The walk that writes link annotations measured every object at every
level, on every export that can have links, and read the box only for
an object with a URL that opens and for a clipping group. It now
measures only those, so a document with no URLs or clipping groups is
not measured a second time. The link annotations are unchanged.
The output is byte for byte the same on a document of linked objects in
plain and clipping groups on and off the page, with default settings,
with uncompressed text and with PDF layers.
The walk measured every object's visual box to leave out what is off
the page. A plain group or layer is measured as the union of its
members, and each member then measured itself again when it was
written, so an object in plain groups was measured once for each level
it sits at.
Measuring a plain group or layer now keeps its members' boxes in a
tree, and each member uses its box when it is written. A container
written whatever its box is (a PDF layer's group, placed symbol art) is
not measured. A knockout group's elements are measured for the group's
box and again for the list of elements on the page, and that second box
is used for the element and for its mask. Objects in a live container
(Blend, Envelope, Repeat, Intertwine) are still measured with the
container and again after it is expanded. Node::visual_of_members holds
the formula that Node::visual also uses, so the boxes and every decision
are unchanged.
The output is byte for byte the same on one open path of 200,000
anchors, 20,000 rectangles in 200 groups of 100, and 5,000 point type
objects, and on a document of plain, see-through, clipping and knockout
groups on and off the page, each exported with default settings, with
uncompressed text and with PDF layers.
When the scan for swatch names searched data outside a section and found no `~>`, it stopped searching for the end of all later data. Data in a later section then was not skipped, so a `(` on a `%_` line of it opened a string that ran past the section's end marker, and a color named after the section did not reserve its name. The scan now records the search that found nothing per section's end marker, so a search in another section still stops at that section's markers.
Data with no `~>` before the section's end marker was lexed as tokens, so a `(` on a `%_` line of it opened a string that ran past the marker and the section did not end. The scan now skips that data up to the section's end marker, or up to the start of a section of the same kind nested after it. A `%` on the key's own line starts no marker, and a marker line with bytes that are not UTF-8 counts, as the scan reads comments.
When versioned content does not end, the scan for swatch names goes back to where its dictionaries stood at the start of that content, so data in the range it reads again is skipped.
A section scan lexed a /ForeignObject's /Data, a /Binary dictionary's data and a text document's data as tokens. A `%_` line of that data holding a `(` opened a string, so the section ran past its end marker: the editing data failed to read, or the art after it was misread. Reader::section now skips the data up to its `~>`, as the reader does for art outside a section. Data without its `~>` before the section's end marker is not skipped. The scan for swatch names does the same.
The `Tool` implementation's `busy`, `pointer`, `key`, `claims_key`, `overlays`
and `cursor` in shape.rs, and `pointer` and `cursor` in select.rs, are the ones
this pull request edits; each now has a doc comment, so the docstring coverage
of the functions the diff touches is complete.
- The 45° move assertion compares dx and dy with a small tolerance instead of
exact equality (the matrix's floats are rounded), keeping the non-zero check
and the message.
- The engine test helpers say what they do, so the docstring coverage of the
functions this pull request touches is complete.
The roadmap's status line and revision history now record the shape tools'
centre-point move and corner rotate (#5477), and the parity status line no
longer names the Transform tab-order fix, which is its own pull request.
The Selection tool's cursor now checks the ruler guides before a selected
object's centre point, as a press does. Tests: a ruler guide under the centre
point keeps the pointer and drags; on a box too thin for its centre to clear the
handles the handles take both; a shape-tool move from the centre snaps and shows
the hollow arrow; and Alt pressed mid-drag copies the shape through the engine.
The Selection tool's cursor now checks the bounding-box handles before a
selected object's centre point, as a press does (on a thin box the centre lies
within a handle's reach). The shape tools' tests cover a Live Corners widget
that lands on another selected shape's centre (the widget rounds), Alt pressed
mid-drag (the preview copies and the copy cursor shows), and the engine tests
check the copy's place and both gestures' Undo and Redo. The cursor catalog says
which arrow a drag of the selection shows.
