Repository navigation
Conversation
Adds internal/interaction-harness (@astryxdesign/interaction-harness) and playwright.interaction.config.ts: phone-sized touch projects in Chromium and WebKit for *.interaction.browser.spec.ts specs. Touch scripts are data (tap, drag with hold and rest, flick with an exact release velocity and velocity profiles, hold-then-drag, pinch, rotate, any path, a second finger mid-gesture), played as real engine input: CDP Input.dispatchTouchEvent with scripted timestamps in Chromium, WebKit's own input command in WebKit. The harness's self-test proves the timestamps, a native fling and multi-touch delivery; a pr-interaction CI job runs the proofs against each PR's Storybook build, and a container script runs the WebKit project where host libraries are missing.
Springs, scroll physics, timing curves, navigation values and gesture thresholds for the touch presentations of core components, each with its source (Apple documentation, a WWDC session, WebKit's UIKit-matched source, or a measurement of UIKit on the iOS 26.2 simulator) or labelled a guess with its measurement task. A test enforces the source rules on every domain file, in their JSDoc and typed forms, and checks the list of simulator measurements still owed.
Pure functions for the touch presentations: a closed-form damped spring with every Apple parameterization and Core Animation's settling duration, a spring chasing a moving target, UIScrollView's coast, rubber band and release projection. Tests hold them to an RK4 integration, WebKit's spring solver, the WWDC conversion equations and values read from UIKit, and play the constants table through them.
…rena An internal gesture system for touch presentations, not exported from the package. GestureRecognizer carries UIKit's states and allowed transitions, discrete and continuous recognizers, require-to-fail, simultaneous recognition on request, shouldBegin and shouldReceiveTouch, cancelsTouchesInView and the touch delays. GestureArena settles each touch sequence without the DOM: participants, resolution order, claims, delivery to the view, click dispositions (allow, suppress, defer and replay) and timers as deadlines on the event clock. Its rules are tested with scripted recognizers, independent of any real one.
…sholds Tap recognizes at the lift within 45 pt of movement, taps 0.35 s and 45 pt apart, and a 1.5 s hold; long press begins at exactly touch-down + 0.5 s on the event clock within 10 pt, even when the main thread was blocked, and a zero duration begins on the touch itself. The arena tests with these real recognizers cover tap against double tap, delegate requirements, deferred clicks and cancellation.
Touch behaviours key on a coarse primary pointer, layout behaviours on a compact viewport, and presentations that need both (a sheet in place of a popover) ask for both. utils/touchPresentation.ts holds the three queries, SSR-safe hooks (useTouchPresentation; useMotionPresentation for components with no interaction to infer from: UIKit motion under a finger, theme tokens under a mouse) and a one-moment read. Styles write the same queries as StyleX media keys, and a test keeps core's compact-touch keys spelled as the shared query. A compact viewport is narrower than 744 px or shorter than 500 px, the closest a size gets to UIKit's compact size class: an iPad mini in portrait keeps the anchored popover and a phone in landscape gets the sheet; a Max-size iPhone in landscape is the known miss. The adaptive popover-or- sheet presentation reads the shared query; its docs, stories and tests follow.
Pure functions for the touch presentations: a closed-form damped spring with every Apple parameterization and Core Animation's settling duration, a spring chasing a moving target, UIScrollView's coast, rubber band and release projection. Tests hold them to an RK4 integration, WebKit's spring solver, the WWDC conversion equations and values read from UIKit, and play the constants table through them.
createAnimator plays springs and UIScrollView coasts as Web Animations keyframes on translate, rotate, scale and opacity, on one explicit clock, so they keep running while JavaScript is busy. Presentation values and velocities come from the curves, so a finger can catch any animation and a new target keeps the velocity; the end state is committed inline and held by the keyframes through a main-thread block, and a finished animation retires when its own time reaches the curve's end, so a sampler can pause it; Reduce Motion substitutes UIKit's cross-fade.
