Follow-ups to the review of the exporter plug-ins PR (#3171):
- Export As opens its format's options once the save panel has picked
the file. The Enter that closed the panel, still held, confirmed
them at once: Save As marks such a dialog (dialogs::AFTER_PANEL) so
it waits for Enter to be let go, Export As didn't. Every options
dialog it opens after the panel is marked now, an exporter plug-in's
included, through one helper (dialogs::opened_by_panel) that Save As
uses too.
- io::export with options that aren't a JSON object dropped the path
and format it adds for the encoder: an exporter plug-in was told the
document's name and its first extension instead of the file it
writes. Such options now count as none.
- The exporter UI tests remove the plug-ins they install, so no other
test's Export As lists them.
tests_export.rs reproduces the held Enter for every options dialog
(PNG, SVG, Save PDF, Text, DXF, EPS and an exporter's), and
tests_plugin_export.rs the options that aren't an object; both failed
before the fix.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HhvnnnD5fTFupK8WWq35un
object.transformEach refused reflectX, reflectY, random and copy given
as text, but read strokes, corners, objects and patterns given as text
as the key left out. With Scale Strokes & Effects on,
{"scaleH": 200, "scaleV": 200, "strokes": "false"} scaled the stroke,
recorded an undo step and kept the params for Transform Again, which
repeated the call. artboard.rearrange, artboard.move and
artboard.duplicate read their patterns option the same way, and
artboard.setProps reads its patterns, strokes and corners options the
same way.
These commands now refuse such a value with "`<key>` must be true or
false" before they change anything. So do the other options and
numbers of artboard.setProps and artboard.move, the art of Duplicate,
Copy and Cut Artboard, the shape commands and path.create, path.close,
path.setHandle, path.moveAnchors, the Pencil, Curvature, Reshape
Segment and Blob Brush options, text.create, text.createInPath,
text.setRangeStyle, text.editRange, text.getRange, text.touchType,
text.areaOptions, type.step, the font size steps,
type.opticalMarginAlignment, type.systemLayout, type.fractionalWidths,
type.ruby, the symbol and swatch commands, the Appearance panel's
commands, the gradient and freeform gradient commands, Shape Builder,
Live Paint, Image Trace, the opacity of object.setProps, the radius,
sides and relativeRadii of object.setLiveShape, the reference point of
object.setBounds and the artboard of Align, Distribute and Distribute
Spacing. They refuse a number given as text with "`<key>` must be a
number, not <value>", and a gradient stop's tint with "stop <n>:
`tint` must be a number". A null still reads as the key left out.
Docs: docs/mcp.md, docs/target-app-parity.md.
* Use chrono for UTC/calendar conversion instead of hand-rolled copies
Replaces the hand-rolled Howard Hinnant civil/days-from-civil copies and the
leap-year/days-in-month tables with chrono as a workspace dependency:
- apps/vectorcraft/src/logging.rs: timestamp (log line time)
- crates/doc/src/metadata.rs: civil (Unix seconds -> [y,m,d,h,mi,s]) and iso8601
- crates/pdf/src/file_info.rs: unix (date -> seconds) and in_month, used by the
PDF/XMP date parsers (pdf_date, xmp_date)
Output formats and parsing semantics are unchanged and pinned by the existing
tests (PDF D:... and XMP dates, offsets, leap-year rejection, year clamp 0-9999,
pre-epoch clamp). chrono uses only std, so wasm stays safe. AGENTS.md/CLAUDE.md
now say time goes through chrono.
Follow-up to storytold/cadcraft#1785
* Never panic on a clock outside chrono's range; keep 5-digit years unprefixed
chrono's From<SystemTime> unwraps, so a clock past year 262143 panicked
inside the logger. timestamp now uses the fallible DateTime::from_timestamp
and reads such a clock (or a pre-epoch one) as the epoch instead. Years past
9999 are written with {:04} like before, because %Y would add a '+'. Tests
cover the out-of-range and 5-digit-year cases.
* Don't build an unrepresentable SystemTime in the far-future test
UNIX_EPOCH + Duration::from_secs(u64::MAX / 2) overflows Windows'
SystemTime and panicked before timestamp() ran, failing the Windows
test job. Build it with checked_add and only assert where it exists.
* time: preserve old out-of-range date conversion (no epoch fallback)
`DateTime::from_timestamp(...).unwrap_or_default()` silently reset any instant
past chrono's year range to 1970-01-01. Restore the old conversion domain with
a cycle-based helper that routes the calendar through chrono (the calendar
repeats every 400 years = 146_097 days, so month/day/time come from chrono and
only the year is carried in i128, keeping every i64 exact).
- apps/vectorcraft/src/logging.rs: `timestamp` via `utc_parts`; keep the
pre-1970 epoch clamp; u64 secs past i64::MAX saturate, not epoch.
