after | 82.1/177.2/182.1/182.1 | 14.4/171.6/176.3/176.3 | 16.6/16.8/16.8/16.8 | 0 | 0 |
-| sweep-L-W-oneline.json-wrap | before | 15.4/23.1/49.8/49.8 | 3.5/11.2/18.3/19.6 | 4.4/8.6/10.5/10.6 | 6 | 0 |
-| | after | 13.9/18.3/167.2/167.2 | 10.2/15.2/19.4/167.0 | 16.6/16.8/167.2/167.2 | 5 | 1 |
-| w-held-hugelines.txt-wrap | before | 3.4/8.6/21.4/21.4 | 3.3/8.2/9.1/20.3 | 4.3/9.0/10.2/13.5 | 0 | 0 |
-| | after | 3.3/7.7/11.9/11.9 | 3.2/7.4/10.8/11.3 | 16.6/16.8/17.2/17.8 | 0 | 0 |
-| e-held-hugelines.txt-wrap | before | 8.0/14.5/20.0/20.0 | 3.8/9.3/18.5/18.9 | 4.4/9.2/10.8/11.6 | 0 | 0 |
-| | after | 6.0/17.2/21.2/21.2 | 4.2/14.8/19.9/19.9 | 16.6/16.8/17.0/17.0 | 0 | 0 |
-| l-held-hugelines.txt-wrap | before | 14.4/18.8/20.4/20.4 | 13.4/17.8/19.4/19.4 | - | 0 | 0 |
-| | after | 14.0/16.8/20.8/20.8 | 13.0/15.7/19.7/19.7 | - | 0 | 0 |
-| dollar-zero-hugelines.txt-wrap | before | 15.7/16.6/16.6/16.6 | 14.5/15.5/15.5/15.5 | - | 0 | 0 |
-| | after | 18.7/19.3/19.3/19.3 | 7.1/17.6/18.2/18.2 | 16.5/16.8/16.8/16.8 | 0 | 0 |
-| f-t-hugelines.txt-wrap | before | 16.1/16.4/16.8/16.8 | 15.1/15.3/15.7/15.7 | - | 0 | 0 |
-| | after | 20.1/20.3/22.6/22.6 | 4.8/19.1/20.5/21.6 | 14.7/16.8/16.9/17.3 | 0 | 0 |
-| gl-gh-hugelines.txt-wrap | before | 13.9/16.9/16.9/16.9 | 12.9/15.7/15.8/15.8 | - | 0 | 0 |
-| | after | 18.8/19.6/20.0/20.0 | 6.8/15.4/19.0/19.0 | 15.1/16.8/17.3/17.3 | 0 | 0 |
-| insert-end-hugelines.txt-wrap | before | 15.4/19.2/21.1/21.1 | 14.4/18.2/20.1/20.1 | - | 0 | 0 |
-| | after | 3.3/8.1/18.4/18.4 | 3.2/7.6/10.2/17.3 | 16.6/16.8/16.9/17.3 | 0 | 0 |
-| insert-middle-hugelines.txt-wrap | before | 17.6/21.3/26.2/26.2 | 13.0/16.9/20.2/20.2 | - | 0 | 0 |
-| | after | 16.7/21.7/24.6/24.6 | 12.2/16.5/19.2/19.2 | - | 0 | 0 |
-| sweep-L-W-hugelines.txt-wrap | before | 14.3/17.8/32.3/32.3 | 8.5/16.1/20.0/26.1 | 7.4/16.7/18.3/18.3 | 0 | 0 |
-| | after | 14.0/18.6/30.0/30.0 | 11.4/15.1/19.1/30.0 | 16.7/16.8/17.0/17.0 | 0 | 0 |
-| w-held-minified.js-nowrap | before | 16.8/32.8/34.0/34.0 | 15.6/15.8/27.7/27.7 | 16.7/16.9/33.3/33.3 | 0 | 5 |
-| | after | 14.3/29.1/30.6/30.6 | 9.5/13.5/20.8/20.8 | 16.6/17.1/32.3/32.3 | 0 | 4 |
-| e-held-minified.js-nowrap | before | 28.0/31.1/33.5/33.5 | 11.6/14.4/17.1/17.1 | - | 0 | 0 |
-| | after | 25.9/30.0/32.7/32.7 | 9.5/13.7/16.4/16.4 | - | 0 | 0 |
-| l-held-minified.js-nowrap | before | 27.5/31.5/33.0/33.0 | 11.0/15.2/16.4/16.4 | - | 0 | 0 |
-| | after | 25.3/29.3/30.6/30.6 | 8.9/13.0/14.3/14.3 | - | 0 | 0 |
-| dollar-zero-minified.js-nowrap | before | 24.5/25.0/25.0/25.0 | 23.4/23.8/23.8/23.8 | - | 0 | 0 |
-| | after | 23.3/24.4/24.4/24.4 | 8.5/13.3/23.2/23.2 | 16.6/17.0/18.0/18.0 | 0 | 0 |
-| f-t-minified.js-nowrap | before | 29.4/35.6/36.6/36.6 | 28.4/34.6/36.2/36.2 | - | 0 | 0 |
-| | after | 31.3/31.5/31.6/31.6 | 8.6/30.2/30.5/30.6 | 16.5/16.8/17.9/18.1 | 0 | 0 |
-| gl-gh-minified.js-nowrap | before | 24.5/35.6/35.7/35.7 | 23.5/34.5/34.6/34.6 | - | 0 | 0 |
-| | after | 21.9/31.0/31.8/31.8 | 9.2/20.8/30.8/30.8 | 16.6/16.7/17.4/17.4 | 0 | 0 |
-| insert-end-minified.js-nowrap | before | 19.4/25.7/35.6/35.6 | 18.4/24.3/34.5/34.5 | 16.7/17.9/17.9/17.9 | 0 | 0 |
-| | after | 8.8/14.5/23.9/23.9 | 8.6/12.7/22.5/22.8 | 16.6/16.8/22.4/22.6 | 0 | 0 |
-| insert-middle-minified.js-nowrap | before | 16.6/1994.8/1994.8/1994.8 | 13.9/18.2/1955.8/1955.8 | 15.3/18.5/30.3/30.3 | 0 | 1 |
-| | after | 166.7/251.2/253.8/253.8 | 8.6/153.7/166.3/166.3 | 16.5/16.8/18.0/18.0 | 0 | 0 |
-| sweep-L-W-minified.js-nowrap | before | 24.7/2780.7/2780.7/2780.7 | 15.6/16.3/1963.0/1963.0 | 16.7/16.8/16.8/16.8 | 0 | 0 |
-| | after | 16.4/26.3/84.3/84.3 | 11.0/14.2/18.3/77.0 | 16.6/16.9/17.1/17.1 | 1 | 0 |
-| w-held-oneline.json-nowrap | before | 32.7/33.7/34.4/34.4 | 15.3/15.7/24.8/24.8 | 16.7/17.8/33.9/33.9 | 0 | 5 |
-| | after | 32.8/33.5/34.7/34.7 | 15.1/15.4/22.0/22.0 | 16.7/18.0/33.4/33.4 | 0 | 5 |
-| e-held-oneline.json-nowrap | before | 33.3/33.7/51.7/51.7 | 13.7/14.0/14.9/14.9 | - | 0 | 0 |
-| | after | 33.3/33.5/52.7/52.7 | 13.7/13.9/14.0/14.0 | - | 0 | 0 |
-| l-held-oneline.json-nowrap | before | 33.3/33.5/52.6/52.6 | 13.4/13.7/14.9/14.9 | - | 0 | 0 |
-| | after | 33.3/33.5/51.0/51.0 | 13.7/13.9/14.9/14.9 | - | 0 | 0 |
-| dollar-zero-oneline.json-nowrap | before | 26.1/33.4/33.4/33.4 | 17.3/25.3/25.3/25.3 | - | 0 | 0 |
-| | after | 30.5/33.3/33.3/33.3 | 16.6/22.0/30.7/30.7 | 16.7/16.8/17.2/17.2 | 0 | 0 |
-| f-t-oneline.json-nowrap | before | 27.2/35.5/39.1/39.1 | 26.2/34.5/38.1/38.1 | - | 0 | 0 |
-| | after | 30.8/31.0/31.2/31.2 | 16.5/29.6/30.3/30.3 | 16.7/16.8/17.1/17.4 | 0 | 0 |
-| gl-gh-oneline.json-nowrap | before | 25.8/27.6/35.3/35.3 | 24.8/27.0/34.2/34.2 | - | 0 | 0 |
-| | after | 31.7/32.6/32.7/32.7 | 15.8/30.5/31.6/31.6 | 16.6/16.8/16.8/16.8 | 0 | 0 |
-| insert-end-oneline.json-nowrap | before | 21.8/26.4/35.2/35.2 | 20.5/25.0/34.2/34.2 | 16.6/16.8/16.8/16.8 | 0 | 0 |
-| | after | 16.6/18.2/32.8/32.8 | 16.4/16.6/31.3/31.7 | 16.7/16.9/18.2/19.0 | 0 | 0 |
-| insert-middle-oneline.json-nowrap | before | 16.7/1891.8/1891.8/1891.8 | 15.6/16.7/1853.1/1853.1 | 16.7/16.9/34.5/34.5 | 0 | 1 |
-| | after | 124.1/318.2/319.3/319.3 | 16.4/179.1/180.3/180.3 | 16.7/16.7/16.9/16.9 | 0 | 0 |
-| sweep-L-W-oneline.json-nowrap | before | 16.8/2801.3/2801.3/2801.3 | 15.6/15.8/16.7/1844.0 | 16.7/16.8/17.0/17.0 | 0 | 0 |
-| | after | 16.4/33.2/70.5/70.5 | 13.4/16.8/26.4/31.8 | 16.6/16.9/17.5/17.5 | 0 | 0 |
-| w-held-hugelines.txt-nowrap | before | 22.3/27.4/27.8/27.8 | 15.3/16.0/16.8/16.8 | 16.8/25.6/27.8/27.9 | 0 | 11 |
-| | after | 22.4/26.9/31.7/31.7 | 15.1/16.0/27.1/27.1 | 16.7/22.7/27.6/27.6 | 0 | 3 |
-| e-held-hugelines.txt-nowrap | before | 22.5/27.5/28.8/28.8 | 11.4/16.3/17.0/17.0 | 22.3/27.4/28.8/28.8 | 0 | 17 |
-| | after | 22.4/26.7/28.1/28.1 | 11.4/15.2/17.0/17.0 | - | 0 | 0 |
-| l-held-hugelines.txt-nowrap | before | 22.0/27.4/28.1/28.1 | 10.8/15.8/16.7/17.0 | 21.8/24.2/28.1/28.1 | 0 | 16 |
-| | after | 22.1/26.4/27.6/27.6 | 10.9/15.3/16.4/16.4 | - | 0 | 0 |
-| dollar-zero-hugelines.txt-nowrap | before | 16.6/16.7/16.7/16.7 | 15.6/15.8/16.5/20.6 | 16.7/16.8/17.3/21.7 | 0 | 0 |
-| | after | 15.6/30.4/30.4/30.4 | 13.9/16.1/29.2/29.2 | 16.6/16.8/16.9/16.9 | 0 | 0 |
-| f-t-hugelines.txt-nowrap | before | 16.6/16.8/16.9/16.9 | 15.6/16.0/16.6/16.7 | 16.7/16.8/17.0/17.5 | 0 | 0 |
-| | after | 30.7/30.9/31.4/31.4 | 14.8/29.6/29.9/30.4 | 16.6/16.8/22.3/22.6 | 0 | 0 |
-| gl-gh-hugelines.txt-nowrap | before | 10.9/16.7/16.8/16.8 | 15.6/15.8/16.9/20.7 | 16.6/16.8/18.0/21.7 | 0 | 0 |
-| | after | 16.8/31.4/31.6/31.6 | 15.0/16.7/30.6/30.6 | 16.6/16.8/22.7/22.7 | 0 | 0 |
-| insert-end-hugelines.txt-nowrap | before | 16.5/16.8/17.1/17.1 | 15.6/16.4/16.8/16.9 | 16.6/16.8/17.1/17.9 | 0 | 0 |
-| | after | 16.5/18.3/31.8/31.8 | 16.3/16.6/29.8/30.8 | 16.6/16.9/22.7/22.7 | 0 | 0 |
-| insert-middle-hugelines.txt-nowrap | before | 26.2/29.0/29.4/29.4 | 11.6/15.7/16.2/16.4 | 16.6/26.9/29.2/29.4 | 0 | 40 |
-| | after | 24.4/29.0/37.4/37.4 | 10.7/16.3/16.9/16.9 | 16.6/17.1/22.3/22.3 | 0 | 0 |
-| sweep-L-W-hugelines.txt-nowrap | before | 16.2/25.8/30.4/30.4 | 12.6/15.7/16.5/18.3 | 16.2/25.6/30.0/30.5 | 0 | 30 |