… Chromium The Chromium driver sent a lift of one finger of several as a touchMove listing the fingers still down, assuming CDP releases a point a move omits. It does not: the lifted finger stayed down. A touchEnd that lists the lifting fingers releases exactly those while the others keep moving, so the driver sends that, and a self-test lifts one of two fingers and checks the release, the other finger's moves and the touchend's remaining touches. Playwright's WebKit sends that touchend with no remaining touches, which iOS never does. The harness does not work around it: drivers expose partialLifts, and PARTIAL_LIFT_SKIP is the reason a proof cites when it skips; the README lists it with the other engine limits.
Pause every Web Animation, seek it to a time and read the presented transform and opacity, then compare the curve with oracles that share no code with the motion under test: cubic-bezier, the SwiftUI spring (response, dampingFraction, initial velocity), UIScrollView deceleration and UIKit's rubber band. Storyboards photograph seeked animations at t = 0, 16.7, 33.3 ... ms or each step of a scripted touch, and write a PNG contact sheet plus the JSON trace a spec asserts, under the gitignored test-results/interaction-evidence. A timer gate holds a component's own timers while a filmstrip is taken. pnpm storyboards runs the tagged tests.
…sented values An animation that keeps its end value in a run of identical final keyframes (held so a later interruption can read it) reported its whole duration as the curve: storyboards filmed the hold, waitForAnimation's interval reached past the landing, and waitForRest, which also required every animation to leave the running state and read only transform, timed out. The curve now ends where that run starts (HeldAnimation.curveEndMs, waitForAnimation's to, the default storyboard times), and rest is no animation pending or before its curve's end plus the presented transform, translate, rotate, scale and opacity (and ::backdrop opacity) still for 3 frames, whatever the playState. sampleCurve composes translate, rotate and scale with transform, so curves of the individual properties are sampled.
Timestamped composited frames (Chromium's screencast and WebKit's own), a marker the frames can find, and the busy-main-thread proof: block the page's main thread mid-flight and show the element still moving every frame, with a requestAnimationFrame animation that freezes as the negative control. Chromium trace counts of Layout, Paint and style work inside an animation window, and long tasks. The exemplar spec proves the BottomSheet drag to a detent with all of it: the drag follows the finger, a flick dismisses, the settle matches its timing function, keeps advancing through a blocked main thread in Chromium and WebKit, does no Layout or Paint, and gets a drag and settle storyboard.
…ed machine The proof asserted a stall under 40 ms and near-exact frame counts, so a dropped screencast frame or two on a loaded runner failed it, and frames delivered out of order read as the element stepping backwards. busyWindowProof now returns a verdict that tolerates what load does to a screencast and nothing a frozen animation does: a median frame interval of at most 20 ms, no stall longer than three screencast intervals, and at least 90% of the window's whole frame intervals presenting a new position (one miss always allowed), with the raw numbers in a summary. Screencast frames are sorted by presentation time, the cadence comes from the frames around the window, and blockMainThread can block from inside the page in the frame an animation has run, so a short settle is still judged at length on a busy machine. The exemplar judges its settle until it is within 2 px of rest, since the screencast rounds to whole pixels. A unit test covers the limits; the rAF negative control still fails every criterion.
The pan begins past 10 pt on either axis and leaves that travel out of its translation, as UIPanGestureRecognizer does; axis and one-way directions with axisRatio, axis lock, finger changes without jumps, scrollEdge, and an opt-in non-UIKit scrollHandoff. VelocityTracker defaults to UIKit's filter (0.2 x the newest event's velocity + 0.8 x the previous one's), which reproduces every value UIKit reported on recorded streams.
…zers The rest of UIKit's recognizer set on the gesture system: swipe recognized during the motion, pinch measured from the starting span moved by its hysteresis, unwrapped rotation, edge pans resolved against the writing direction that fail at touch-down outside their band, and hover for mice and hovering pens.
The touch-action an element gets from its recognizers (pinch-zoom stays unless a pinch or rotation recognizer is attached), what the scrollers under a touch could still do along the scroll chain, and UIKit's haptic feedback kinds on the web: navigator.vibrate where it exists, the switch tap haptic in iOS Safari.
