The canvas kept the art of an image whose own fill or stroke follows the
outline of its pixels for as long as the image object stayed the same. When
the document's pixels for the image changed under the same key (Update Link
replacing a preview with the linked file's pixels), the stroke kept the old
outline while hit testing and Use Preview Bounds used the new one. The
renderer now keys that art by the document's pixels for the image too.
- vectorcraft-effects: `follows_image_pixels` names such images.
- vectorcraft-render: `reads_outside` counts them, and the resources key
hashes the address of the document's pixels for them; their walk is not
kept between frames.
- A test places an off-center opaque area in a smaller copy of the pixels
under a turned and reflected placement.
- Docs: the Outline Object params text names every case that keeps the frame
and the 50% cutoff as VectorCraft's own; docs/mcp.md's `effect.apply` entry
for Outline Object, how long an outline is kept and the 16-bit RGBA limit;
docs/target-app-parity.md's evidence lines.
Rewrap the prose in docs/plugins.md whose lines broke well short of
the file's 120-column width or ran past it, including the Safe Mode
sentence, the permission-name and install-confirmation sentences, and
the grants.json paragraph. Each rewrapped line is at most 120 columns.
No words, punctuation, code blocks or tables change.
Type in a font found by name with a predefined non-Unicode CMap, such as
`/Ryumin-Light-83pv-RKSJ-H findfont` or a `composefont` with `GBK-EUC-H`,
opened as Latin-1 letters (`ú{A` for 日本A) in a font whose style carried
the CMap name (`Light-83pv-RKSJ-H`). Only Unicode CMaps were split off the
font name, and `composefont` failed with `undefinedresource` for the others.
- The predefined CMap decoder used by legacy `.ai` type moves to
`import/predefined_cmap.rs`, and EPS import decodes these CMaps' codes
with it (Shift-JIS, EUC, GBK, Big Five, UHC and the other public CMaps).
- A font name splits at a predefined CMap unless an installed face has
the whole name. A CMap that can't be read gives a warning.
- Type over a vertical CMap (`-V`) is set as vertical type, and each code
moves the point one em down the column, the default `DW2`
displacement.
- Every show operator sets decoded type with the same spacing and shaping
as other fonts found by name.
- Decoding a string counts against the import's memory budget.
An image's own fill or stroke with Outline Object now follows the outline of
the image's pixels that are at least 50% opaque, traced from the alpha and
fitted with curves, instead of the image's frame. An image without
transparency, one whose pixels can't be read, one of more than 64 megapixels
or whose decoded pixels take more than 256 MB, and one whose outline would
need more than 100,000 anchor points keep the frame. An image with no pixel
at least half opaque gives nothing to paint. Images of more than 4 megapixels
are traced from their alpha averaged down to 4 megapixels, and an averaged
pixel that covers a pixel at least half opaque stays at least half opaque, so
thin opaque lines stay in the outline. Each outline is traced once for its
pixels, kept for the next redraw while those pixels are loaded, and dropped
after them; a second thread that needs the same outline waits for the trace.
- vectorcraft-trace: `decode_alpha` reads an image's alpha, and
`opaque_outline` traces it as Black and White mode at threshold 128.
- vectorcraft-effects: the reshaped image art uses it for rows whose first
effect is Outline Object; Convert to Shape rows keep the frame.
- Docs: docs/mcp.md, the Outline Object params text and an evidence line in
docs/target-app-parity.md. docs/mcp.md also says that Use Preview Bounds
takes in the shapes an image's fills and strokes paint.
A selected image offers Remove Background in the Contextual Task Bar, Properties'
Quick Actions and the Control bar, as the reference app does. A click runs it at
once with its defaults, as one undo step; Object > Plug-ins > Image Tools keeps
the dialog with Preview.
Pinning a closed replaced spine opens it with a copy of its first point
at its end, so a two-point ring became three points and Delete Spine
Point could leave a spine of two points in one place. The two-point
check now counts the points left that are apart.
Six tests cover the Direct Selection marquee (anchors and crossed segments), its pointer and anchor highlight, and the Group Selection click, repeated click, marquee (default and E for enclosed objects) and hover over a drop shadow: they reach the object under the shadow or nothing. The selection passes-through line in docs/target-app-parity.md names these gestures.
A Warp effect on an image, or on a group that holds one, bent the image's
frame but drew the pixels unbent, cut by the bent frame. The image now
bends with the frame as a raster mesh warp, as an envelope bends it: after
the last effect in the list, the image is cut into 10 x 10 cells of two
triangles, each a clipped piece of the image under its own affine map,
through every Warp it took and the Transform moves between them. Effects
after a Warp reshape only the frame, and a placed gradient on a group
that holds the image follows the group's bounds as the pieces draw them.
The image itself is unchanged, so editing or removing the warp redraws
it. The images stay unbent when their pieces would make the art pass 2
million objects, and an image stays unbent when its clipping frame keeps
its shape because the warp would pass the effects' anchor budget.
- SVG export writes the image once and draws the pieces with <use> when a
group holds pieces of one image clipped by triangles. PDF, EPS, .ai and
PNG exports write the bent pieces.
- Envelopes: a grayscale image's fill color now colorizes the pieces an
envelope cuts. Before, the pieces drew the gray image.
- Geometry effects: an image under any geometry effect keeps its fill
color and its smoothing setting. Before, it drew the gray image,
smoothed.
Docs: Warp params text, docs/mcp.md, docs/target-app-parity.md.
Horizontal type keeps both marks, and the legacy .ai writer writes such
type as text, but the marks were dropped without a warning. Version 7
text has operators for both (section 15.3.4 of the published format
specification): Tv for Roman characters upright in vertical writing,
and TY for a Kumi block in horizontal text recorded for later vertical
use. The writer now writes 1 Tv and 0 Tv, and 1 TY and 0 TY, where a
run's mark changes, closes an open Kumi block before the text object
ends, and writes neither operator for type without the marks. The
legacy .ai reader reads both back onto the same characters. The prolog
defines Tv and TY so the file still runs as PostScript.
With automation read roots in force, links.findFiles and text.findFontFiles
replies, including later polls and stops of a search another client started,
leave out the files found outside the caller's read roots, a sought link's
saved path outside them and the folder searched when it is outside them. A
failed search then gives its reason without the path. The rows keep their
names, ids and status, and the counts stay.
image.info gives link: null for a linked file outside the read roots, as
links.info does. text.addFontFiles gives folder: null when the Fonts folder
is outside the read roots and names the copies there by their file names;
the fonts are still copied.
Docs: docs/mcp.md, docs/target-app-parity.md.
dotCooCoo