-| | after | 21.0/25.8/30.4/30.4 | 14.1/18.1/29.3/29.7 | 16.6/16.8/16.9/16.9 | 0 | 0 |
-
-#### tty ReleaseSafe: `scroll`
-
-| scenario | | latency ms p50/p90/p99/max | frame work ms | present gap ms (moving) | rows/s | late | bytes/frame p50/max |
-|---|---|---|---|---|---|---|---|
-| all | before | 2.6/3.0/8.8/8.8 | 0.4/1.4/2.8/8.7 | 0.5/1.9/3.1/16.5 | 1449 | 0 | 105/15264 |
-| | after | 2.4/3.0/8.1/8.1 | 2.2/2.7/7.9/8.0 | 15.8/16.9/18.6/18.6 | 1455 | 0 | 120/15264 |
-
-#### tty ReleaseSafe: `stress`
-
-| scenario | | latency ms p50/p90/p99/max | frame work ms | present gap ms (moving) | rows/s | late | bytes/frame p50/max |
-|---|---|---|---|---|---|---|---|
-| all | before | 2.6/11.4/23.2/26.1 | 1.0/2.7/20.1/26.0 | 2.0/2.6/5.2/16.7 | 9811 | 0 | 174/13383 |
-| | after | 1.3/2.8/5.9/14.6 | 1.1/2.7/4.9/13.1 | 15.5/16.6/17.6/17.6 | 9807 | 0 | 132/13383 |
-
-#### tty ReleaseSafe: `jumps`
-
-| scenario | | latency ms p50/p90/p99/max | frame work ms | present gap ms (moving) | rows/s | late | bytes/frame p50/max |
-|---|---|---|---|---|---|---|---|
-| all | before | 1.6/6.0/22.0/22.7 | 0.4/0.8/2.8/22.7 | 0.4/1.5/2.8/16.7 | 2190 | 0 | 108/15264 |
-| | after | 1.6/3.6/15.8/17.4 | 1.1/2.6/7.3/17.3 | 15.8/16.8/18.4/19.1 | 2189 | 0 | 100/15264 |
-
-#### tty Debug: `scroll`
-
-| scenario | | latency ms p50/p90/p99/max | frame work ms | present gap ms (moving) | rows/s | late | bytes/frame p50/max |
-|---|---|---|---|---|---|---|---|
-| all | before | 9.8/14.8/60.9/60.9 | 1.9/5.4/14.7/45.2 | 2.0/2.0/13.3/24.9 | 1452 | 0 | 86/15264 |
-| | after | 7.3/14.1/36.2/36.2 | 5.6/11.8/25.2/27.9 | 15.5/19.8/20.8/20.8 | 1454 | 0 | 128/15264 |
-
-#### tty Debug: `stress`
-
-| scenario | | latency ms p50/p90/p99/max | frame work ms | present gap ms (moving) | rows/s | late | bytes/frame p50/max |
-|---|---|---|---|---|---|---|---|
-| all | before | 3.5/29.0/73.5/103.4 | 2.6/13.2/58.9/103.0 | 3.0/3.2/18.7/24.0 | 9825 | 0 | 180/13383 |
-| | after | 2.4/12.8/69.9/195.7 | 2.3/11.0/66.3/195.5 | 15.6/19.2/24.9/24.9 | 9825 | 0 | 6859/13383 |
-
-#### tty Debug: `jumps`
-
-| scenario | | latency ms p50/p90/p99/max | frame work ms | present gap ms (moving) | rows/s | late | bytes/frame p50/max |
-|---|---|---|---|---|---|---|---|
-| all | before | 26.1/43.9/74.5/78.0 | 2.1/22.9/44.8/77.5 | 2.1/2.3/15.7/40.8 | 2189 | 1 | 108/15264 |
-| | after | 24.5/37.8/63.1/70.3 | 4.9/27.5/42.9/68.7 | 15.7/18.5/24.2/39.9 | 2175 | 1 | 101/15264 |
-
-#### GUI: `pdf-scroll` (anim 11), before = anim 10 with the scenario, after = anim 11
-
-ReleaseSafe, run 1 of 2 (`.scratch/runs9/v1-*`; run 2 within a few ms but far-jump max):
-
-| scenario | | latency ms p50/p90/p99/max | frame work ms | present gap ms (moving) | rows/s | late |
-|---|---|---|---|---|---|---|
-| spin | before | 4.7/9.2/33.4/33.4 | 4.7/9.2/33.4/33.4 | - | 1073 | 0 |
-| | after | 2.0/6.8/22.0/22.0 | 1.7/6.9/17.2/22.0 | 12.9/16.7/21.8/28.4 | 1009 | 1 |
-| held-pagedown | before | 15.3/29.4/31.1/31.1 | 15.2/29.3/31.0/31.0 | - | 9092 | 0 |
-| | after | 1.8/19.2/26.8/26.8 | 1.7/9.6/26.7/26.7 | 16.6/20.5/33.9/33.9 | 6359 | 5 |
-| far-jump | before | 16.5/32.8/32.8/32.8 | 6.8/18.0/32.7/32.7 | 17.0/17.2/19.1/19.1 | 24126 | 0 |
-| | after | 22.6/32.1/32.1/32.1 | 5.8/7.1/32.0/32.0 | 16.6/16.8/18.8/18.8 | 23744 | 0 |
-| slow-notches | before | 7.1/7.1/7.1/7.1 | 8.8/9.8/9.8/9.8 | 16.8/17.4/17.4/17.4 | 83 | 0 |
-| | after | 6.9/7.0/7.0/7.0 | 1.7/6.7/6.9/6.9 | 16.6/16.8/17.3/17.3 | 87 | 0 |
-| slow-touch | before | 6.9/10.7/12.3/12.3 | 6.8/10.6/12.2/12.2 | - | 94 | 0 |
-| | after | 6.8/11.0/12.4/12.5 | 6.7/10.9/12.4/12.4 | - | 94 | 0 |
-
-Debug (`w1-*`):
-
-| scenario | | latency ms p50/p90/p99/max | frame work ms | present gap ms (moving) | rows/s | late |
-|---|---|---|---|---|---|---|
-| spin | before | 4.8/12.7/148.2/148.2 | 4.7/12.3/148.1/148.1 | - | 1074 | 0 |
-| | after | 3.2/7.7/140.4/140.4 | 2.7/7.6/71.8/140.3 | 16.4/16.8/72.2/141.1 | 995 | 2 |
-| held-pagedown | before | 78.5/115.5/115.5/115.5 | 78.3/115.3/115.3/115.3 | - | 8296 | 0 |
-| | after | 10.1/121.7/127.6/127.6 | 6.5/98.4/127.5/127.5 | 16.6/98.6/141.2/141.2 | 6106 | 4 |
-| far-jump | before | 45.4/84.2/84.2/84.2 | 6.2/45.1/83.9/83.9 | 16.6/17.7/18.7/18.7 | 24172 | 0 |
-| | after | 51.4/86.9/86.9/86.9 | 6.1/8.5/86.7/86.7 | 16.6/17.5/20.8/20.8 | 23525 | 0 |
-| slow-notches | before | 11.4/13.1/13.1/13.1 | 7.5/12.6/12.6/12.6 | 15.6/19.2/19.2/19.2 | 83 | 0 |
-| | after | 18.3/19.0/19.0/19.0 | 2.7/11.9/18.5/18.5 | 16.6/16.9/19.2/19.2 | 88 | 0 |
-| slow-touch | before | 7.3/11.8/14.1/17.8 | 7.0/11.5/13.8/17.3 | - | 94 | 0 |
-| | after | 7.2/11.0/12.5/15.4 | 7.0/10.5/12.3/15.1 | - | 94 | 0 |
-
-The `panes` timedemo's pdf phase (`u1-*`, RS then Debug):
-
-| scenario | | latency ms p50/p90/p99/max | frame work ms | present gap ms (moving) | rows/s | late |
-|---|---|---|---|---|---|---|
-| pdf | before | 6.9/14.8/44.1/46.3 | 6.7/14.5/30.6/44.0 | 16.7/16.8/16.9/16.9 | 4678 | 0 |
-| | after | 2.0/7.4/34.8/37.8 | 1.8/6.6/18.4/34.7 | 16.6/16.8/27.2/32.9 | 4678 | 9 |
-| pdf | before | 7.6/57.1/149.4/158.8 | 6.3/18.0/142.1/149.3 | 16.6/16.8/17.9/17.9 | 4663 | 0 |
-| | after | 3.2/16.2/144.5/155.5 | 2.8/7.2/125.0/144.4 | 16.5/16.8/134.0/149.4 | 4655 | 7 |
-
-Strip steps between presented frames (px, the probe's `scroll` column is
-the strip's drawn y for a focused PDF; cell height 25):
-
-| phase | before | after |
-|---|---|---|
-| slow notches (one each 300 ms) | 25 px in one frame a notch, 7 of 7 steps whole lines | p10/p50/p90/max 0.04/0.8/6.8/7.2 px, none a whole line |
-| slow touchpad (0.06 line a frame) | 1.5 px every frame | 1.5 px every frame (followed) |
-| a spin's steps | 25 px each (a notch a frame, 100 at a burst) | 0.5/16/28/42, a stream |
-| settle tails | none (nothing settled) | no rise after the fastest frame; last step <= 0.05 px |
-
-Latency: the spin and held PageDown answer sooner (the frame that answers
-a key no longer renders the pages it moves to before showing it; they
-render as the glide's input frames, a stream's, come). The far jumps
-(ge, gg, a Look at a page, Back) answer within the before's spread in
-ReleaseSafe (p50 16.5/17.8 -> 22.6/14.9 over two runs, max 32.8/31.7 ->
-32.1/31.2); in Debug ge's frame renders what the before split between the
-`g` prefix's frame (~20 ms, re-rendering pages left banded by a fling) and
-ge's own (~40): the same total in one frame (60-67 ms). The `late` frames
-are a held PageDown's input frames rendering a page while it moves (33 ms
-gaps; before they were not moving frames, and took 30 ms to answer).
-
-## The coordinator's latency rows, answered
-
-- **pages-up-big** (p90 8.8 -> 24): real. The paint-ahead's upward stretch
- stopped two lines short of the view's own bottom (it was "two views out",
- the AFTER rows counted from the far end), so a half page up missed it and
- painted on the frame (11-17 ms). The stretch is now the view itself and
- three more the way it goes, both ways; re-measured (two interleaved
- rounds and the final batch): p90 8.8-11.0 against 9.7-19.1 before, frame
- work p90 6.0 against 14.1; the p99 left (15-20 ms) is the first page up
- after `ge`, which paints on the frame in both (the paint-ahead was
- looking down).