- crates/doc/src/metadata.rs: `civil` cycle-based, keeping its clamp EXACTLY
as the old code had it: the input is clamped to 0000-01-01..9999-12-31 first
and the year is clamped to 0..=9999 afterwards. (With the input clamp the old
`unwrap_or_default()` was already unreachable; this drops it and makes the
conversion explicit, with identical output.)
Pinned old outputs in tests:
logging u64::MAX/2 s -> 292277026596-12-04T15:30:07.000Z (was epoch),
10_000_000_000_000 s -> 318857-05-20T17:46:40.000Z
metadata civil(i64::MIN)[0]=0, civil(i64::MAX)[0]=9999;
iso8601(i64::MIN)=0000-01-01T00:00:00Z,
iso8601(i64::MAX)=9999-12-31T23:59:59Z (year clamp, as before)
* Don't build a year-318857 SystemTime unchecked in the logging test
Windows' SystemTime only reaches about year 30828, so
UNIX_EPOCH + Duration::from_secs(10_000_000_000_000) panicked before
timestamp() ran and failed the Windows test job. Build it with
checked_add and assert where the platform can hold it. The year-10000
case fits everywhere and stays unconditional.
* Name the 400-year range extension as the one exception to the time rule
The rule said never to hand-roll calendar math, but the range fix
(shifting by whole 400-year Gregorian cycles so dates past chrono's year
range keep their old output) is cycle arithmetic, so the rule and the code
disagreed. The rule now names that exception: the library still computes
month, day and time, and only whole cycles are carried in the year.
* Say the library computes month and day; time-of-day is plain arithmetic
Narrows the range-extension wording: the date library computes the
calendar fields, and splitting seconds into hours, minutes and seconds
is ordinary arithmetic, which is what the code does.
## Summary
Live ellipses already support pie start/end angles, canvas widgets and Invert Pie
like Illustrator. Label the Properties panel Transform section with Ellipse
Properties so the pie fields match the Transform panel and are easier to find.
## Tests
- `an_ellipse_shows_its_pie_properties` in `panels/transform` tests.
Closesstorytold/vectorcraft#3451
Co-authored-by: nrq6993-a11y <nrq6993-a11y@users.noreply.github.com>
An image whose data is empty failed to decode before the size check, so a
version 7 .ai said it couldn't be decoded. The writer now checks for empty
data first and gives "an image that is empty or too large for legacy .ai
was left out". docs/mcp.md now says the DXF decode note applies to R13 and
later; R12 leaves out every image with its own note.
EPS, EMF, WMF and DXF exports gave each image they left out because it
couldn't be decoded its own warning, "image '<key>' could not be
decoded and was left out", which names the internal key the document
stores the image under. A version 7 .ai said "an image that can't be
decoded or is too large was left out".
The four writers now give the note raster exports, PDF and the legacy
.ai flattening give, "an image couldn't be decoded and is missing from
the picture", once per export. In a version 7 .ai it is one line with
the note for flattened art. The other image notes in the same
functions no longer name the key: EPS's note for an image too large to
write, and DXF's notes for an image with no pixels (its rows in 23
catalogs are reworded) and one that could not be written. Legacy .ai
gives a separate note for an image that is empty or too large.
Docs: docs/mcp.md (EPS, DXF export, EMF and WMF, aiVersion "7").
variables_panel_data_sets_ask_off_the_ui_thread empties its folder
before Export Data Sets… writes out.csv, then checks that the file is
exactly the header and the row for data set B. It ignored the result
of remove_dir_all, and create_dir_all succeeds on a folder that is
still there. When the folder could not be emptied, an out.csv left by
an earlier run passed the check even when the export wrote nothing.
The test now fails when emptying the folder fails with an error other
than NotFound.
When the Compression settings keep a PNG as it is, the PNG goes to
the PDF writer unchanged, and the writer refuses images of 16 bits a
sample below PDF 1.5. Saving a document with a placed 16-bit PNG as
PDF 1.4, or as a .ai file at 1.4, failed with "PDF writer error:
SixteenBitImage(...)" and wrote nothing. PDF 1.3 and PDF/X-3 failed
the same way for an opaque 16-bit PNG, and PDF/X-1a failed for a
16-bit gray PNG.
In PDF 1.3 and 1.4 files, such a PNG is now decoded and encoded again
at 8 bits a sample, with the same color type and ICC profile, before
the writer reads it. From PDF 1.5 on, it keeps its 16 bits. A 16-bit
PNG whose pixel data can't be decoded is left out with the note "an
image couldn't be decoded and is missing from the picture".
Docs: docs/mcp.md (PDF options).
## Summary
object.align now moves selected ruler guides when aligning to the artboard,
matching Illustrator and the existing Distribute behaviour. Guide-only
selection can run align; the Align panel enables its buttons when only
guides are selected. Guides still stay put when objects are in the selection
and Align To is Selection; key-guide behaviour is unchanged.
