fix: cleanup API surface - #1
Merged
Merged
Conversation
Unused default Storybook install (config, sample stories, assets). Nothing imported it and it was not wired into any script. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Drops the Storybook devDependencies, eslint plugin and vitest project left behind by the scaffolding removal. Adds a single `verify` script (check + lint + test + size budgets) and a GitHub workflow that runs it. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Tabs + single quotes to match the rest of the repo; no behaviour change. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Root entry now re-exports every module, not just animate + flip, so the old 12 kB budget measured the wrong thing. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Fourteen module READMEs and INDEX.md restated what the TSDoc and the root README already say, and drifted every time the API moved. One document to keep in sync instead of sixteen. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
…elpers Three concerns were duplicated across attachments, JS-driven animations and the property registry: - frame-batch: one rAF per burst of subscribers instead of one each - frame-tween: shared delay/completion/cancellation lifecycle - inline-style: save/restore of a forced inline property + !important Also drops the shared/index.ts barrel — callers import the module they need. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
animate/flip.ts reimplemented what flip/ already does with a richer option set; its exports (flipFromRect, flipToRect, captureRect, AnimateFlipRect) are gone — use @ixirjs/pulse/flip instead. animate/spring.ts was a second copy of shared/spring-core, now re-exported from there. style-utils moved to shared/inline-style. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Helpers with no callers outside their module lose their export; adds a transform-tracker browser test covering concurrent transform components. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Forward FLIP that animates an element from its current box to a target element's box, for popovers/tooltips that adopt an anchor's geometry. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
- geometry lives in one module instead of being split across animate/ and flip/ - cancel-controller folded into controller-slot; scope into bridge; observer-manager + scheduler into tracking/observers - delta math (computeDelta, isIdentityDelta, diagonal, rectsEqual, FlipDelta, DeltaOptions) is no longer exported — it is an animator internal Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
flipTo was cut as dead code while flipFrom survived, leaving the surface with a 'from' half and no 'to' half. It is a four-line forward-FLIP wrapper over animateFlip; the symmetry is worth more than the lines. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The nearest-slot search and ascending-order check are pure functions that were buried in the attachment; testing them no longer needs a browser. Binary search for ordinary linear lists, linear scan for wrapped/grid layouts. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
planMorph, interpolatePlan, toPathString, subdivideTo, alignSubpaths, minimizeAnchorTravel, rotateClosed, reverseClosed and MorphPlan are no longer exported. morph, parsePath and normalizePath are unchanged. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
…ernals Same surface trim as the other modules: helpers with no callers outside their own file stop being part of the package API. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
…t scroll index Directory imports resolved fine in this repo but not in consuming Vite projects; easing was reachable only via the subpath export. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Interactive playground for the new modules; app.html applies the stored or system theme before first paint to avoid a flash. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
…llbacks Two duplications kept reappearing across modules: every attachment hand-wired addEventListener/removeEventListener pairs (one cast per handler, one mirrored removal that could drift), and both animation controllers repeated the same pause/play/reverse/seek fan-out over a group of WAAPI animations. - `shared/listen` attaches a typed handler map and returns the exact teardown. - `shared/playback` owns the transport controls, including the swallowed throw when seeking a cancelled animation. - `isBrowser()` checks only `document`; a `window`-less DOM is not a case we support, and the extra clause implied it was. - `createFrameBatch` no longer falls back to a synchronous callback when `requestAnimationFrame` is missing. Tests run in a real DOM, so the fallback only masked scheduling bugs. Sampling assertions that relied on the extra synchronous frame drop from >3 to >2 samples. `scroll()` moves to `listen()` as the first consumer. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Both controllers drop their hand-rolled pause/play/reverse/seek in favour of the shared fan-out. `wireTransform()` pairs the two setup calls every caller already made together (`ensurePropertiesRegistered` + `ensureTransformWired`), so no caller can register the properties and forget the element's chain. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
`ease`, `easeIn`, `easeOut` and `easeInOut` are exactly the CSS keyword curves, yet `easingToCss` resampled them into 25-point `linear(…)` approximations of timing the browser already implements natively. Tag each with its keyword — mirroring how `springEasing` carries `_linearEasing` — and prefer that tag. Exact timing, smaller keyframes. The resampling tests move to `backOut`, which genuinely has no keyword form. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The types file held only the `Timeline` / `TimelineDefaults` interfaces that `timeline.ts` implements and re-exported anyway. One file, one indirection less between a method's docs and its body. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
`createVariants()` and the `variants()` attachment that drives it were split across two files for no reason beyond the `.svelte.ts` suffix — which the merged file needs anyway. The index now re-exports from a single module. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
`size()` and the `width`/`height` fields of the other presence transitions both grew a box from a start fraction to its natural footprint, but each carried its own metric snapshot and its own declaration list — the `size()` copy spelled all fourteen properties out by hand and could drift from `AXIS_SPACING`. Both now go through `axisMetrics()` + `growDecls()`. `size()` keeps its fraction-valued `start` (no `resolveSize()` round-trip) by building its plans directly. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Every gesture attachment repeated the same three things: a cast-per-handler
addEventListener block mirrored by a removal block, the `touch-action` save /
set / restore dance, and — in `draggable` and `swipe` — an inlined copy of
two-sample pointer velocity tracking.
