The outline cache kept a copy of each stroke's line to recognize a moved
path, and made that copy on every miss, so an anchor edit of a long path
paid for copying its whole line. The entry now records whether its outline
was made along a line of the object's own path. On a miss, the anchors of
the earlier object are compared with the new one's, stopping at the first
that differs, and the earlier line is made again from the object and its
stroke only when they match.
Strokes with arrowheads keep the reuse: the check uses the untrimmed path.
A shear, a wider stroke, another cap, another dash pattern and another
width profile each make one new outline and draw the pixels a new
renderer draws, for open and closed, short and long, plain, dashed,
profiled and gradient-filled paths.
The moved-outline key is imported where it is used, and the outline is
stored through one helper, so the change touches fewer lines that other
work on the renderer also changes.
A path object that moves without changing shape drew its stroke by
outlining the whole line again, because the stroke outline cache is
keyed by the node. The renderer now keeps, per object, stroke and
outline width, the latest outline with the line it was made from. When
the new line is that line moved, it draws the kept outline moved by the
same offset, and moves the region a long path was outlined in with it.
Scaling, rotating or editing the path, changing the stroke's width,
join, cap, dashes or profile, a zoom that changes the outline's
tolerance, and a move that takes the view past the outlined region
still outline the stroke again. Only the latest outline of each stroke
keeps its line.
A row drawn before was counted with no limit on its objects, so symbols
that each hold many instances of the next made the count grow with the
product of the instances, and the Layers panel could stall on such a file.
- Counting stops after 20,000 steps: each object, subpath, text run and
symbol looked at is one step, and every 64 bytes of a symbol or placed
file name looked up is one more. A row that needs more gets its
thumbnail on the thumbnail thread, and the painted thumbnail draws
nothing for it.
- A row drawn before counts at most 20,000 objects.
The Layers panel decided whether to make a row's first thumbnail on the
UI thread by counting path anchors and type frames only, so point type,
symbol instances, gradient meshes, images and placed documents counted
as one object with no anchors, and a heavy one was rendered on the UI
thread.
- small_node takes the row's document. Each glyph of type counts as 4
anchors, each mesh point as 1 and every 100 image pixels as 1, against
the same 2,000-anchor limit; glyphs are counted only up to the limit.
- A symbol instance and an inline graphic count their symbol's art, and
a placed document counts its file's art once the file has been read;
one not read yet goes to the thumbnail thread.
- Symbols and placed documents nested more than 8 deep go to the
thumbnail thread, so a symbol holding an instance of itself is counted
to an end.
- The painted fallback thumbnail counts as before.
A path or compound path whose Transform effect made copies expanded to
one compound path of every copy. It now expands to a group of one object
per copy, each painted as the path, as gradient copies already did; a
second Transform effect's copies nest in a group per copy of the first.
Drawing is unchanged.
# Conflicts:
# docs/effects-parity.md
A numeric field read a comma as a decimal point only when the text had no
point at all. A field shows its value with a point (100.5 pt), so typing
-20,5 after it gave 100.5 pt-20,5, which kept its comma and was refused.
Each number now reads its own separator: commas become points unless the
text reads as thousands (1,000, as before). A number written with both
separators (1.234,5, 1,000.5) ends up with two points and is still refused
rather than guessed at.
Locking or hiding the layer (or a group) around the isolated object left
its art selectable with the Selection tool: the isolation branch of the
hit test only checked for a locked artboard, and the marquee's checked
nothing. Both now pass over isolated art that isn't editable.
type_font_menu_lists_every_available_family read Type › Font while the
scan the app starts could still be running. The scan bumps the font
generation a moment before it counts as done, so a read in that gap lists
only the loaded fonts under the new generation, and the test's
"generation held still" check let the comparison with the full list fail.
The test now waits for the scan after making the app, as the native menu
tests do since #5653.
On a platform without a clipboard change count (macOS, Linux), a Type
copy that the system clipboard refused is kept only when the clipboard
held text when the write failed. A clipboard with no text, one that
couldn't be read, or one holding a picture can't show that another app
copied something later, so Paste no longer inserts the earlier copy
there.
When a fallback face draws Ka with the vowel sign I, the vowel sign's
glyph comes first in Ka's cluster, and text.glyphAlternates reported it
for Ka, laid out or overflowing. The query now finds the face that draws
the character and takes that face's glyph for it by the same rule as
for the requested face.
To find the characters that justification or Mojikumi moves, SVG export
without Fewer <tspan> Elements laid out every live type object a second
time, as point type with
Mojikumi None and default Justification options, whenever its paragraph
had Mojikumi on. New type gets Line-end Punctuation Half Width, so every
new type object was laid out twice. Mojikumi only spaces Japanese
punctuation and Japanese letters, so type without them lays out the same
either way.
- TextLayout::mojikumi_applies says whether the text or a list marker
holds a character Mojikumi spaces.
- reader_advances keeps such a paragraph's Mojikumi, and reads the glyphs
from the layout it was given when the object it would lay out is the
same; otherwise it lays the object out again as before.
The SVG is unchanged. Exporting 5,000 point type objects takes a median
of 109.4 ms instead of 171.5 ms on Windows 11.
When another app copies while VectorCraft's write is trying to open the
clipboard and the write then fails, Paste takes the other app's copy. A
change during the failed write counts as the write's own only when the
clipboard then holds this copy's text, or no text and either nothing to
paste or one of this copy's formats.