## Tests
- engine tests_keyguide: guides_selected_alone_align_to_artboard; updated
selected_guides_align_to_a_clicked_key_object for artboard align
- ui-egui align panel: the_align_buttons_align_guides_selected_alone_to_artboard
Closesstorytold/vectorcraft#3463
Co-authored-by: nrq6993-a11y <nrq6993-a11y@users.noreply.github.com>
On Linux and the BSDs, where the save dialog reports the file type
picked in it, a name that still ends in the suggested name's extension
gets the picked type's extension. The comparison was case-sensitive,
so Untitled-1.VECTORCRAFT with "PDF-compatible .ai" picked was saved
as a native file named Untitled-1.VECTORCRAFT instead of Untitled-1.ai.
with_picked_type now compares the lower-case extensions that
fileio::extension gives. A name typed with the picked type's own
extension in another case (a.AI with .ai picked) stays as typed.
This follows up #3372.
Docs: docs/target-app-parity.md.
tests_ruby::vertical_ruby_is_upright_to_the_right_of_the_column
checked that no reading glyph is more than 1 pt wider than tall. With
craft-fonts in the build, the kana come from Shippori Mincho, whose か
is 8.24 pt wide and 7.22 pt tall at 10 pt. The test failed on that
glyph although it stands upright. FreeBSD CI, which builds with
craft-fonts at a71c20e3, fails it on every branch since 14f804cd. It
also fails on Windows with that ref or an older craft-fonts checkout.
The test now compares each reading glyph's ink with its own outline at
10 pt. The same width and height mean the glyph is upright; a swapped
width and height mean it lies on its side. It also requires one
reading glyph whose outline is not square.
find_saved matched names case-insensitively but save_as, remember and
switch compared them exactly: "Mine" saved over "mine" made a second
workspace, and changes kept as "Mine" didn't come back when switching
to "mine". save_as now replaces a user workspace of the name in any
case, remembered layouts are kept under the saved spelling and looked
up whatever the case, and sanitized maps remembered layouts to the
saved spelling and keeps one per workspace.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01KcHisW1vt6pkVPYbzL6Cpg
remember() pushed the layout being left and then truncated the list to
64, so with 64 workspaces already remembered it dropped the entry it
had just pushed and the changes were lost. Drop the oldest entries
instead; a test fills the list and checks the last one left survives.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01KcHisW1vt6pkVPYbzL6Cpg
A workspace remembers the changes made in it: switching away keeps them
(UiState::workspace_changes, kept with the preferences), switching back
restores them, and Reset <name> returns to the saved layout. Saving a
workspace starts it with no changes; renaming keeps them, deleting drops
them, and deleting the current workspace shows Essentials as it was
left. Saved preferences recover a damaged remembered layout instead of
losing the whole file.
Split from #2965 as a reviewer asked.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01KcHisW1vt6pkVPYbzL6Cpg
A face hidden on its own (for example with Object > Hide after Live
Paint Selection picks it) while its Live Paint group stays visible was
still highlighted, selected by Live Paint Selection and filled by the
Live Paint Bucket, and it kept the face of a group behind it from being
picked.
live_paint_at and lp_part_at now take the front-most face that is not
hidden (visible_face_at). A point on a hidden face falls through to the
visible face of a group behind it; where there is none, an edge pick
takes the nearest visible edge as it does outside the faces.
livePaint.fill also skips hidden faces. With group, a point on a hidden
face is an error. With ids, it fills the face there of another Live
Paint group among the objects, or else fails and does not merge the
group.
Docs: docs/mcp.md.
## Summary
Aligns parity checklists with shipped View › Snap to Glyph behaviour (#3124):
when Smart Guides are on, drawing tools snap to live-text baselines and glyph
bounds via `view.snapToGlyph`. Adds a headless test that `view.snapToGlyph`
toggles the UI view flag like Illustrator’s menu item.
## Tests
- `vectorcraft-ui-egui`: `snap_to_glyph_menu_command_toggles_the_view_flag`
- Existing `tools::text_tests::snap_to_glyph_lines_up_new_point_type_with_a_baseline`
Part of gap G4
Co-authored-by: nrq6993-a11y <nrq6993-a11y@users.noreply.github.com>
The docstring check of the review counts the functions a pull request
touches: effect.apply, effect.setParams, plugin.run, the line cap
name, the reader of an output object and seven of the object tests had
no doc comment. Each now says what it does or checks; the join and cap
names get one each.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HhvnnnD5fTFupK8WWq35un
A filter could reshape and repaint objects, but not set the stroke's
join, cap or miter limit, nor the effects an object's appearance holds.
A tool that prepares objects for a live effect (a sheet of samples,
each with the effect at its own settings) had to leave the user to
apply the effect to each object by hand.
