* hyphenation: linear heuristic, cache misses outside the lock
The syllable heuristic built each word suffix as a String to test it, which is
quadratic in the word's length; it now uses prefix sums of vowels and UTF-8
lengths. Mapping a word's break points to glyphs scanned the word for each
point; both are indexed once per word. A word not in the hyphenation cache is
worked out without holding the cache's global mutex.
* hyphenation: compare the word ending with ends_with
A cached story copied every wrap neighbour's whole signature and item list into
its entry; stories on one spread share its wrapping items, so entries grew with
all of theirs together. A neighbour's signature now goes in as its hash and its
items are kept once each.
Anchored-frame placement asked for the host's composition once per frame, and
again at every level a host was reached from. Each host is now asked for once
per placement pass, and the cache and the uncached path remember the hosts they
resolved for the whole traversal.
- Footnote references on a line are deduplicated with a set instead of
searching the references collected so far.
- Nested styles stream the paragraph up to their delimiter with one character
of lookahead instead of copying the rest of the paragraph for each style,
and a delimiter's characters are indexed once per style.
- Shaping resolves which nested/GREP style is on top with one sweep over the
overlay boundaries instead of scanning every overlay for each stretch.
- Column rules group crossing lines by paragraph through an index.
- Table fragments index each row's top once instead of searching the rows
for every cell.
A relative link names another file once the document is saved elsewhere,
so a status remembered for the old file must not answer for the new one.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A fontFamily the app doesn't have was applied with a null result, so a
script that asked for Courier Prime exported the PDF in the substitute
and exited 0. The font is still applied, since documents can carry
missing fonts and Find Font and preflight list them, but type.char now
returns {"warnings": ["Missing font: Courier Prime (using Source Sans 3)"]},
or "Missing font style: ..." for a style the family doesn't have, in
preflight's words. Other calls still return null.
The drop cap is set at the indent and the list number beside it, so the
number's character is the later one on the line and sets the indent.
Pin that with a test, and say so where the character is looked up.
The character was looked up among the line's body glyphs, so one inside
the drop cap was missed, and one in a list number's ^i matched the first
body glyph (after the label's tab) instead of itself. Find it among the
placed glyphs, which include the drop cap's and the label's.
The lines after an Indent to Here start at the character, which already sits past the drop cap, but the lines beside the drop cap still added its width on top. They no longer do. Adds tests for the drop cap, for Indent to Here after a tab (the usual hanging number) and for a right-to-left paragraph, and documents how the indent behaves beside a text wrap.
links.list decided a placed graphic's status by reading the whole file and comparing it with the asset's bytes whenever the sizes matched, and the Links panel polls it twice a second, so a document of press-size images read every one of them in full on every poll. Image links now have the size-and-time fast path linked text files already had: a file of the remembered size and time isn't read, a new time is read once and remembered (unless the file changed in the last two seconds), a different size is modified without a read. The answer is keyed by the asset's bytes' identity, so a relink, update or undo can never be served a stale one; status() stays as the exact check for callers that want it.
Verified locally on a288a87b: fmt, clippy, engine link tests including the new unchanged_linked_images_are_not_read_again.
The composer treated Indent to Here as an invisible zero-width character and nothing else, so inserting it changed nothing. A line holding one now sets the start-side indent of the paragraph's following lines to the character's position, and those lines are broken again with the narrower measure. The indent holds into the next columns and frames; a later Indent to Here in the paragraph replaces it.
The picker opened with R selected every time, so anyone who works in H (the hue slider with the saturation/brightness field) chose it again on every open. The component now lives in the UI preferences (colorPickerComponent, default r, so existing preference files and openers that name a component behave as before) and the picker reads it on open and writes it when a radio button is clicked, as Photoshop's picker keeps its last mode.
Verified locally on a288a87b: fmt, clippy, the picker's tests; tried on macOS.
A centred dialog window offers its body only last frame's height, and the Color Picker lays itself out on a fixed 566 x 334 canvas, so it opened clipped to its top strip behind a scrollbar: the colour field cut off, no value fields, no hex code, OK and Cancel visible but blind, and the window not resizable. The body's minimum scrolled height now covers the picker's canvas, as it already did for Numbering & Section Options.
Verified locally on a288a87b: fmt, clippy, the picker's own tests; opened on macOS at full height.
A sheet draws its spread's items without asking which page each one sits on, so
a frame straddling two pages is drawn, in part, on both. The warning chose the
page from the frame's centre alone (`page_at_x(bounds.center().x)`), so a frame
centred on a page the export leaves out had its warning dropped while the part
on an exported page went into the file (noted in review).
`pages_of` gives every page of the spread the frame's bounds overlap, and the
story is kept when any of them is exported.
The page named is one the export has: the centre page when that is exported, so
an ordinary frame reads exactly as before, otherwise the first exported page the
frame reaches. A frame wholly on the pasteboard overlaps no page's span and
falls back to `page_of`, so the two cannot disagree about where such a frame is.
`overset_location` returns the reached pages beside the centre page, which is
what made it worth a named type.
The test exports a frame centred on page 3 and reaching back into page 2, four
ways: the whole document, page 3 alone, page 2 alone -- which used to report
nothing -- and page 4, which the frame never reaches and which still reports
nothing.
Found while testing crates/geom/src/path.rs:~335 (PathData::bounds)
Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
A Cover Image field with a Choose… button (PNG or JPEG) in the EPUB Export Options dialog; left empty, the first page stays the cover. The three labels are translated; a dialogs test exports with a chosen file and checks the EPUB's cover image and manifest.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Found while testing crates/engine/src/cmd/datamerge/parse.rs:212 (parse_record_no, used by select_range)
Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
winit 0.30 gives an undecorated window WS_CAPTION and WS_BORDER; on some systems the compositor
reserves that frame, so the content is drawn offset (black bars at the top and left) while the
pointer keeps the client coordinates. A per-frame guard strips the styles again after every
winit flag change and keeps a maximized window clear of the taskbar. Fixes#1295.
Merge origin/main at a288a87b. Preserve the font.list progress entry and the content-first routing note. The feature diff remains the same four reviewed files.
Co-Authored-By: OpenAI Codex <noreply@openai.com>
Keep native ZIP, legacy JSON and IDML bytes out of the INDD converter even when a browser supplies a misleading .indd or .indt name. Preserve signature and unknown-data routing to the converter. Check that every browser delivery adds a document before asserting its title.
Co-Authored-By: OpenAI Codex <noreply@openai.com>
aiongg
Dylan M. Taylor
Brandon Thomas
Brandon Thomas
JonRC
ljkx
Paul Gemignani
NetterSilverandClaude Opus 5.5
FredApps
Hamza
Trevor BrownandClaude Fable 5.1
Tom MainandOpenAI Codex