…s takeover One host per document routes fingers (touch events, paired with their pointer events), mice and pens to the recognizers on the hit path. A recognizer that owns a touch cancels its touchmove through a non-passive listener; pointercancel, an uncancelable move or a scroll mean the browser took the touch; a claimed touch's touchend is cancelled and its click swallowed, a deferred click replayed when its recognizer fails. pointerIdOf names the pointer behind a tracked contact, so a held finger can be handed to code that follows pointers.
… and hooks PressTracker applies UIControl's events and its 70 pt tracking margin, the content-touch delay inside scrollers, the quick-tap flash and the scroll brake, with two profiles: control (UIButton) and cell (a row: delayed highlight everywhere, cancelled past the pan hysteresis, no flash). paintPress is the one writer of data-astryx-press and never clobbers a held row. installPressAttribute paints every marked surface with one tracker per touch, picks the profile from the marker's value and releases on its clock. useGestureRecognizer attaches a recognizer with a stable identity and no render per move.
The same touch scripts run in Chromium (CDP touches with timestamps) and WebKit (its own touch input): touch-action per element, claims over a scroller, no ghost click, scroll edge, pinch on its element only, a second finger, and a long press at exactly 500 ms under a blocked main thread. The README documents the system for component authors.
…tarts at the commit
play() bakes curves given as value and velocity functions of time (a spring
chasing a moving target, a fitted UIKit curve) like a spring.
createAnimator(dialog, {pseudoElement: '::backdrop'}) drives a
pseudo-element: a static filling animation holds its end value.
animatorFor(element) shares one animator per element, so callers animating
different axes compose in one translate track; atRest() waits out every
curve on its properties, and rebase() moves them by a layout shift keeping
their springs. A call from rest gets no start time, so the browser starts it
at its commit, as Core Animation does, and handle.started reports that
start. Every animation and set() promotes its channels to a compositing
layer in the same task (will-change, or a host custom property for a
pseudo-element) until 500 ms after the last motion: Mobile Safari only runs
a script-started curve through a block on an element that already has a
layer. Under Reduce Motion the substituted cross-fade can follow a
presentation's own spring, and crossfadeAs gives a slide that changes no
opacity its direction. The header lists the iOS Safari limits measured on
the simulator.
…eraction harness useAnimator wraps the animator for React without re-rendering per frame. Browser proofs on the interaction harness, in Chromium and WebKit on the phone profile: a spring keeps advancing on every screencast frame through a 1 s main-thread block and every frame sits on its analytic curve; a spring ending inside the block keeps its end state; no Layout or Paint runs while it animates; the engine plays the curve when seeked; a finished animation leaves within a frame or two of the curve's end; a dialog's ::backdrop follows the curve.
A touch presentation is now also an idiom: ios for WebKit on iOS or iPadOS
under a coarse pointer (CSS.supports('-webkit-touch-callout: none'), which
no other engine implements), default everywhere else, Android included,
which keeps today's behaviour until a Material variant exists.
useTouchPresentation and readTouchPresentation report it, useTouchIdiom
reads it alone, and useMotionPresentation now chooses UIKit motion only in
the iOS idiom. Styles gate the same way with two nested StyleX keys, the
media query outside and the supports query inside (IOS_TOUCH_STYLE_KEYS),
as core's 16px input floor already does; a test keeps every such key in
core spelled and nested that way. A real-engine spec reads Chromium as an
Android phone and WebKit as default.
Every spring, duration (a key ending in Ms) and curve of a motion table is
a motion role named by its table path (springs.sheetPresent,
timing.curves.easeInOut, sheet.present). defineTheme({motion}) takes
springs, durations and curves keyed by role, at the root or in an
adaptation rule, and writes one CSS variable per role it sets
(--motion-springs-sheet-present: 0.6s 0.1); a role it leaves out keeps the
table's measured value. The duration scale becomes all-or-nothing so a
theme can set roles alone. Adaptation rules gain a platform condition (ios
or other) that lowers to @supports (-webkit-touch-callout: none) around the
media block. Gesture thresholds and the arbitration timings (the
content-touch delay, the scroll brake) are not roles.