-- **notches p90/p99**: the wheel takes the drawn-at-once branch (its SDL
- event sets `input_seen` as a key's does: the nine notches into a running
- glide answer in 1.7-2.0 ms against 2.7-4.7 before). The p90 of ten
- samples is the first notch after 1.2 s idle: 13.3 against 9.8, n=1, the
- CPU and GPU waking in both (its upload 4.7 ms, its fence 2.4).
-- **small-file jumps +2 ms**: not reproducible: two more interleaved rounds
- put after both above and below before (search-n-small 8.6/7.2 against
- 9.1/10.4, rail-small 8.6/7.0 against 8.9/12.1). Per frame, an input after
- idle pays the soft present's upload at a low clock (4.7 ms against 0.8
- warm) in both builds; that is the spread.
-- **insert-big max**: the 200-line paste's frame (a far, stitched slide and
- its paint), 31-37 ms in both builds.
-- **rows/s that changed meaning**: the `scroll` demo's before ran `G` (a
- no-op without a count; the script was fixed to `ge` in anim 3, not in
- 1b, so before and after ran different scripts). Moved into 1b, both
- rebuilt and re-run: the tables below are the same script. `jumplist`'s
- rows/s counts how far the jumps went, which depends on what earlier
- phases left in the jumplist; before's `look-big` phase drew no frames
- (its Looks resolved in the query pane), so its jumplist held fewer far
- places. Compare its latency and work, not its rows/s.
-
-## Open questions and positions
-
-1. **The soft present on Wayland is the ReleaseSafe frame's tail.** Every
- native-Wayland GUI presents through SDL_Renderer so WindowOpacity can be
- background-only, whatever the setting: a GPU fence wait and a 6.7 MB
- upload a frame, and the first frame after idle pays both at a low clock.
- With the GPU swapchain (X11) the scroll is 0.34 ms of work p50. Position:
- claim the GPU swapchain when a session starts at WindowOpacity 100 (the
- author's config) and say "needs a restart" below 100.
-2. **Debug is slow where tree-sitter allocates**: a third of the Debug
- stress run was the DebugAllocator's stack capture per tree-sitter
- allocation. Position: a size-class pool for tree-sitter in safe builds,
- leak-checked by counting.
-3. **The spin** costs a core while something moves (CPU 2.0 -> 7.3 s over
- the 18 s scroll demo, ReleaseSafe). Position: keep; asleep, the clock
- that cannot be trusted costs the latency the author ranked first.
-4. **Terminals and keys**: a terminal glides only under the wheel. Ctrl-d
- in its normal mode could glide too when no output arrived that frame.
- Position: next, with the stream test.
-5. **The far jump's slide** shows the last frame's body sliding off and the
- new one in, not what lies between. Position: keep; it reads as motion
- and never shows anything that was not on screen.
-6. **Long lines**: 16 KB round the caret is painted; the rest of a minified
- line keeps its last colours. Typing in the middle of a 500 KB line is
- now 18-22 ms a key in ReleaseSafe (was 173-213); what is left is the
- text's splice and the wrap of the line. The no-wrap sideways scroll
- snaps (a horizontal glide was not asked). Position: an incremental
- whole-text parse is the real fix; not ponytail-sized.
-7. **PDF first frames are not improved.** The bridge renders on the loop
- through one MuPDF context and a one-page cache (30-38 ms RS, 110-184
- Debug, the page's content stream interpreted). Off the loop needs a
- cloned fz_context per worker and locks round the document: a day's
- work with the band-render design in the latency notes. Pictures are
- done (decoded on a thread: first frame 90 -> 13 ms max RS, 296 ms ->
- one frame later in Debug); a second open still decodes again (no cache:
- a Restore is rare).
-8. **The tty** has no vsync and no sub-cell motion: it is paced (a 16 ms
- tick through its acks), it glides at whole rows (anim 12), and a scroll still rewrites
- most rows (vaxis diffs cells; terminal scroll regions would move them
- for a few bytes). Position: scroll regions, if the tty's bytes matter.
diff --git a/docs/next-pdfrender.md b/docs/next-pdfrender.md
index 4c8565d8..9e2e34b8 100644
--- a/docs/next-pdfrender.md
+++ b/docs/next-pdfrender.md
@@ -1,7 +1,7 @@
# pdfrender: PDF pages rendered off the loop
The author's ask: "add concurrency to the mupdf build so its smoother". The
-open item of `next-anim.md` (7): the bridge rendered on the loop through one
+open item of the animation plan (7): the bridge rendered on the loop through one
MuPDF context and a one-page cache, so a PDF's first frame, a far jump's
page and every page a held PageDown reached cost the frame they landed in.
Later asks from the coordinator, answered here too: the Intel SDM (combined
diff --git a/docs/render-pipeline.md b/docs/render-pipeline.md
index c04ccd33..49180ad3 100644
--- a/docs/render-pipeline.md
+++ b/docs/render-pipeline.md
@@ -1,915 +1,54 @@
-# pardes render pipeline: one frame, one clock, one order
+# pardes render pipeline: the core's frame, then the frontend's own
-Status: DESIGN, FINAL (draft 2 plus the adversary's two convergence fixes; converged with the adversary; the adversary's 17 objections to draft 1 are folded in, each marked "draft-2 fix" where it changed the design). Nothing here is implemented. Line numbers are
-the working copy of 2026-09-28 (`losuylsy`, mid tag→Text refactor; they drift).
+What a frame is and who draws it, as it stands. The effects programme this
+document planned (eleven panel transitions in two tracks, a Shadertoy post
+chain with Crt, Bloom, Vignette and Grain, Lift and its soft shadows, the
+glows, the terminal's chip shadow and thumb flash, the lapis ornament and
+its parallax, the message fall and dissolve, the theme crossfade, five
+Motion flavours with anticipation and squash, tiers and track groups) is
+gone: the core keeps no motion and no clock but its deadlines, and each
+frontend keeps the little it draws on its own.
-Goal, in the user's words: an algorithm for the ordering of rendering, how each
-part is selected and joined, the shader passes after that, and one animation
-model. Consumers: the lapis theme (shader effects on specific parts), multi-line
-tags, the full-screen post pass (now Shadertoy-compatible), images/PDF, and
-animations. Added requirements: a game-quality GUI effects catalogue, a separate
-terminal effects track (audit + new cell-native effects), and a motion / colour /
-performance / "feel" discipline.
+## The core's frame
----------------------------------------------------------------------------------
+- `Pardes.pump` drives the host: `wait_input`, with the core's next
+ deadline (`nextWake`: a message's linger, the look-hover preview, a
+ question's hint, a paused Grep, work that resumes a frame at a time);
+ `now`; `poll_frame`; `render` and `present` only when a frame is owed
+ (`needs_frame`); and `post_present`, whose `Pardes.presented` lets a
+ drag's moving end and a look-hover follow what is under the pointer now.
+- `advance(now)` steps `.tick`s a `frame_ns` (16 ms) at a time, for the
+ work that resumes a frame at a time; a wait is jumped in one step.
+- `render` (draw.zig) paints one `Surface`: the canonical grid (`/screen`,
+ the snapshot goldens, the tty), the regions (each part's rectangle,
+ owner and focus; a body's first line and the text it shows,
+ `Region.line` and `Region.doc`), the layers a pixel shell draws at the
+ tagline's pitch and with compact context rows, the images, the cursor
+ and the palette (`Chrome`). Nothing on it moves.
-## 1. Today's frame, end to end
+## The frontends' own
-### 1.1 Who drives a frame
+Each draws these on its own clock, from the Surface alone, and draws their
+frames itself while they move (the last frame again: the tty's
+`poll_frame`, the window's loop), pacing its own waits:
-`Pardes.pump` (pardes.zig:4607) is the only loop body every shell runs:
-`wait_input(timeout = animationActive ? 16 : 0)` → queued events → `update` →
-effects → `poll_frame` → **skip unless `needs_frame or animationActive()`** →
-`frame_arena.reset` → `render` → `present` → `needs_frame = present_skipped` →
-`post_present`.
-
-Ticks (`Event.tick`) are produced by six different clocks:
-
-| shell | who ticks | cadence |
-|---|---|---|
-| tty | `tickWatch` thread: nanosleep 16 ms **after** the wait asked, then `postEvent(.tick)` (tui.zig:1513) | 16 ms + frame time: drifts slow, worse over ssh |
-| gui | `AnimationClock.due` in `postPresent` re-arms `now+16ms` (gui.zig:203, 221, 4095) | a tick only once ≥16 ms have passed since the re-arm: at 144 Hz every 3rd frame (20.8 ms) → animations ~25% slow; at 120 Hz every 2nd; a slow frame slows them further |
-| gui grid mode | every `poll_frame` while active (gui.zig ~4114) | as fast as the loop |
-| web | JS rAF fixed-step bank, capped (app.mjs:566) | correct average rate, always ticking |
-| macOS | `spendTickTime` bank (macos.zig:868, 1549) | correct average rate |
-| detached server | after a timed `poll` (server.zig:816, 828) | ≥16 ms |
-
-So the same animation runs at different speeds per shell. Everything counts
-16 ms frames (`layout.Animation.frame_ms`, layout.zig:2041).
-
-Acknowledgement is also per shell: tty `postPresent` acks the tracks it drew
-(tui.zig:1187), GUI acks then ticks (`finishPresentedAnimationFrame`), grid mode
-acks `&.{}`, macOS acks via `pardes_frame_presented`, **web never acks**.
-
-### 1.2 Core render (`Pardes.render`, pardes.zig:5950), in code order
-
-1. `File.refreshHighlights`; size the grid; reset cursor/pointer shape, **clear
- every cell's hover bit** (5979 — stale state that exists only because paint
- passes do not own their output), zero layers/images/tracks.
-2. Fill the whole screen with `chrome.border` (5991).
-3. Per pane: `collectNotices` → `renderPane` (6580): `clearRect` + page fill
- ("two full passes over every cell in the pane", 6596) → grip → `paintPaneTag`
- → PDF / image / `renderBody` (+scrollbar) → `renderBodyLayer` (body_layer.zig:243)
- which **renders the whole body again** into a temporary Surface by
- `std.mem.swap(Surface, &p.surface, &temporary)` (body_layer.zig:266-272).
-4. Notice chips painted on the grid (6000-6065).
-5. Workspace tag, column tags, their hover word and header selection/caret on
- the grid (6068-6155).
-6. `renderTagLayers` (6396): **paints every pane tag a second time** into a
- temp Surface via the same swap (6414-6419); builds notice layers (re-deriving
- chip geometry, text, caret the grid pass already computed) and column /
- workspace layers (`renderHeaderLayer` 6520, which re-does hover and header
- selection).
-7. Drag overlays (`overlayDash`, glyph-only rewrites, 6158); the column-move
- rail shrinks layer viewports by one cell so the rail stays visible.
-8. Debug box (6244).
-9. `presentation.submit` (6301): tracks, previous grid, per-cell diffs.
-10. `composeAsciiTransitions` (6305): character effects composed in the core
- into an arena copy of the cells.
-
-Implicit orderings that matter and are written nowhere: body layers before tag
-layers (notice chip over a context band); images after cells, which is why an
-image pane "gives up the rows" under its notices (6629) instead of having z;
-tag layer array index order (panes, columns, workspace, notices) is paint order.
-
-### 1.3 Surface: what the shells get (surface.zig)
-
-`cells` (the canonical grid), `cursor`, `pointer_shape`, `images`
-(`ImagePlace`, with `patch` rows), `body_layers[16]` (compact context rows +
-body rows), `tag_layers[MAX_TAG_LAYERS]` (one row each; the tag refactor just
-added `line` — one layer per tag line), `panel_tracks`, `previous_*` + `cell_diffs`
-while a transition needs them, `mark_hover` (a flag that makes `Surface.at`
-set `cell.hover`, 378-385 — a side channel for macOS glass).
-
-### 1.4 Shells
-
-- **tty** (tui.zig:1100): `panel_compositor.compose` rewrites cells for
- slide/zoom/dissolve/vertical → `paintCells` (win.clear + every cell) → kitty
- images (hidden while geometry moves) → cursor → `vx.render` (vaxis diffs
- against its last screen, wraps in sync 2026). Layers are ignored.