- All of them now wire events through `shared/listen`.
- `lockTouchAction(element, axis)` joins the pointer-capture helpers, since it
is the same "hand this axis to us, put it back on teardown" concern.
- `trackVelocity()` holds the velocity model once, with the reason two samples
is the whole model written down where it lives.
- `release()` drops its `hasPointerCapture` guard — the try/catch it sits
inside already covers the only case it could catch.
- `reorder-internals` folds back into `reorder.ts` as a single linear
`nearestCenterIndex`. The binary-search path plus its `centersAreAscending`
precondition was real complexity for lists of a dozen elements; the test
moves alongside it.
**Breaking:** `disabled` is gone from every gesture attachment. It only ever
short-circuited setup, so flipping it did nothing to an already-live
attachment; conditional attachment (`{@Attach on ? draggable() : undefined}`)
is the form that actually tears the listeners down.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
`auto` asked the caller to do the library's job. `auto: () => { void open }`
read the state, discarded it, and left flip to re-measure a frame later — and
the `ObserverManager` variant made an internal a public type just so it could
be passed back in.
Instead the attachment owns the attribute:
<div {@Attach flip({ class: () => ({ 'is-open': open }) })}>
The thunk is tracked, so flip writes the class (or inline style) itself and
measures `from` → write → `to` → animate in one tick. Writes are diffed against
what flip previously applied — never a wholesale `className` / `cssText`
assignment — so markup classes and the animator's own `--motion-*` writes
survive. Style strings go through the platform's parser (a detached element),
which handles `!important`, comments, and malformed input for free.
Layout tracking (`ResizeObserver` + parent `MutationObserver`) is now
unconditional, so a shuffled `{#each}` animates with no options at all.
Also in this pass:
- Interrupted runs converge. A burst of retargets (drag, rapid shuffles) used
to restart the full duration on every trigger and never settle; the slot now
charges elapsed time against the replacement, floored at 120ms.
- `carryVisualOffset()` re-expresses a live visual rect in the previous layout
frame. An in-flight transform is an offset from the resting box, so once
layout changes the raw visual rect belongs to the *new* box — using it
directly made an interrupting FLIP shoot out of place.
- The skip counter only counts cycles that would actually animate, so
`skip: (n) => n === 0` lands on the first real move rather than the no-op
callback observers fire on connect.
- `flip/animation/`, `flip/integration/`, and `flip/tracking/` flatten into
`flip/`. Three directories held one to four files each and every import paid
for the nesting.
- `DEFAULT_DURATION` → `FLIP_DEFAULT_DURATION`, unambiguous next to the
animate-side default; the layout bridge drops its `performance.now` fallback.
**Breaking:** `auto`, the `FlipAuto` type, and `createObserverManager` /
`ObserverManager` are removed from the public API. Delete `auto:` from existing
call sites — tracking is on by default.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Adds a card for the new option and drops the now-unnecessary `auto:` thunks from the existing FLIP demos. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
`repository.url` gets the `git+` prefix npm expects, and the CHANGELOG release link points at ixirjs/pulse instead of the old svelte-atoms/vibra path. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
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
1 participant
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.
No description provided.