The motion docs carry a table of every role with its CSS variable and
default, generated from the tables (pnpm generate:motion-roles, checked in
check:repo).
# Conflicts: # packages/core/src/motion/README.md
…me sets it resolveMotion(element, role) reads a motion role's CSS variable from the element's theme scope (the nearest [data-astryx-theme], or the root) when a motion starts: a style read, no layout, cached per scope until a theme attribute changes, theme CSS is inserted or edited, the window resizes or a pointer, contrast or reduced-motion preference changes. An empty or malformed value, a detached element or a value that is not a role gives the table's value back. The animator resolves every spring it is handed, the Reduce Motion fade's duration, curve and spring, and a decay's edge bounce, so call sites keep passing table entries. A browser proof seeks two boxes animated with springs.sheetPresent, one inside a theme that sets it: each plays its own spring within 0.05 px in Chromium and WebKit.
…tal touch exports
…tations @astryxdesign/core/motion/experimental exports the compositor animator (createAnimator, useAnimator), the spring, rubber-band and cubic-bezier math, and the UIKit constants table as the read-only uikit namespace. @astryxdesign/core/gestures/experimental exports the gesture recognizers, useGestureRecognizer, usePressTracker, the touch presentation resolver and playHaptic. Both follow spec:AST-017's experimental boundary: never re-exported from the package root or an ordinary subpath, @experimental on every exposed declaration, stability: 'experimental' docs. The internal gestures barrel gets no subpath.
WebKit dates a pending animation created before busy work in the same task (a React commit, layout effects) to the frame before that work, so the first frame after it already shows most of the curve. A call from rest without an explicit start now promotes its element in the calling task and creates its animations in the next animation frame, which starts them; handle.started reports that start and the curves read from the call time until then. A set, stop, explicit start, rebase, attach or dispose before that frame creates the held animations first, in call order. The browser proof of a spring started right before a block now runs, and passes, in WebKit too. The promotion timer no longer rereads the clock when it fires: a test that fakes timers but not the clock could keep it rescheduling forever.
vjeux
requested review from
cixzhang,
imdreamrunner and
josephfarina
as code owners
October 24, 2026 16:47
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
|
AI review status for this pull request.
|
This was referenced Oct 24, 2026
meta-cla
Bot
added
the
CLA Signed
This was referenced Oct 24, 2026
![astracat-bot[bot]](https://avatars.githubusercontent.com/in/4796231?s=60&v=4){kind=small}
There was a problem hiding this comment.
This change is on hold for an engineering owner's decision on draft spec:AST-094, the only spec it and its stack cite: the record is unapproved, its FR5 does not state the start-at-first-frame rule this fix implements, and its FR4 and DEC-1 say touch motion is not themable while the stack ships theme motion roles. An engineering owner records that decision by amending the record as needed and commenting /approve-spec <full head SHA> on this pull request; the author may do so as an engineering owner. Independently of that decision, two changes are needed:
- Reduce the diff to this pull request's own commit once its predecessors land; it currently carries about 28,500 authored lines from the whole stack.
- Drop the
platformadaptation condition, or amendspec:AST-012first: its FR2 limitswhentowidth,pointer,contrastandmotion.
[Automated review]
| * is everything else. Pair `ios` with `pointer: 'coarse'` for a touch | ||
| * profile, such as touch motion tuned per platform. | ||
| */ | ||
| platform?: 'ios' | 'other'; |
There was a problem hiding this comment.
A platform condition adds a fifth adaptation field, but spec:AST-012 FR2 limits when to width, pointer, contrast and motion and says future environment fields need an explicit addition to that closed vocabulary. Please drop this axis (and its @supports output) or land an AST-012 amendment first.
| measures (0.55 for a scroll view and a sheet), and a release MUST pick its | ||
| rest point by projecting its velocity the way its UIKit counterpart does: | ||
| with a deceleration rate, or with a measured projection time. | ||
| - **FR4 — Physics are not themable in touch presentations.** A touch |
There was a problem hiding this comment.