-- **gui** (`renderFrame`, gui.zig:5035): one render pass. `makePaintPlan`
- (2048) = one batch per active track + static batch. Per batch: grid cells not
- covered by a layer (`coverLayers` 5656) → body layers → tag layers → previous
- layers/cells for transitions → that batch's images. Then the overlay
- (`buildOverlay` 7258: pet, topbar rule, column rule, bottom fill, pane chrome
- 7468, grips 7543, bar cursors, touch). Then optional crt pass scene_tex →
- swapchain (5439). Cell = one instance drawing bg + glyph (ui.frag.glsl).
-- **web**: DOM spans for the grid + absolutely positioned divs for tag and body
- layers. No images, tracks, scene; never acks.
-- **detached** (wire.zig, `version = 7`): cells + body/tag layers, delta coded.
- A GUI attached client has `core == null` → no pane chrome (appendPaneChrome
- reads `core.rects`): a per-shell divergence caused by the shell reading core.
-- **macOS** (out of scope): its own copies of tag/body layer ABIs, its own
- tick bank, Metal crt.
-
-### 1.5 Duplicated logic, smells, bugs
-
-1. Paint twice: pane tags (grid + layer), bodies with context rows (grid + layer),
- notices (grid + layer, text/geometry/caret recomputed), workspace/column tags
- and their hover/selection/caret (grid + layer). The two copies already
- disagree: a notice layer highlights the hovered word, the grid chip does not;
- the column-tag grid hover is suppressed during `column_move`, the layer's is not.
-2. `std.mem.swap(Surface, &p.surface, …)` as the way to "paint into something
- else": every paint function writes `p.surface` directly.
-3. Geometry recomputed everywhere: `tag_y = if tag_bottom …` appears in
- renderPane, paintPaneTag, renderBody, renderTagLayers, appendPaneChrome,
- taglineBaseRgb; `bodyTop`; BOX_H == 1 assumed throughout (multi-line tags
- break all of these).
-4. The GUI infers semantics from painted cells: scrollbar thumb from
- `cellBackgroundIs(scroll_thumb)` (7468ff), topbar rule from "row 1 has a
- tagline cell" (5509), bottom band (5519), grips from tagline cells
- (`paneGripCell` 5872), focus tint recomputed from `core.active`
- (`taglineBaseRgb` 5609). Any theme where two roles share a colour breaks it.
-5. Six clocks (1.1), four ack sites, and a web that never acks.
-6. `animationActive` (5884) is true while crt/ripple/glitch are on → the whole
- core render (double paints included) and `capturePrevious` run every 16 ms
- for a time-only uniform.
-7. `acknowledge(&.{})` → `capturePrevious` memcpys all cells and all layer
- cells on **every idle presented frame, even with `panel_transition = .off`**
- (layout.zig:283-290).
-8. A lingering message (`message_linger_ms`) keeps `needs_frame` true every
- tick for its whole linger (Messages.zig:238-250): seconds of full 60 Hz
- renders that change nothing. Look-hover delay likewise ticks every frame.
-9. `Animation.Transition.retarget` (layout.zig:2064) restarts from the displayed
- value with step 0: position continuous, velocity snaps to zero.
-10. Colour fades mix sRGB bytes (`interpolateRgb` layout.zig:2111,
- `Messages.blendRgb`, `colors.mix`): muddy mid-fades.
-11. Shell-side animations each have their own time: crt uniform
- (`SDL_GetTicksNS`), pet (`SDL_GetTicks`), smooth scroll lag (per poll),
- touch flash (per rendered frame), macOS coasting.
-12. `mark_hover` hack and the hover-bit reset (1.2 step 1).
-
----------------------------------------------------------------------------------
-
-## 2. Principles
-
-1. **Core decides WHAT and WHERE; shells decide HOW.** The core emits an ordered
- region list with geometry, z, and resolved styles. No shell reads
- `core.rects`, `core.active`, settings or theme to decide what something is.
-2. **Paint once.** Every part is painted exactly once, into its own target; the
- grid is produced by joining those targets.
-3. **One clock.** Time is the shell's monotonic `now_ns`, passed in; the core
- never reads a clock. Animations are functions of `now - start` (tweens) or
- closed-form springs, never frame counters.
-4. **A fixed, named order**, written as straight-line code in one function per
- side. No pass objects, graphs or vtables.
-5. **Zero cost when off; zero work when idle.** A disabled effect emits nothing.
- Idle means the earliest deadline is "never".
-6. **Decorations never change cell geometry.** Grid, hit testing, 9P cells and
- snapshot goldens are unaffected by any GUI effect.
-
----------------------------------------------------------------------------------
-
-## 3. The frame
-
-### 3.1 Loop (pump, all shells)
-
-```
-pump(h, now_ns):
- wait_input(timeout = min(core.nextWake(now), shell's own next wake))
- drain events → update ; drain effects → perform
- core.advance(now) // settles animations whose time passed; sets
- // needs_frame if something core-visible moved
- poll_frame
- if needs_frame or core.running(now):
- render(now) → present → if present said "shown": core.acknowledge(tracks)
- else if shell has a presentation-only animation:
- shell redraws at level A (post chain only) or B (decor only), §5.5;
- no core render, no cell-instance rebuild
-```
-
-`animationActive: bool` becomes `nextWake(now) ?u64` (null = idle, 0 = run
-every frame, >0 = sleep until). `acknowledge` moves into `pump` (present returns
-whether it showed the frame); the four shell ack sites and GUI's
-`finishPresentedAnimationFrame` go. `capturePrevious` runs only when
-`panel_transition != .off`.
-
-### 3.2 Core `render` (moves to `src/draw.zig`, `pub const render = draw.render`)
-
-Straight-line, in this order:
-
-```
-render(p, arena, now):
- 0. BEGIN size grid; reset regions, layers, images, cursor, tracks
- 1. PLACE compute every part's rect ONCE and append a Region, in z order:
- page; per pane: body, tag (h = tag lines), grip, rail(+thumb),
- notices; per column: tag + grip; workspace tag; drag overlays;
- debug box; cursor
- 2. PAINT for each region: paint its cells once into its target — its
- layer (tags, notices, bodies with compact rows) or the grid
- 3. JOIN copy each layer's grid-visible cells into the grid at its
- viewport, in region order → the canonical grid (tty/9P/wire)
- 4. GRID-ONLY glyph-only overlays (drag dashes) and the debug box on the grid
- 5. PRESENT presentation.submit (tracks, diffs); region tier from role
- and track phase (§3.4)
- 6. STYLE resolve region styles: lift, dim (focus animation), the
- theme's decor per region kind, style_alpha, per-pane view origin
-```
-
-Paint functions take the target `s: *Surface` instead of writing `p.surface`
-(mechanical signature change through body_layer, File, Terminal, Image, Pdf);
-the swap hack and the second tag/body/notice paint are deleted. Where the
-old grid and layer copies disagree, **the grid's behaviour wins** so goldens
-stay byte-identical. Every disagreement is listed in the change and decided
-on its merits later (e.g. the column_move hover suppression is right for both;
-notice-word hover could go either way). JOIN traps: (a) a wide grapheme at the
-viewport's right edge. `print` drops a width-2 glyph that does not fit
-(`col + 1 >= end`), but the layer has columns beyond it, so the copy must
-re-clip at the edge and never leave a head without its spacer. (b) The body
-layer's first `r.h − tag_rows` rows must equal today's grid rows with
-`tag_bottom` both on and off (`first_row`, body_layer.zig ~275).
-
-### 3.3 Which parts render, and from what
-
-- A region is emitted iff its rect intersects the screen and its owner is
- visible (collapsed pane → tag region only).
-- Pixel shells draw a region **from its layer if it has one, else from the grid
- cells in its rect**. The page region draws only grid cells no other region
- covers (today's `coverLayers`, generalised to regions).
-- Cell shells (tty, detached tty) draw the grid; layers exist only for pixel
- shells.
-
-### 3.4 Composition order (GUI): fixed tiers by ROLE
-
-A region's tier comes from its **role** and is fixed for the frame. Animated
-elevation (`Region.lift`, 0..1) only drives shadow offset, σ and strength, so
-nothing changes draw order partway through a lift (draft-2 fix: an animated
-`z` whose floor is the tier makes the draw order pop mid-animation).
-
-| tier | role | overlap inside the tier? |
-|---|---|---|
-| 0 | page, static panes, their chrome (rails, grip marks, rings, rules) | no |
-| 1 | the active pane (only when focus lift is on) | no |
-| 2 | panes moving/opening in a transition | **yes**: split per track, in `paintOrder` |
-| 3 | closing tombstones (frozen previous cells) | **yes**: split per track |
-| 4 | floating: notice chips, debug box, drag preview | no |
-| 5 | true overlays: cursor, touch/pet (debug) | no |
-
-A pane's own chrome (rail, grip marks, rings, rules) is decor of the pane's
-tier and carries its track transform, so it moves with the pane. Today it is
-hidden during transitions (`transient_on`, gui.zig ~7299) because it doesn't
-move. Only real overlays live in tier 5.
-
-Straight-line GUI draw, one render pass:
-
-```
-for tier 0..5:
- for each group in tier // one group per tier, except tiers 2 and 3:
- // one group per track, in paintOrder
- decor-under fills, patterns, rings under content [decor]
- cells bg + glyph; text-shadow regions in 3 phases [ui]
- images the group's images [image]
- cast shadows this group's shadows, darken-only, each with
- its caster's rect as an exclude rect [decor]
- decor-over rings over edges, rails, rules, grip marks [decor]
-post chain (§6) → present
-```
-
-Why cast shadows come AFTER cells (draft-2 fix): static panes share tier 0. If
-the shadow were drawn before the tier's cells, the neighbour's opaque cells
-would paint over a shadow cast onto its gutter or tag row, which is exactly the
-lapis case. Each shadow instance discards inside its caster's rect, so it only
-darkens what lies beside or under it. An elevated tier's shadow onto lower
-tiers works the same way.
-
-Draw calls: ≤ 5 per group. Tiers 0, 1, 4 and 5 are one group each, so ~20 draws.
-Tiers 2 and 3 add ≤ 16 groups, and only while a transition runs. Empty draws are
-skipped. The per-track batching that exists today (PaintPlan) survives only in
-tiers 2 and 3, as groups. The vertical effect's scissor becomes a per-instance
-clip rect. A notice chip (tier 4) ends up above an image, so the image pane can
-stop "giving up the rows" (its own visual-change step).
-
----------------------------------------------------------------------------------
-
-## 4. Data the core hands the shells
-
-Additions to `Surface` (all POD, fixed-size):
-
-```zig
-pub const Region = struct {
- kind: enum(u8) { page, body, tag, grip, rail, column_tag, workspace_tag,
- notice, overlay, debug },
- pane: u8 = none, column: u8 = none, serial: u32 = 0,
- rect: Rect, // grid cells: the hit-test geometry
- tier: u8 = 0, // fixed by role (§3.4)
- lift: f32 = 0, // animated elevation 0..1: shadow only
- dim: f32 = 0, // 0..1 pull toward the page (unfocused)
- layer: u16 = none, // layer index, or draw the grid cells in rect
- style: u8 = 0, // index into Surface.styles (0 = plain)
- style_alpha: f32 = 1, // decor weight; theme crossfade (below)
- active: bool = false,
- thumb: struct { y: u16, h: u16 } = .{}, // rail only
-};
-regions: [MAX_REGIONS]Region, nregions: u16,
-styles: [8]Style, // resolved per frame from theme + settings
-view: [MAX_PANES]struct { serial: u32, top_line: i32, wrap_at: i32 }, // ABSOLUTE per-pane view origin (§5.4)
-```
-
-`Style` holds a theme's decor for one region kind. Hard-edged geometry is in
-**logical pixels** and snapped to whole device pixels by the shell (logical px ×
-display scale, rounded): ring widths, pattern periods, hard-shadow offsets.