When the requested font has no glyph for a character, the text draws it
from a fallback face, and text.glyphAlternates reports that glyph. For
a character that overflows its area, the query shaped the character
alone but kept only a glyph of the requested face, so it reported the
requested face's default glyph (0). It now takes the first glyph drawn
for the character when the requested face draws none, as it does for
laid-out text.
A symbol instance whose fill has an opacity of 0 no longer counts as
stained. Expand Appearance and the other places that build an
instance's art used to rebuild every solid paint as an unlinked color
even though the tint changed nothing, so the art lost its global and
spot swatch links.
Ka, the halant, Ka again and the vowel sign I are shaped as one cluster:
the vowel sign, the first Ka's half form, then the second Ka's default
glyph. The glyph lookup looked for the character's default glyph across
the whole cluster before its alternates, so for the first Ka it found
the second Ka's default glyph instead of the half form. It again goes
through the cluster in order, and takes the first glyph that is the
character's default glyph or one of its alternates and is not the
default glyph of another character of the grapheme, so a vowel sign's
glyph that the font also lists as an alternate of the consonant is not
reported for the consonant.
A row sent to the thumbnail thread because its opacity masks pass the
2,000-anchor limit still paints the boxes of its own paths until the
thumbnail is ready. The progress line said such a row shows no picture.
Follows #5705.
A connection ended (a junk line, an idle or stalled peer, the app going away) by dropping its socket. On Windows, a reset can follow and discard replies the peer has not read yet, which made tcp_connection_is_closed_after_junk fail at random in CI with ConnectionReset. serve_socket now shuts the write side down and reads what the peer still sends until it closes, for at most a second and 1 MiB, before dropping the socket. The test goes through serve_socket, the path the app uses.
When a Copy in the Type tool cannot be written to the system clipboard
(another app holds it open, or a format fails after the clipboard was
emptied), Paste into type or a text field now inserts that copy, with
its style in type, instead of the text the clipboard kept. This covers
Edit > Paste, the paste key, and a paste that arrives while the copy is
still being written. When another app copies something, before or after
the failed write, Paste takes that.
The new SystemClipboard::change_count reports the clipboard's change
count (Windows' sequence number). Where a platform has none, a change
of the clipboard's text marks a new copy.
The enabled check of artboard.deleteEmpty built the full list of empty
artboards, measuring the visual bounds of the art against every
artboard, each time the Artboards panel menu was drawn. It now returns
false at once in a document with one artboard (that artboard is never
deleted). With two or more, it asks the new
Document::any_artboard_without_art whether any unlocked artboard has no
art, which is the same answer as a non-empty list.
Document::any_artboard_without_art builds the symbol boxes once for all
the artboards and first tests each path and compound path by a quick
box: its control points' box grown by the farthest its strokes reach
(Appearance::outset), with a margin for rounding. An artboard that no
quick box meets is reported empty without measuring along any path.
Only when every artboard may hold art is each one measured.
Document::artboard_has_art also skips measuring an object whose quick
box misses the artboard; its answers are unchanged.
The check of gradient.presets.save found the proxy paints through
Node::proxy_paint, which measures the object's geometric bounds for a
gradient box the check does not use. Node::proxy_paint_only picks the
same paint without measuring, and the engine's proxy_paint uses it, so
the Fill and Stroke proxies and the mixed-paint test no longer measure
each object either.
With an open stroked path of 200,000 anchors selected, the two checks
take under 0.01 ms instead of 21.9 ms and 17.2 ms in a release build, and
Delete Empty Artboards' check takes 0.85 ms instead of 42.1 ms with a
second, empty artboard.
Tests compare each check with what its command does (one artboard, all
empty, one empty, locked empty, nested art; with and without a
gradient), compare any_empty_artboards with the previous check and with
measuring every object on 120 seeded documents, compare the quick boxes
with the visual bounds on seeded strokes, and compare proxy_paint_only
with proxy_paint on seeded objects.
Docs: docs/target-app-parity.md, one line under Delete Empty Artboards.
text.glyphAlternates reported the default glyph as `chosen` for any
character that overflows its area. A character whose style substitutes
another glyph (`salt`, or `vert` in vertical type) was reported with a
glyph it is not drawn with. The query now shapes such a character alone,
with the rest of its grapheme in their own styles and in the text's
writing direction, by the same helper text.setGlyphAlternate uses.
The helper also shapes the character in the direction its paragraph
sets it in, from its bidi level, so a parenthesis set right to left is
reported with its mirrored glyph in both commands.
The renderer keeps the path of a path or compound path of 1,024 anchors
or more by object id, with the bounds of its segments and of its miter
spikes in runs of 128 segments. When anchors move and the subpaths and
anchor counts stay the same, the next frame rewrites only the path
elements of the segments that start or end at a moved anchor, and
measures again only the runs that hold them and the runs of the
segments with length before them. Other edits make the path again as
before. The pixels and the cull bounds are the same.
- StrokeLayer::bounds_on is unchanged. The miter tip of one joint
(miter_tip) and the stroke box with given tips (bounds_with_tips) are
new functions beside it, and Appearance::stroked_bounds_by takes the
box of each shaped stroke, so the cull bounds of a kept path use the
kept tips.
- geom::bez::has_length is the predicate reach.rs used for segments
that draw, made public.
Docs: docs/target-app-parity.md.
The anchor count that sends a row's thumbnail to the thumbnail thread now
includes the art of opacity masks that are not disabled, since the
thumbnail draws that art. A row drawn before was rendered on the UI thread
whenever its last thumbnail had been quick, even when its object had since
become a long path; it now goes to the thumbnail thread when the object
has more than 2,000 anchors.
dotCooCoo