FR4 says touch motion is not themable, and DEC-1 rejects tuning touch motion per theme, yet the stack ships defineTheme({motion: {springs, durations, curves}}) for exactly these roles. One of the two needs to change before this record can be approved.
|
|
||
| ### Motion | ||
|
|
||
| - **FR5 — Motion runs on the compositor.** Every release and every |
There was a problem hiding this comment.
The pull request cites FR5 for a motion starting at its commit, with a busy main thread delaying it rather than skipping into it, but FR5 as written covers compositor playback through a main-thread block, not when a motion from rest starts. If that rule is intended, it needs its own requirement.
| expect(element.style.translate).toBe('300px 0px'); | ||
| }); | ||
|
|
||
| it('a set, a stop or an explicit start before that frame creates the held call first', async () => { |
There was a problem hiding this comment.
This test is named for a set, a stop or an explicit start, but it exercises only the explicit start, and it also passes against the previous animator. The flushes in set, stop, rebase, attach and dispose have no test.
vjeux
added a commit
to vjeux/astryx
that referenced
this pull request
Oct 24, 2026
4 of 8 tasks
Contributor
PR Analysis Report📚 Storybook Preview🧪 Sandbox PreviewModified ComponentsContextMenu (@astryxdesign/core) · View in Storybook
MoreMenu (@astryxdesign/core) · View in Storybook
MultiSelector (@astryxdesign/core) · View in Storybook
Bundle Size Summary
Accessibility AuditStatus: No accessibility violations detected. Visual Regression46 of 496 shot(s) changed. A change here is a question, not a failure: check whether the after is the
and 26 more. Generated by PR Enrichment workflow | Storybook | Sandbox | View full report |
Contributor
Author
|
Superseded by #8340, which makes a start on a presented frame an opt-in ( |
This branch was successfully deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Depends on #8076 (and its predecessors); review the last commit.
User impact
On iOS, a motion started right after a React commit (a row removed, a node
expanded) often appeared most of the way through: the first frame after the
commit's work already showed 36–76 % of the travel.
Problem and solution fit
to the frame before that work, so the curve runs while nothing paints.
A call from rest without
{at}now promotes in the calling task (as before)and creates its animations in the next animation frame, which starts them.
Expected behavior and authority
spec:AST-094FR5: a motion starts at its commit, and a busy main threaddelays it rather than skipping into it.
Smallest restoration
animate,playanddecayfrom rest create in the next frame;handle.startedreports that start; the curves read from the call timeuntil then, as before.
set,stop, anexplicit start,
rebase,attach,dispose) creates the held animationsfirst, in call order.
{at}stay immediate.faking timers but not the clock kept it rescheduling).
Evidence
spring, then 600 ms of work in the same task. The first frame after
the work shows 36–76 % of the travel in 6 of 17 runs.
The small square flashes while the main thread runs and freezes during the
work.
plays in full after it" now passes in WebKit too (first moving frame 1 px in
both engines); unit tests cover next-frame creation, the flush, and the
promotion timer.
Scope
Testing
pnpm test:interaction packages/core/src/motion: Chromium 8 passed; WebKit (container) 7 passed, 1 skipped (tracing is Chromium-only).pnpm exec vitest run --project ui packages/core/src/motion packages/core/src/gestures: 34 files, 594 tests passed.Stack (merge in this order): #8316 spec → #8053 constants table → #8058 → #8059 → #8060 → #8061 gesture recognizers → #8054 math → #8257 theme motion roles → #8055 animator → #8056 animator curves → #8067 browser proofs → #8258 theme motion in the animator → #8071 → #8072 → #8073 → #8074 gesture system → #8076 experimental subpaths → #8266 start in frame (#8067 also follows the interaction harness, #8043 → #8044 → #8045; #8062 merges any time after #8316).