-Lapis is pixel art (2/3/1 px rings, 6 px checker, 5-6 px hard shadow). In
-fractional cell units those would blur and change with font size (draft-2 fix).
-Fields: `fill`, `ring[3] {px, rgba}`, `pattern {none|checker|stripes|dots, px,
-duty, a, b, see_through}`, `hard_shadow {dx_px, dy_px, rgba}`, `text_shadow
-{dx_px, dy_px, rgba}`. Soft elevation shadows are not per style. They come from
-`lift` and one global `shadow {rgba, σ_px, offset_px}` per theme.
-
-`see_through`: a region style with an under-pattern (the lapis dot grid under
-pane bodies) marks page-coloured cell backgrounds as clear, using the existing
-ui.frag clear-bg flag (0x40000000), so the pattern shows between glyphs. In a
-tiled layout the page region is almost never visible, so the pattern has to sit
-under the bodies.
-
-Theme crossfade: switching into or out of a decorated theme fades
-`style_alpha` from 0→1 or 1→0 over the same Tween as the colours. Decor never
-snaps while colours fade. Schema: `decor` is an optional field on `Theme`. The
-comptime `fold` and the ZON ThemeFile parser (colors.zig:266) both walk the
-same struct, so theme files get decor with no separate schema. Lapis ships
-compiled in first.
-
-Layers: once the tag refactor lands, `TagLayer` and `BodyLayer` merge into one
-`Layer { kind, id, serial, viewport, cols, rows, compact_rows, cells, cursor
-{x,y,bar}, bg, slide, fade, separators }`. A tag is all compact rows. A body is
-`context_rows` compact rows followed by body rows. A notice is one compact row.
-There is one `rowTop/rowHeight/hitAt`. Multi-line tags are `rows > 1`, which
-replaces the refactor's one-layer-per-line `line` field. This is the only part
-that touches the other agent's files, and it waits for them.
-
-Time: `Surface.now_ns`, the core's time for this frame.
-
-Wire: regions and styles go over the detached wire (`version` 7 → 8). The
-attached GUI then draws chrome from them, which fixes the core==null
-divergence. Web ignores the new data at first. Its lapis could later map styles
-onto CSS box-shadow, borders and gradients, a natural fit since the reference
-IS CSS.
-
-Deleted from the shell once regions exist: `taglineBaseRgb`,
-`topbarPaneBorderHeight`, `bottomTaglinePresent`, the scrollbar inference in
-`appendPaneChrome`, `cellBackgroundIs`, `paneGripCell`'s scan, and `mark_hover`.
-The hover affordance becomes an `overlay` region.
-
----------------------------------------------------------------------------------
-
-## 5. Region effects on the GPU
-
-### 5.1 One decor pipeline, instanced
-
-New `shaders/decor.{vert,frag}.glsl`. Each decoration is one instance:
-`{ rect_px, clip_px, exclude_px, kind, params vec4, color0..2 (premultiplied),
-transition header (the same fields as CellInstance) }`.
-
-- **Hard-edged kinds, no AA, snapped edges**: `solid`, `ring` (up to 3 inset
- rings), `checker`, `stripes`, `dots`, `hard_shadow`. Rules, rails and grip
- marks become `solid` decor emitted from regions. Because they are snapped and
- un-anti-aliased, stage 8's "PPM goldens byte-identical" is achievable.
-- **Soft kinds, fwidth AA, dithered**: `soft_shadow` (a rectangle convolved
- with a Gaussian in erf closed form, as Wallace and GPUI do: one quad grown by
- 3σ, no blur pass), `inner_shadow`, `glow`, and `cursor` (rounded corners).
- Wide gradients get ±0.5/255 interleaved-gradient-noise dither.
-
-The triangle overlay pipeline stays only for the pet and the cursor's glide.
-
-### 5.2 Shadows and gamma
-
-A pane's right and down shadow lands on its neighbour's gutter or tag row
-(chrome), because pane rects tile. With `tag_bottom` it lands on the first body
-row instead, which is simply what a shadow does. §3.4 guarantees it is drawn
-over the neighbour and never over the caster.
-
-Gamma decision (draft-2 fix, one rule instead of two). Decor blends on the
-UNORM target, so blending happens in sRGB space, as CSS does.
-- Dark shadows: alpha is pre-warped so the result matches linear light,
- `a' = 1 − (1 − a)^(1/2.2)`. This is exact for black.
-- Coloured decor (glows, selection halo, tinted shadows) is capped at ≤ 25%
- alpha, where sRGB-space falloff is only slightly dark. That is documented, and
- it is accepted by the feel review, not by maths.
-- Opaque decor (lapis hard shadows, rings) has no gamma issue.
-- Measured experiment in the fx set: render the scene into RGBA16F linear
- (ui.frag writes a linearised `mix(bg,fg,a)`, so glyphs look identical) and add
- one encode pass that also produces the sRGB iChannel0. It is adopted only if
- its cost at 4K on the iGPU fits §8.3.
-
-### 5.3 Text shadow
-
-Only for regions whose style has `text_shadow` (lapis titles). The region's
-cells are emitted in three phases in its group: all bg instances (glyph = the
-space slot), then shadow glyph instances (offset, shadow colour, clear bg), then
-glyph instances (clear bg). Every other region keeps its single bg+glyph
-instance. Within a group, bodies come before tags, so a tag's shadow can hang
-over the body's top edge.
-
-### 5.4 Cursor (shell-owned presentation)
-
-The cursor is decor in tier 5, not a reversed cell. It is a quad whose 4 corners
-follow the target on critically damped springs. The leading corners are stiffer
-than the trailing ones, which gives a smear along the motion that collapses at
-rest.
-- **Distance rule** (draft-2 fix): a move of ≤ 1 cell, or any move in insert
- mode, snaps or settles in ≤ 40 ms. Only jumps glide (90-150 ms), as neovide
- does. At key-repeat rate (~33 ms per cell) a 120 ms spring would trail 3-4
- cells behind and feel sluggish.
-- **Scroll-aware**: the core exposes each pane's ABSOLUTE view origin
- (`view[pane]`: serial, first line, wrap offset). The shell diffs it against
- the origin it last drew and shifts its spring state by the difference in
- pixels, so the cursor does not glide against text that jumped. It is absolute,
- not a per-frame delta, because frames get dropped: the detached server skips
- clients with pending output (server.zig ~712), and a GUI can skip a present.
- G6 smooth scroll uses the same origin.
-- **Legible during the glide**: glyphs under the quad are re-emitted in the
- cursor-text colour, clipped to the quad (per-instance clip), as neovide does.
-- **Geometry**: the pixel rect comes from the layer the cursor sits in (body
- layer compact rows, or tag layer pitch plus band offset), not from `cols × cell_w`.
-- **Blink**: default OFF, as pardes never blinked. When on, it is a square wave
- with an ~80 ms ease at each edge, so it redraws only at the edges, held solid
- while typing and for 500 ms after, and it stops after 10 s idle.
-
-### 5.5 Presentation-only redraw levels (draft-2 fix: zero idle cost must be real)
-
-| level | trigger | work |
-|---|---|---|
-| A: chain only | iTime or ShaderAnimation, scene unchanged | re-run only the post chain from the retained `scene_tex` |
-| B: decor only | cursor glide, blink edge, hover fade | reuse the uploaded cell vbuf and image draws; rebuild and upload the decor instances AND the cursor-legibility glyph instances (the glyphs under the cursor quad in cursor-text colour, ui pipeline, clipped to the quad); redraw the scene, then the chain |
-| C: core frame | `needs_frame` or a core animation | full render and instance build |
-
-Cell instances are rebuilt only at level C. With a chain active, the cursor
-lives in the scene, so cursor motion is level B, not A (a documented choice:
-iCurrentCursor shaders still get exact uniforms).
-
-### 5.6 tty / web / wire
-
-GUI decor is GUI-only. The tty realises nothing from `styles` beyond the colours
-that are already in cells; its own effects are in §9.2. Web ignores the new
-data. The wire carries regions and styles so an attached GUI draws the same
-frame.
-
----------------------------------------------------------------------------------
-
-## 6. Post passes: a Shadertoy-compatible chain (GUI)
-
-Mirror ghostty 1.4.0 (`zig-pkg/ghostty-*/src/renderer/shadertoy.zig`,
-`shaders/shadertoy_prefix.glsl`, `generic.zig:2106-2230, ~1692`):
-
-- **Source format**: user file with `void mainImage(out vec4 fragColor, in vec2
- fragCoord)`, untouched. pardes prepends its OWN prefix with ghostty's exact
- uniform names, types and std140 order (iResolution, iTime, iTimeDelta,
- iFrameRate, iFrame, iChannelTime[4], iChannelResolution[4], iMouse, iDate,
- iSampleRate, iCurrentCursor, iPreviousCursor, iCurrentCursorColor,
- iPreviousCursorColor, iCurrentCursorStyle, iPreviousCursorStyle,
- iCursorVisible, iTimeCursorChange, iTimeFocus, iFocus, iPalette[256],
- iBackgroundColor, iForegroundColor, iCursorColor, iCursorText,
- iSelectionForegroundColor, iSelectionBackgroundColor; CURSORSTYLE_* defines;
- `#define texture2D texture`), but SDL GPU bindings: `iChannel0` at set 2
- binding 0, the block at set 3 bindings 0 to 2 (cut before and after
- iPalette: SDL GPU binds at most 4 KiB of a block, so one block left
- the colours after the palette at zero until G4), pardes's own at 3. `main()` calls
- `mainImage(_fragColor, gl_FragCoord.xy)`. Trap found: ghostty's Zig
- `Uniforms` (shadertoy.zig:13) orders selection_background before
- selection_foreground while its GLSL block orders Foreground before Background,
- so in ghostty the two arrive swapped. pardes fills by the GLSL names (the
- documented meaning); a unit test checks our Zig mirror's std140 offsets
- against the prefix (≈4.6 KB block, pushed with SDL_PushGPUFragmentUniformData).
-- **Y is down**, exactly like ghostty on Metal (`custom_shader_y_is_down`):
- fragCoord, iChannel0 UV and cursor uniforms share one top-left space, no flip.
- `iCurrentCursor = (left x, BOTTOM-edge y, w, h)` in framebuffer pixels (the
- "+Y edge" rule community cursor shaders rely on). Directional shaders appear
- flipped versus Linux ghostty (OpenGL, y-up); documented, not emulated.
-- **Chain**: a list of paths, run in order; ping-pong two UNORM textures
- (cleared to 0), the last pass writes the swapchain. iChannel0 holds
- sRGB-encoded values (scene_tex and swapchain are UNORM today — keep it so; if
- the scene ever moves to an _SRGB/float target, encode before the chain). Empty
- chain → render straight to the swapchain, no scene texture (today's cost).
-- **Uniform sources** (shell): iTime since the first post frame, iTimeDelta
- between draws, iFrame per draw (ghostty semantics, shell wall clock); iMouse =
- real Shadertoy mouse (superset; ghostty leaves it 0); cursor rect/colour/style
- from the region cursor (5.4 geometry), iTimeCursorChange stamped when rect or
- colour changes; iFocus/iTimeFocus from window focus events; palette,
- background, foreground, selection from the theme.
-- **Alpha**: forced to 1 unless WindowOpacity < 100 (then passed through
- premultiplied). Transparent grounds keep working.
-- **Input mapping**: identity. The bundled CRT has NO geometric distortion
- (beam-profile scanlines, aperture mask, bloom, vignette), so input stays exact.
- Barrel exists only in user shaders, with the drift documented: 1.6% is ~30 px,
- ≈3 cells, at the corners of a 4K screen, where the grips and rails are.
- `crt.zig` is deleted. If the user insists on barrel in the bundled CRT, keep
- `crt.zig`'s inverse keyed to "bundled crt active".
