adminandClaude Opus 5 3ff7e91dbf Vector art points have four types and the pen draws whole curves (#195) ...

The pen only pulled out the outgoing handle of a new point, because
the opposite handle was mirrored only if it already existed. The
first segment of every dragged point was therefore half a curve.
Beyond that bug, a point had no way to say how its handles relate.

- A vector point has an optional mode: straight, corner, smooth or
  symmetric. Missing means corner. The mode is not part of the
  fingerprint, because it changes no pixel, only the next drag.
- Dragging a handle follows the mode. Corner leaves the other handle
  alone, smooth turns it along and keeps its length, symmetric
  mirrors it exactly.
- The pen pulls out both handles mirrored and marks the point
  symmetric. A plain click still sets a point without handles.
- The tab header has a "Point: ..." menu for the last clicked or set
  point. That point is drawn filled. Changing its type is one undo
  step. Text instead of four icons, because the UI kit has no
  tooltips.
- Switching a point without handles to a curved type first grows
  handles along its neighbours. Smooth and symmetric line the
  handles up and keep the rough direction of the curve.
- A point inserted on a curve is smooth. Symmetric neighbours become
  smooth, because the split shortens one of their handles.
- Alt-click still toggles between straight and symmetric.

Found on the way: the old Alt rounding took the point at the other
end of an open line as a neighbour, so its handles came out skewed.
Fixed and held by a unit test.

Existing points with handles now count as corner. Before, every drag
mirrored. No pixel changes, only the next drag behaves differently.

Tests:
- ./check.sh: 129 files, 1917 tests, exit 0 (before 1903)
- GM_STEPS=47, freshly seeded: ALL PASSED, 39 checks (before 32)
- Check sum 1195 in README and wiki is computed (1188 + 7), CI
  measures it
- Counter-checks: pen without 'symmetric' fails 3 checks, no smooth
  branch fails 1 unit test and 1 e2e check

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

2026-09-21 06:51:35 +02:00

..

core

Vector art points have four types and the pen draws whole curves (#195)

2026-09-21 06:51:35 +02:00

editor

Vector art points have four types and the pen draws whole curves (#195)

2026-09-21 06:51:35 +02:00

ui

Snap is a magnet button in every graph and vector art tab (#194)

2026-09-21 06:25:56 +02:00

viewport

A picture in the editor may now be drawn instead of uploaded, which is the piece #129 (Vector art) and #130 (Pixel editor) both need and neither could build alone: TextureRef.file has been a file name on the server, isTextureFile() checking a sha256 hash plus an extension and TextureLibrary fetching it over HTTP, so an asset that paints its own image had nowhere to put itself - the key now also takes the form gen:<assetId>@<revision>, which an uploaded file can never look like, and everything that only passes a ref along (sprite, tile, tilemap mesh, particle source, shaders, aspect ratio) is untouched because it still sees a Texture; the revision is a string, not a counter, which is the one deviation from the plan in the issue and costs nothing - gen:a1@7 and gen:a1@c0ffee go through the same door, so a generator may fingerprint its own content instead of relying on every command to remember to bump a number; two registries keyed by asset type as a plain string, so a plugin may bring its own type - registerPictureSource(type, read) in the new core/model/generated.ts tells the model canvas size and revision and resolvePicture() builds the ref from it (no rect, no slice, smooth from props.filter like a texture), and registerPictureGenerator(type, draw) in the new viewport/generatedPictures.ts hands out the finished texture, projectPictureFactory(() => state.project) tying the two together per session in App.ts; invalidation is a rule instead of guesswork at twenty call sites: GeneratedPictures<T> holds one picture per asset keyed on the revision and frees the old one as soon as the new is drawn, TextureLibrary destroying it and dropping the crops cut from it, while a generator that cannot draw yet is not remembered but asked again next time - the class imports Pixi as a type only, so the whole bookkeeping is measurable without a GPU; carried along are the picture picker (a drawn asset sits among the textures by name, without frames), the delete warning (counts tiles and animations for it too) and the asset context menu, where "Create sprite from picture" now covers it while "Create animation from texture" stays texture-only because it needs frames; one finding while probing - crop() could not have cut a generated texture at all, a RenderTexture having no source.resource to draw back from and drawImage running onto null, so it is handed out whole, a drawn picture having no sheet whose neighbouring row could bleed in; 18 new unit tests, 10 in tests/core/generated-picture.test.ts (building and splitting the key, an uploaded file is not one, no registry means no picture, resolving to canvas size, a new revision is a new key, a canvas without a size, the filter, the sprite size, picker order) and 8 in tests/viewport/generated-pictures.test.ts (drawing once, freeing the old revision and only for its own asset, null is not remembered, no key means nothing drawn, a session change gives everything back, peek), and three counter-checks hold: dropping generated assets from resolvePicture fails 5 of the 10 core tests, not freeing the old revision fails 1 viewport test, and narrowing the picker back to textures fails 3; no e2e check and the check sum stays 1157, because without a generator the groundwork shows nothing in the browser - that evidence comes with #129 (#185)

2026-09-20 20:09:42 +02:00

helpers.ts

The new node "Raycast all" answers with every collider along the ray instead of only the nearest one: it carries the same inputs and settings as Raycast (Ignore owner, Hit triggers, Layers from #158) plus "max hits" where 0 means all, and it hands back parallel lists sorted by distance - objects, colliders, points x/y, normals x/y, distances - next to count, so index i of every list speaks about the same hit; a collider counts once no matter how many convex parts its shape is built from, while a tilemap counts per cell it is hit in, and the filter sits in Solids.raycastAll like the one of #158, the one path the editor run and Play both take, so all 12 new unit tests run through describe.each against both systems; the node needed the value system to learn something it could not do - a graph list carried numbers, text, booleans and bundles but no object handles, asList turned an object into its name, so objects/colliders would have been lists of names, which loses the object when two carry the same name, and collider children usually do; Value now knows a travelling list that may hold handles and "Item at" gives such an item out as an object, ready for Destroy, Apply force or Name of, while the single door into a stored file - writeTyped/asTypeOf - runs the list through the new storedList() so every object becomes its name again and nothing living reaches a project file, which is its own unit test over JSON.stringify; three list nodes follow the handles: "Index of" compares objects by identity because a name is not the object of that name, "Set item" and "Fill" write an object as an object, and "Join text" prints the names instead of the innards; this decision was mine to make because the question ran into the MCP timeout - the alternative was a list of names plus "Object by name", which is the smaller change but lossy; 4 new e2e checks in the new block 43e (shield and floor in one go reading 2 and "145, 280", the second object answering to its name, which only a real object in a list can do, "max hits" 1 stopping at the nearest, and the node showing the row) and 18 new unit tests, 12 in raycast.test.ts, 4 on the value system and 2 in the node catalog; two counter-checks hold: not counting tilemap cells one by one drops 2 unit tests, and answering with the nearest hit only drops module 43 with 2 FAILs; one finding on the way - the node read "maxHits" while every other input carries a spoken label, because int() took the key as its label, so it has a third parameter now, and the e2e check caught that, no unit test did (#157)

2026-09-20 18:54:12 +02:00