- Each input stroke has its join, cap and miterLimit beside its paint
and width (live effects get the same, through the shared encoder),
and each object its own effect records, {id, params, visible}.
- An output stroke may set them (miter limit clamped to 1-500), and an
output object's "effects" replaces its own effects.
- The host admits each record as effect.apply does: an effect that
effect.list lists or an installed live-effect plug-in, parameters it
takes (filled in to their defaults; a plug-in's checked against its
manifest), one 3D effect that fits the object (its fills' and
strokes' own effects counted), and the texture of an effect with
Load Texture (Glass, Texturizer) as the effect stores it, never a
file to read. A record the object already had stays as
it is, so a filter that hands its input back changes nothing. One
refused record refuses the whole output and the document stays as
it was.
The plug-ins crate checks the records' form; which effects exist is the
engine's to say (objects::AdmitEffects, engine plugin::admit_effects).
effects::validate_three_d is the one check of whether a 3D effect fits
an object, for effect.apply, effect.setParams and the filters.
A path handed back as it was now stays the live shape it was: write()
dropped a live rectangle's or ellipse's parameters whenever the output
held a path, changed or not, so a filter that returned its input
turned a live rectangle into a plain path.
objects/tests.rs checks the round trip, stroke styles on existing and
new strokes, the host's refusal, the bad forms refused, and a live
rectangle handed back unchanged (a plain path before the fix);
tests_plugins.rs runs filters that set a built-in and a plug-in effect,
an echo that changes nothing, and six refused outputs.
CRAFT_PLUGIN_FRAME_WASM=<folder> saves a test filter to try in the app.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HhvnnnD5fTFupK8WWq35un
The renderer's pixels and egui's are both premultiplied RGBA8, so color_image copies the bytes in one go. At 3456×2234 that takes 0.9 ms instead of 3.4 ms per update, at 5120×2880 1.6 ms instead of 6.8 ms. Every rendered image the UI shows (canvas, thumbnails, previews) goes through it.
## Summary
Opening or placing a .pmd file now fails with a dedicated unsupported-format
hint (like DWG and PICT), telling users to export PDF or EPS from PageMaker
instead of a generic unknown-format error. document.formats lists PageMaker
in its unsupported table.
## Tests
- formats_list_dxf_and_what_can_not_be_written
- dwg_and_pict_say_what_to_use_instead (includes brochure.PMD open)
Part of storytold/vectorcraft#3343
Co-authored-by: nrq6993-a11y <nrq6993-a11y@users.noreply.github.com>
complete_languages_translate_every_message read a message literal only
right after a marker such as Err( or Other(. It did not read the
message given to the engine's bad(cmd, ...), so this test did not
require a row for a parameter error. It also did not read the messages
given to the engine's err() and error(), to the PDF writer's bad() and
failed(), or to an error enum's variant such as PsError::Limit(...).
1,122 of these messages have no row in any complete catalog.
The scan now reads those messages, a literal or a format! string, in
the same template form as the catalogs' rows. It skips PsError::Ps and
EngineError::UnknownCommand, whose first argument is a PostScript error
name or a command id. The 1,122 messages are listed in
untranslated_messages.txt next to the tests, together with the 71 that
pull requests open at 009384b add without rows, and the test lets a
listed message through whether or not it has rows. Every other message
needs a row in each complete catalog, so a new message made this way
fails the test until it has its rows.
VECTORCRAFT_I18N_DUMP_UNTRANSLATED=<file> writes the messages in the
sources that lack a row. No catalog rows change.
Docs: docs/development.md, docs/localization-parity.md.
Its glyphs were put in visual order and then cut where the path ran out, which kept the text's end. Now the longest start of the text, in reading order, whose glyphs fit in visual order is laid out, as area type does with each line it breaks, by the same rules that place the line: steps where the glyphs sit, hanging punctuation and the path beyond the brackets.
Glyphs placed one by one with even gaps could be letters set apart or
words of one letter each, and a file's positions can't tell them apart.
The spacing a line's letters share is now taken off its gaps only when
the line also shows where its words break: with spaces it draws (as
with Tc), or with gaps wider than the rest (tracked words placed glyph
by glyph). Otherwise every gap past a fifth of an em is a space, as
before, so A, B, C and D placed with ordinary word spaces between them
still come in as "A B C D".
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A line whose shared gap is over half an em is no longer read as letter
spaced: capping that gap at half an em left glyphs set 0.55 em apart
with no spaces. The regression now covers 0.55, 0.75 and 1 em.
Whether a line is right to left is now known before its gaps are read,
from its logical order: the paragraph direction is only set afterwards,
so right-to-left lines were read for letter spacing too.
The revision-history row moves out of the Headline table, where it had
been put by mistake.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
# Conflicts:
# docs/target-app-parity.md