-- **Compile** (decision 3): a user's file is compiled by spawning `glslc` with
- the build's own flags, prefix and file on its stdin, SPIR-V on its stdout
- (no file of ours anywhere; std's spawn allocates before fork), on a thread of its own when the file joins the chain, never on the
- frame path; the result is taken in at the next loop step. The prefix ends in
- `#line 1`, so glslc's errors count the file's own lines, and they name the
- file, not `<stdin>`. Level A redraws once a refresh of the window's display. A failed compile keeps the file's last good
- pipeline and says glslc's first line as a message; glslc missing is said
- once, and only files stay off (the bundled passes are compiled with the
- build, through the same prefix). The files' directories are watched
- (file_watch shader slots): a change there has each file read and hashed,
- and one whose bytes moved compiles again; the same bytes (good or bad)
- compile nothing and say nothing twice. The process that holds the core
- compiles (ShaderBuild.zig): a local GUI, or a detached session, which
- sends its attached GUIs the chain with each file's SPIR-V (wire `post`,
- on attach and on every change), so an attached GUI runs the same passes,
- levels and ShaderAnimation as a local one and still reads no disk and runs
- no program.
-- **Animation mode**: `ShaderAnimation off|on|always` (ghostty's
- `custom-shader-animation`): off = redraw only when content changes; on =
- continuous while the window is focused; always = continuous. Continuous means
- the shell redraws its retained Surface (3.1), never a core tick or render.
-- **Existing crt** becomes a bundled Shadertoy file in `shaders/post/` under
- the same builtin; `scene_effects` and `crt.frag.glsl`/`crt.zig` are deleted.
- It takes a level, `Crt 0..3` (0 off, `on` the default 2, `off` and bare
- still work), passed as `pardesLevel`: 2 is the polished rewrite, 3 at least
- as strong as the old CRT (pixels changed by more than 8 in a channel against
- a plain frame: 18.3% vs 17.4%). Ripple and Glitch were rewritten too, then
- removed at the user's word after a live look (decision 8). The chain is
- `Crt` and `Shader <path>` in the order they were put in
- (`Shader off` empties it of files), `ShaderAnimation off|on|always`
- (default on). The core keeps no clock for it: level A redraws are the
- shell's.
-- macOS (out of scope) would need spirv-cross → MSL, as ghostty does.
-
-Region filters (e.g. "tint/blur one pane") are NOT built. If one is ever needed:
-one post pass reading a uniform array of ≤16 region rects+kinds, not a pass per
-region.
-
----------------------------------------------------------------------------------
-
-## 7. Animation model
-
-### 7.1 Time
-
-- The shell's monotonic `now_ns` is the only clock. It enters the core with each
- `pump` (and a `.tick = now_ns` event for shells without pump, e.g. esp32).
- Tests inject a fixed clock: a helper steps `now` by 16.67 ms.
-- Two domains, one clock:
- - **core animations** (change what hit-testing or the grid show, or state):
- panel geometry, message life, notice slide/fade, look-hover delay, chrome
- theme crossfade, focus lift/dim. Stored in the core, deterministic in tests.
- - **presentation-only animations** (never touch core state): cursor glide/
- trail/blink, shader iTime, smooth-scroll pixel offset, pet, touch flash.
- Owned by the shell, redraw the retained Surface, never tick the core.
-
-### 7.2 Two primitives, closed form
-
-```zig
-Tween { start_ns, duration_ns, from, to, curve } // value = lerp(from,to, curve(t))
-Spring { start_ns, x0, v0, target, omega } // critically damped:
- x(t) = target + (c1 + c2 t) e^{-ωt}, c1 = x0-target, c2 = v0 + ω c1
-```
-
-Both are pure functions of `now`: frame-rate independent, free catch-up after a
-sleep, no per-tick `advance` loops. Retarget: a Spring evaluates `(x, v)` at now
-and restarts from them (velocity kept); a Tween retargets from the displayed
-value with an ease-out curve (starts moving at once, no ease-in stall).
-Colours interpolate in OKLab (small CPU function), not sRGB bytes.
-
-Settling (draft-2 fix: a critically damped spring never reaches its target).
-Each quantity has an end rule: position springs end when |x − target| < 0.5
-device px and |v| < 1 px/s, lift/elevation when < 0.002, and then they snap to
-the target and stop. A test checks that `nextWake` returns null after settle.
-A Tween ends at `start + duration`.
-
-### 7.3 Idle and wakes
-
-`nextWake(now)` is ONE straight-line function listing every core animation
-(the same list `advance` settles):
-
-- running tween/spring → 0 (frame now)
-- holding phase (message linger, look-hover delay) → its end time
-- nothing → null
-
-Shells wait `min(core wake, own wake, input)`. A lingering message now costs
-zero frames until it starts to dissolve. `advance(now)` performs the state
-changes at phase ends (message entering→shown→lingering→leaving→removed,
-track done → drop, hover wait → preview) and sets `needs_frame` only when
-something visible changed.
-
-### 7.4 What each animation becomes
-
-| today | becomes |
-|---|---|
-| `ChromeAnimation` 10 steps sRGB (colors.zig:192) | Tween, 200 ms, OKLab, ease-in-out; retarget ease-out |
-| panel `Track.frame` counters (layout.zig:1583) | geometry Spring per track (retarget-safe); content crossfade Tween |
-| `MessageLife.frame` (Messages.zig) | phase + start_ns; motion from Tween curves |
-| look-hover wait frames (look.zig:902) | deadline |
-| GUI `AnimationClock`, tty `tickWatch` cadence, web `animationTicks`, macOS `spendTickTime`, detached tick | deleted; each shell passes `now_ns` and sleeps to `nextWake` |
-| crt time uniform, pet, touch flash, scroll lag | shell-owned, same now_ns source |
-
----------------------------------------------------------------------------------
-
-## 8. Motion, colour, performance, feel (both tracks)
-
-### 8.1 Motion spec
-
-| class | examples | curve | duration |
-|---|---|---|---|
-| micro-feedback | hover affordance, button/word highlight, selection appear | ease-out (cubic) | 80–120 ms |
-| follows input | cursor, smooth scroll, drag preview, pane geometry under interaction | critically damped spring | settles 90–150 ms (cursor), 180–260 ms (layout) |
-| arriving | notice drop-in, pane open | ease-out (cubic / quint), no overshoot | 150–220 ms |
-| leaving | notice dissolve, pane close | ease-in (cubic), shorter than arriving | 100–160 ms |
-| state change | focus lift/dim, theme crossfade | ease-in-out (smooth) | 150–250 ms |
-| anything > 300 ms | — | must be justified in the feel review | — |
-
-Rules: leaving is faster than arriving; nothing overshoots at text scale
-(a 2 px bounce reads as jitter); interruptions retarget (7.2), never snap or
-restart from zero; **input is never gated**: hit testing uses logical geometry
-(`presentation.pointer` already does), keys act on the logical state
-immediately, animations only lag the picture.
-
-### 8.2 Colour and contrast spec
-
-- Targets (WCAG ratio via `colors.themeContrast`, colors.zig:113): body text vs
- page ≥ 4.5 (aim 7); secondary ink (comments, line numbers) ≥ 3; tag/chrome text
- vs its band ≥ 4.5; non-text chrome (rules, grips, scroll thumb) ≥ 3 vs
- neighbours; selection text vs selection bg ≥ 4.5.
-- Absolute targets apply to the 15 native themes only, matching the existing
- test (colors.zig ~465), including its deliberate 2.5-4 band for line numbers
- (lineno is the exception to "secondary ≥ 3"). Imported themes get a RELATIVE
- rule: with every effect at its maximum (shadow darkening, unfocused dim, glow),
- no text/ground pair may drop below min(its original contrast, the target).
- Measured against the composited ground. If an effect would break the rule, its
- strength is clamped for that theme.
-- Subtlety budget: elevation shadow max 30% darkening at the edge, σ 0.4–0.8
- cell; unfocused dim ≤ 10% toward the page; glows ≤ 25% alpha; bloom threshold
- high (only true highlights), strength ≤ 5%.
-- Gamma: effect maths in linear light (shaders convert), shadows via pre-warped
- alpha (5.2), crossfades in OKLab; glyph coverage untouched (ui.frag stays).
- Wide gradients dithered.
-- Focus hierarchy by light, not only hue: the active pane is lifted (shadow),
- unfocused panes recede (dim), the cursor is the brightest thing on screen,
- chrome is quieter than text.
-
-### 8.3 Performance budget
-
-- Targets: 60 Hz → 16.6 ms, 144 Hz → 6.9 ms end to end. Shares (144 Hz,
- 200×60 grid, 4K, iGPU): core render ≤ 1.5 ms (only on state change), GUI
- instance build ≤ 1 ms, scene GPU ≤ 2 ms, decor+text-shadow ≤ 0.5 ms, post
- chain ≤ 2 ms for two passes (bloom at half res). Measured with the existing
- tracy zones plus GPU timestamps; numbers are budgets to verify, not facts.
-- Idle: zero ticks, zero renders, zero presents (7.3); blink stops after 10 s.
-- Presentation-only animation: shell redraw only, no core render.
-- Degrade order: post chain to half res → bloom off → soft shadows become hard
- (σ = 0) → cursor trail off → transitions snap. Hysteresis: step down after 30
- of the last 60 frames miss the budget; step up only after 10 s with none
- missed; at most one step per 2 s, reset on theme or window-size change. A step
- is never taken mid-transition, only at the next idle, and each one is logged
- (and shown in Debug). Input handling is never deferred for effects.
-
-### 8.4 Feel review (gate before an effect is kept)
-
-Each effect lands behind its own toggle, default off. Before "keep": render a
-frame sequence with the injected clock (GUI capture mode, PPM per frame, at 60
-and 144 Hz; tty via the snapshot harness `snapstyle` per frame) and a live
-screen recording; judge in motion against a reference (the lapis page, ghostty
-cursor trail, neovide smear); record keep / polish / drop with one line of why
-in `docs/effects.md`. Code review never substitutes for this.
-
----------------------------------------------------------------------------------
-
-## 9. Effects catalogues
-
-Every effect: a toggle (one `Fx` bitset in settings + builtin words, like
-`scene_effects` today), zero instances/passes/uniforms when off.
-
-### 9.1 GUI track
-
-Existing, verdicts: crt → REWRITE as bundled Shadertoy (Lottes-style mask,
-proper scanline beam profile, subtle); ripple → DROP; glitch → DROP (or a
-tasteful rewrite only if the feel review asks); pet → keep as-is, out of scope;
-transitions slide → POLISH (spring + elevation shadow while moving); zoom → DROP;
-dissolve → REPLACE with a linear-light crossfade; vertical → POLISH timing; the
-character effects (ascii…typewriter) → not a GUI effect (terminal track).
-
-| # | effect | needs from pipeline | pri |
+| what | tty | window | from |
|---|---|---|---|
-| G1 | soft elevation shadows onto lower panes (active lifted, floating chips, dragged pane) | regions + z tiers, decor `soft_shadow` | P1 |
-| G2 | focus transition: lift (spring) + unfocused dim | core focus Tween, `Region.dim`, cell-instance dim | P1 |
-| G3 | cursor: jumps glide on springs, adjacent moves snap (≤40 ms), 4-corner smear with glyphs re-drawn legibly inside it, optional edge-eased blink | layer geometry, per-pane view origin, shell time, decor `cursor`, redraw level B | P1 |
-| G4 | Shadertoy chain + bundled CRT, bloom (half-res dual-Kawase), vignette+grain | post chain, offscreen ping-pong, shell time | P1 |
-| G5 | lapis decor: rings, checker, stripes, hard offset shadows, title text shadow | styles, decor kinds, 3-phase text shadow | P1 (lapis set) |
-| G6 | smooth scroll: pixel offset on a spring, momentum for touchpads | shell-side; replaces scroll_lag stepping | P2 |
-| G7 | panel transitions redone: spring geometry, shadow while moving, linear crossfade | Track as Spring, tiers | P2 |
-| G8 | notice chips: elevation, shadow, ease-out drop, ease-in fade | tier 4, existing slide/fade | P2 |
-| G9 | selection glow: soft halo behind selection runs, 100 ms fade-in | selection runs as regions (overlay kind) | P2 |
-| G10 | theme crossfade in OKLab, 200 ms | Tween | P2 |
-| G11 | look-hover affordance: soft underline glow, 80 ms | hover region | P3 |
-| G12 | pane edge ambient occlusion (inner shadow, 1–2%) | decor `inner_shadow` | P3 |
-| G13 | parallax body pattern (lapis dots under see-through bodies at 0.25× scroll) | `see_through` style, per-pane view origin | P3 |
-| G14 | damage afterglow (changed cells glow briefly) | removed: a pane whose whole text changed flashed whole (the user: "really bad, just remove it") | — |
-
-### 9.2 Terminal track (tty, cell-native; separate, smaller)
-
-Where: presentation-only cell rewrites in the tui shell after the canonical
-grid (`tui/panel_compositor.zig` grows into the tty compositor), never in the
-canonical grid, so 9P/goldens stay clean with effects off. Gated on truecolor
-(vaxis caps / COLORTERM) — static effects may quantise to 256, animated ones
-turn off below truecolor. Safe glyph set by default (░▒▓ ▀▄▌▐ ▖▗▘▝);
-sextants/octants behind a setting. Budget: an animated frame ≤ 8 KB of output
-(≤ 4 KB and 30 Hz when `SSH_CONNECTION` is set); prefer effects whose per-frame
-delta is a moving edge, not a full-pane rewrite; without sync 2026, animated
-effects are off (static ones stay). vaxis has `caps.rgb` but no 2026 cap: the
-tty sends its own `CSI ? 2026 $ p` (DECRQM) at startup, next to the kitty shm
-probe; no answer = assume unsupported. Under tmux, tmux answers for itself (its
-own 2026 handling; old tmux does not answer → animated effects off), which is
-correct, because tmux is the thing that repaints the outer terminal.
-
-Enforcing the budget (draft-2 fix): tty animations are time-based, so dropping
-samples is free and correct. After each `vx.render` the shell counts the bytes
-written. If a frame exceeded the budget, the next sample is deferred by
-`bytes / budget × frame` ms. Slide and vertical therefore get fewer samples over
-ssh; they are not exempt. Effects are designed so each cell changes as few
-times as possible over the whole animation.
-
-Audit of what exists:
-
-| effect | where | verdict | why |
-|---|---|---|---|
-| ascii (byte walk) | core compose | REMOVE | slot-machine noise, rewrites every changed cell every tick |
-| edges | core compose | REMOVE | gimmick, full-pane churn |
-| fall | core compose | REMOVE | gimmick |
-| wave | core compose | REMOVE | gimmick, illegible mid-motion |
-| scramble | core compose | REMOVE | noise every tick, worst byte cost |
-| typewriter | core compose | REMOVE | slow on big panes, row-major reveal reads as lag |
-| curtain | core compose | POLISH → wipe | a directional reveal is good; give it a 2-cell soft truecolor edge |
-| slide | tty compositor | POLISH | ≤150 ms ease-out, pane open/close only; a full-pane rewrite per sample, so the byte budget thins it to few samples over ssh |
-| zoom | tty compositor | REMOVE | nearest-neighbour cell scaling is garbage |
-| dissolve | tty compositor | POLISH → ordered swap | Bayer 4×4 threshold per cell; each cell swaps old→new ONCE at its threshold (no continuous colour lerp), so the total output is one pane's worth spread over the animation |
-| vertical | tty compositor | KEEP (timing polish) | reads as a drawer, cheap |
-| message fade | core (grid colours) | KEEP (OKLab) | |
-| theme crossfade | core | KEEP (OKLab, fewer steps over ssh) | |
-
-Removing the six core-composed effects deletes `composeAsciiTransitions`,
-`charSource`, `AsciiDiff`, `PanelCellDiff.ascii` and their shader/GUI paths
-(user decides).
-
-New, cell-native:
-
-| # | effect | technique | pri |
-|---|---|---|---|
-| T1 | floating shadow for chips/debug box/drag preview | 1 cell right/down darkened in OKLab; blank edge cells get ▗▄/▐ half-cell edges | P1 |
-| T2 | cursor jump trail | on jumps ≥ 3 cells: 3–5 cells along the path, bg ramp cursor→page, 120 ms; tiny delta | P1 |
-| T3 | wipe transition | 2-cell soft leading edge; only the edge changes per frame | P2 |
-| T4 | ordered swap transition | = polished dissolve: Bayer threshold, one swap per cell | P2 |
-| T5 | unfocused dim | fg 15% toward bg, instant (no animation over ssh) | P2 |
-| T6 | active tag gradient | subtle horizontal truecolor ramp on the active tag band | P3 |
-| T7 | scroll thumb flash | rail thumb brightens 300 ms on scroll | P3 |
-
----------------------------------------------------------------------------------
-
-## 10. Lapis on this pipeline
-
-Colours: a normal theme file (lapis #0f1a4a, deep #0a1030, ink #050a24, gold
-#e8c46a, vermilion #d4412f/#ff6a4a, vellum #eee6d2/#9aa6d9). No fonts. Styles:
-
-- workspace/column/pane tag (the "trail"): lapis fill, 2 px gold ring, hard
- shadow 5×5 px vermilion (px snapped to device pixels, no AA), 3 px title
- text shadow for the file name.
-- pane (the "panel"): gutter (2 cells ≈ the CSS 18 px band) carries the
- ornament pattern (checker gold/lapis), inset rings gold/lapis/vermilion
- (2/3/1 px) on the pane edge, soft black shadow from its tier.
-- column tag bar / notice chips: striped gold/lapis (`stripes`, period 4 px)
- behind the text band, ring gold, hard vermilion shadow.
-- page: lapis-deep. The dot grid sits under the pane BODIES (`dots` pattern
- with `see_through`, so page-coloured cell backgrounds let it show between
- glyphs). It is fixed to the pane, not to the text, so it does not track
- scroll. G13 parallax is retargeted to this body pattern at 0.25× scroll and
- stays P3.
-
-Where the pixels come from: left band = the pane's own gutter; top = its tag
-row (the band has vertical slack above/below the tagline glyphs); right/bottom
-= 1–2 px rings over the last column/row's side bearings; shadows fall on the
-neighbour's gutter/tag row (§5.2, drawn after the neighbour's cells, §3.4). No
-layout change. If the user wants the CSS
-look with real gaps between panes, that is a separate layout knob (question 1).
-
----------------------------------------------------------------------------------
-
-## 11. Damage / dirty tracking
-
-No new dirty bits without a measurement. Existing: `needs_frame` (frame-level),
-vaxis cell diff (tty output), wire deltas, web node caches. The pipeline makes
-two savings free: (a) presentation-only animations skip the core render
-entirely; (b) deadlines replace per-tick renders for holding phases. Region
-lists make per-region instance caching possible later, if tracy shows the
-instance build matters.
-
----------------------------------------------------------------------------------
-
-## 12. What gets deleted or merged
-
-Core: second pane-tag paint, second body paint, grid-notice pass (joined from
-layers), duplicated header hover/selection/caret, `std.mem.swap` of the Surface,
-hover-bit reset + `mark_hover`, per-animation `advance` counters and frame
-constants, `animationActive` (→ `nextWake`), `SceneEffect` (→ shader list),
-`capturePrevious` on idle frames with transitions off; if the user agrees, the
-six core-composed character effects.
-Types: TagLayer + BodyLayer → Layer; layout.zig's Presentation/Track/Easing/
-Animation move to `src/Presentation.zig` and `src/animation.zig` (pure moves).
-GUI: AnimationClock, finishPresentedAnimationFrame, PaintPlan/PaintBatch,
-coverLayers-as-cell-loop (→ page cover), the five inference functions (4),
-crt.zig mapping, crt.frag.glsl, `scene_failures` plumbing folds into the chain.
-Shells: four ack sites, six tick drivers.
-
----------------------------------------------------------------------------------
-
-## 13. Staged plan (jj changes on a bookmark `render-pipeline`; effects on top of it on `fx`, droppable)
-
-Gate for every stage: `zig build snap` byte-identical, `zig build unit-test`
-and `unit-test -Dplatform=gui` pass, `-Dplatform=web` builds, tracy frame time
-not worse at 200×60. Builds redirected to files; `-Dplatform` always; PARDES_*
-stripped. No visual change until stage 9 unless stated.
-
-| # | change | notes / risk |
-|---|---|---|
-| 0 | GUI capture goldens: a handful of scenes rendered hidden in capture mode, PPM hashes checked in (same machine/driver) | the safety net the GUI stages need; machine- and driver-specific, so a LOCAL gate only, never CI |
-| 1 | pure moves: render → `src/draw.zig`; Presentation/animation out of layout.zig | no behaviour |
-| 2a | clock plumbing: `Host.now` (the shell's monotonic ns) and `Pardes.advance(now)`, which runs one `.tick` per whole 16 ms frame since core time last stood still; `nextWake()` lists every core animation and answers the next frame while anything moves, the wait's end while something only waits (message linger, look-hover delay), null when idle; the pump sleeps to it and draws only stepped frames. No shell posts `.tick` any more (tty's timer thread only wakes the wait; GUI's AnimationClock, web's JS tick bank and the detached server's and the board's ticks are gone; grid mode moves its virtual clock straight to the next wake). `acknowledgePanelPresentation` stays in the shells (revisit at stage 8: grid mode acks `&.{}` on purpose). capturePrevious only when transitions are on. `.tick` stays as the one-frame step `advance` and the unit tests use, so every curve and every asserted number is unchanged | pacing is corrected (144 Hz no longer ~25% slow, ssh no longer drifts): a bug fix allowed in this phase. PARDES_TEST_CLOCK (set by the snapshot harness): tty and the detached server answer a virtual clock that a timed-out wait moves exactly to the core's next wake, so goldens replay the same frame sequence on any machine load. Review fixes: tty's wake timer is interruptible (a newer, shorter request cuts short an older sleep); a wait (linger, hover delay) is jumped to its end in one step, never counted against the 240-frame catch-up cap; an overshoot under 1.5 ms after a step is let go, so 60 and 120 Hz take exactly one step per display frame. Known exceptions: the GUI still polls every 16 ms when idle (SDL events, gamepad, fs tick, smooth scroll depend on it; revisit at stage 8); a minimized GUI renders every stepped frame of a running animation (stage 8) |
-| 2b | (on `fx`, feel-reviewed) Tween/Spring curves, settle rules, OKLab crossfades, new durations from §8.1 | a look change, kept apart from 2a so either can be bisected |
-| 3 | paint functions take `s: *Surface`; delete the swap hack | mechanical, wide |
-| 4 | PLACE: Region list built once; renderPane/tags/notices read their rects from it | wait for tag→Text; the refactor is already replacing BOX_H with `pane.tag_rows` and `p.tagTop/bodyTop(pane, r)` — PLACE absorbs those into the region rects |
-| 5 | JOIN: paint tags/notices/headers once into layers, copy into grid; the grid wins where the copies disagree, and each disagreement is listed for a later decision | goldens are the oracle; wide-grapheme re-clip at the edge (§3.2) |
-| 6 | body layer joined the same way (paint once, copy visible rows) | riskiest join; A/B the old double paint in a temporary test with tag_bottom on and off, then delete it |
-| 7 | Layer merge (TagLayer + BodyLayer), wire v8, web accessors | after the other agent lands; touches mouse hit paths; breaks the macOS shell's layer ABI (accepted: macOS build ignored for now); done: one `Layer` (src/Layer.zig) with `rows` (0 = no layer) and a `cursor{x, y}`; a tag of N rows is ONE layer of N grid rows (`tagHit` answers the row as `line`, `bodyHit` keeps its meaning), so the per-line layer bases are gone. Wire v8 ships rows, cursor y and the region list in the one bump; v7 and v9 peers are refused in both directions (tests). web: `tag_layer_value` 11 = rows, 12 = cursor y, and app.mjs lays every row. macOS: its Zig side compiles against `Layer`, but pardes.h still sees one row per tag layer (a taller tag shows its first row there) |
-| 8 | GUI draws from regions: role tiers with track groups in tiers 2-3, page cover, hard-edged snapped decor for rules/rails/grips (pane chrome in the pane's tier), per-instance clip; delete inference functions, `transient_on` and `mark_hover` (breaks macOS glass hover; accepted, macOS ignored for now) | stage-0 PPM goldens byte-identical (possible only because hard decor is snapped and not anti-aliased); pane chrome now also shows during transitions, which is the one allowed visible delta, listed Done: the frame is drawn in groups (makeGroups): tier 0, one group per track in paint order (tiers 2 and 3), tier 4 as two groups (the notices; then the guides and the debug box, which the core paints over them), tier 5 (bar cursors), each drawing cells, images, decor. Decor is every rule, rail, thumb, grip mark, spine, the workspace and column rules, the bottom band, notice rules and bar cursors: whole-pixel rects drawn by `decor.frag` over `ui.vert` as cell instances (colour in the ground, coverage in fg.r, blended like the overlay), so each carries its track's transition and clip; a closing pane's comes from the last frame's regions (`Surface.previous_regions`). Only `solid` exists yet: `decor.vert` and the other kinds (§5.1) come with the first effect that needs them. Deleted: taglineBaseRgb, topbarPaneBorderHeight, bottomTaglinePresent, frameChromeBg, cellBackgroundIs and the rail inference, paneGripCell's scan, transient_on, PaintPlan; one cover map (coverFrame: layer, grip and offset, focus, anchor, floating) is marked from layers and regions once a frame. New regions: `column` (spines, anchors, the focused column's tint), `guide`, `debug`; `Surface.chrome` is the palette, on the wire in v8 (not yet shipped, so no bump), and an attached GUI draws the same chrome (test). Per-instance clip (`CellInstance.clip_*`, 104 → 120 bytes an instance, ~15% more upload a frame) replaces the vertical transition's scissor. While any pane moves, opens or closes, notices stay in their own panes' groups, under whatever slides over them; tier 4 holds them only when nothing moves. Goldens: 01-12 byte-identical; 13-17 differ only in pixels past the grid (with a picture on screen the image pass left the scissor at the grid's size, cutting every rule end, rail foot and band that runs into the leftover pixels; they now run to the edge as in every scene without one); 16-debug shows the debug box (it was drawn from the grid under the source's context-row layer, so the GUI never showed it there); 18-mid-transition is new (virtual clock, PanelSlide Newcol with the picture, frame 6 of 12: chrome moves with its pane). Other deltas, outside the goldens: a guide over a tag or a context-row body now shows; with WindowOpacity < 100 a layer's bar cursor is ink like the grid's; an attached GUI gains the focus tint, notice rules, spines and the theme's page and caret colours. Deferrals fixed: the GUI sleeps when idle (SDL and queue events wake it; a smooth scroll, a gamepad, the test feed, a shell's kill deadline, a present to retry still poll), and a minimized or occluded window sleeps through animation. Tracy, `gui frame build`, 200x60, terminal output plus scrolling, ReleaseFast, two interleaved runs of ~180 frames: before 992/1015 µs median, after 989/987 µs |
-| 9 | post chain: glslang, Shadertoy prefix, ping-pong, ShaderAnimation, redraw levels A/B (§5.5); bundled CRT without barrel; delete scene_effects/crt.zig/crt.frag | first visual change (the CRT look) Done: src/gui/Post.zig, shaders/post/ (prefix, crt, ripple, glitch), post.vert. glslc (decision 3) with compileGlsl's flags, off the frame path; ghostty's Uniforms matched offset for offset by a test against a copy, and the prefix's block checked name by name; ghostty's test_shadertoy_crt and _focus compile, _invalid fails with glslc's words. Level A measured (Tracy, 200x60, Crt, idle 10 s): 215 chain-only redraws at 34 µs median CPU each, 6 core frames (957 µs) in the same time; the core's clock stays idle. Input is identity (test: the corner click with Crt on). The three bundled passes were rewritten: Crt without barrel or tube edge, scanlines and mask that average to one, dithered vignette; Ripple as rings in pixels, eased in, lit on their slopes in linear light, dithered; Glitch as short eased bursts of torn bands with an RGB split, keyed to iTime. Before/after stills for the user's judgement. |
-| 10+ | `fx` bookmark: G1–G3 → feel review → G4 bundled → G5 lapis theme → P2s; tty T1–T2 → feel review → removals of audited effects (after user decision) → T3–T5 | each effect its own change, default off |
-
-Tests to add: fixed-clock animation tests (Tween/Spring closed form, retarget
-velocity continuity, `nextWake` for each holding phase); region placement vs
-old ad-hoc geometry (all layouts, tag_bottom, collapsed, multi-line tags);
-contrast test (absolute targets on the native themes, relative rule on all themes × effect maxima);
-settle tests (`nextWake` is null after every spring settles); Shadertoy prefix compile test with
-ghostty's test shaders (`test_shadertoy_crt.glsl`, `_focus`, `_invalid`) and
-uniform-offset checks against ghostty's `Uniforms` struct; tty byte budget test
-(bytes per animated frame via the snapshot harness).
-
-Risks: the tag refactor moving under stages 4–7; body-layer join parity; glslang
-build time and size (C++, ~minutes on the ~10 min build); machine-specific GPU
-goldens; ssh/tty byte budgets; wire version bump breaks mixed-version attach.
-
----------------------------------------------------------------------------------
-
-## 14. Recorded open points (agreed with the adversary: not blockers)
-
-1. Selection colours: because ghostty swaps them (§6), Shadertoy shaders tuned
- on ghostty will see iSelectionForeground and iSelectionBackground swapped in
- pardes, which fills them by their GLSL names. Documented.
-2. A same-tier cast shadow over a neighbour sits under that neighbour's
- decor-over rings, because rings are drawn last. Accepted as the look; check it
- in the feel review.
-3. Gamma is a compromise (sRGB-space blending with pre-warped alpha). Coloured
- glows keep a slightly dark falloff until the RGBA16F experiment is measured.
-4. The macOS shell breaks at stages 7 and 8 (Layer ABI, mark_hover). Accepted,
- since the macOS build is ignored for now.
-5. Stage-0 GPU goldens gate locally only.
-6. Revisit at the first visual stage (from stage 5): the notice layer lost
- its hover word to keep the grid's behaviour. Notice words are Look/Exec
- targets, so bring the hover affordance back in BOTH the grid and the
- layer then. Done right after stage 8: a notice's word under the pointer
- is lit as a header's is, in the layer and so in the grid's copy joined
- from it. The same change deleted `mark_hover`, the `Cell.hover` bit and
- the per-frame reset of it (never set since stage 5; macOS loses its
- glass hover rect, accepted with open point 4).
-7. Revisit at stage 8 (from stage 6): a body without context rows has no
- layer (paint once, straight onto the grid). If the GUI is to read every
- body from a layer, give every body one then and measure the copy. The
- context-row path also still builds the body's text twice (a pre-pass
- with body_rows 0 decides the context rows, renderBody builds it again at
- the layer's rows); the paint is single. Measure both there. Decided at
- stage 8, measured (Tracy, 200x60, TreeContext on a scrolled source,
- ReleaseFast): the pre-pass text build is 5 µs median and the layer's
- copies 10 µs per context-row body per frame. Kept as they are: nothing in
- the GUI reads a layerless body from anything but the grid, which is
- exact, so a layer for every body would be ~10 µs a body a frame for no
- pixel.
-
-## 15. Questions for the user
+| a scroll's glide (glide.zig) | whole rows | the pixel | a body's line and doc between two frames, and the last frame's cells |
+| the caret's ease | (the terminal's cursor) | its quad on critically damped springs (gui.zig's `Caret`) | the cursor and the focused body |
+| the blink (`CursorBlink`) | (the terminal's) | the cursor's alpha | the window's last key or click |
+| a jump's trail (`JumpTrail`, jump_trail.zig) | the ground behind the cursor, truecolor terminals | drawn over it, in the overlay | the cursor between two frames |
+| `InactiveDim` | (kept, not drawn) | the unfocused panes' ink faded, never below its contrast floor | the regions' focus |
+| `Pet` | (unsupported) | a sprite in the workspace tag's blank end (gui/pet.zig) | the grid's first row; SDL's ticks |
-1. Lapis geometry: decorations inside existing chrome (gutter band, tag row,
- thin edge rings; no layout change — recommended), or real pixel/cell gaps
- between panes like the CSS page (layout + hit-testing change)?
-2. Remove the six core-composed character transitions (ascii, edges, fall,
- wave, scramble, typewriter) and GUI zoom/ripple/glitch?
-3. Runtime shader compile: link glslang (C++, from ghostty's vendored pkg,
- allocator-policy exception), or spawn `glslc` at load time (no new dep, needs
- glslc installed)?
-4. Focus lift/dim on by default once it passes the feel review, or opt-in?
-5. Stage 2a corrects animation pacing (144 Hz is ~25% slow today, and ssh
- drifts) while keeping every curve. OK as a bug fix inside the "no visual
- change" phase?
-6. Cursor blink: pardes never blinked. Keep it off by default (recommended),
- with an opt-in eased square-wave blink?
-7. Bundled CRT without barrel distortion (exact clicks, recommended), or barrel
- plus keeping crt.zig's inverse mapping for the bundled shader?
+The settings these read are plain bytes in the core's settings, dumped
+and replicated as any other; nothing a frontend keeps goes back to the
+core.
-## Decisions (user, 2026-09-28)
+## The window's composition
-1. Lapis is drawn inside the existing pane chrome; no gaps between panes.
-2. Keep every whimsical effect (the six terminal transitions, GUI zoom, ripple,
- glitch): the problem is their quality, so polish each until it feels good
- instead of removing it.
-3. Shadertoy shaders are compiled by spawning `glslc`, not by linking glslang.
-4. Focus lift/dim default: undecided, ask when that stage lands.
-5. The animation pacing fix goes into the no-visual-change phase.
-6. Cursor blink is on by default.
-7. The bundled CRT has no barrel distortion, so clicks stay exact.
-8. (After trying them live) Ripple and Glitch are removed entirely; Crt stays.
+Four fixed draw groups, bottom to top: the page and the panes, the
+notices, the guides and the debug box the core paints over them, and the
+ink (bar cursors); then the overlay (the pet, a jump's trail, the caret's
+quad). Decor is flat: rules, rails, grips, spines, notice rules and bar
+cursors as hard-edged rectangles read from the regions.
diff --git a/docs/typ/building.typ b/docs/typ/building.typ
index 972b8a97..456cbb3e 100644
--- a/docs/typ/building.typ
+++ b/docs/typ/building.typ
@@ -305,8 +305,8 @@ the hint.
= Other notes
`docs/` keeps the platform and design notes beside this book:
-`macos.md` (the macOS shell; `web.md` says what the browser shell was), `effects.md` and
-`render-pipeline.md` (visual effects), `helix-keys.md` and `selections.md` (helix parity),
+`macos.md` (the macOS shell; `web.md` says what the browser shell was),
+`render-pipeline.md` (a frame, and what the frontends draw on their own), `helix-keys.md` and `selections.md` (helix parity),
`lsp-evaluation.md`, `ui-review.md`, `divergences.md` (bookmarks off
`main`) and `open-questions.md`. `next-steps.txt` is a wishlist and
`transactions.txt` records one open structural gap against helix.