- The Selection tool takes a Shift-press on a selected object's centre point
too: a Shift-drag moves the selection held to 45° steps, and a Shift-click
there takes the object out of the selection, as a Shift-press on the object
does.
- One rotate zone for the Selection tool and the shape tools
(`select::rotate_zone`): a press on a bounding-box handle is not a turn, so
the shape tools' zone matches the Selection tool's classification exactly.
- The shape tools' hover cursor follows the press order (corner and pie widgets,
then the centre point, then the rotate zone), and a move of the selection
shows the Selection tool's feedback: SelectionCopy with Alt, ArrowHollow while
it snaps.
- Engine-level tests for the whole gesture (`crates/engine/src/tests/shapetools.rs`):
repeated previews and one undo step, Alt copy, Escape then release, a turn
about the centre, the corner widget still rounding, and an unfilled shape
moved by the Selection tool.
- docs: evidence in target-app-parity.md's Shapes bullet and the M2 milestone,
and the Move cursor's entry names the copy and snap cursors.
A press on the centre point of the selected shape now moves the selection, as
asked (#5477), so the Constrain Angle test's repeated draws from the same point
moved the polygon drawn before them. It deselects first: nothing selected, the
press draws.
The Artboard tool's own Control bar drew the W/H Constrain Proportions link
between W and H, so with that tool Tab stopped on it before H, as the object
Control bar did before the last commit. It reserves the link's place and makes
the button after the fields too, and the link now toggles with Space or Enter
when it has the keyboard, as a dialog's button does.
Tests: `one_tab_moves_from_a_field_to_the_next` covers the Artboard Control bar
(W to H in one press), and `the_control_bar_link_toggles_by_click` Tabs from H
to the link and toggles it with Enter.
A shape just drawn keeps its bounding box with the Rectangle, Rounded Rectangle,
Ellipse, Polygon and Star tools, but its centre point (Attributes › Show Center)
did nothing and the box's handles only drew: a press anywhere started a new
shape, and with the Selection tool a press on the centre of an unfilled shape
started a marquee.
- The Selection tool now takes a press on a selected object's centre point as
the start of a move (held by that point, so it snaps and Alt copies), as
Direct Selection has since #4891. A click there still makes the object the
key object of a multi-selection.
- The shape tools take it too, and take a press in the rotate zone just outside
a bounding-box corner (the Selection tool's own zone and cursor) as the start
of a rotation about the box's centre. Live Corners widgets keep the corners
themselves, so a press on a corner still rounds it; every other press still
draws a new shape. Escape cancels, as for the Selection tool's drags.
`hit_center` moves from direct.rs to select.rs, shared by the three tools.
A review found the link lost its keyboard focus when it stopped being a tab
stop. It is focusable again, and the Control bar now keeps the link's place
between W and H but makes the button after the H field, so Tab still goes from
W straight to H and finds the link after it (as the Transform and Properties
panels already draw it). Tests: Shift+Tab goes back one field too, and a click
on the Control bar's link still toggles the preference.
The Blob Brush Tool Options test that expected two Tabs between the Roundness
field and the tip preview now expects one, as every field does. Docs: the Tab
entry in target-app-parity.md, both revision histories and the M1 milestone.
A numeric field's label is a scrub zone laid over the label (a drag scrubs the
field, a click focuses it). It was registered with `Sense::click_and_drag()`,
which is focusable, so every field had a second, invisible tab stop: Tab went
from X to W's label zone and only then to W. The zone is now `CLICK | DRAG`,
not focusable, and the W/H Constrain Proportions link — a button between W and
H in the Control bar — is no longer a tab stop either, so Tab goes from W
straight to H. The preference stays reachable from Preferences › General.
Test: `one_tab_moves_from_a_field_to_the_next` (Transform panel, Properties
panel's Transform section and Control bar).
The Community links can fall below the first scroll viewport at 1000x700
when the fonts differ, so the test measured nothing. 1000x900 keeps every
link visible.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
The first commit centred the headings but left the New / Open buttons
and the community links on the left, as `horizontal` fills the width.
The buttons' row is now padded by its measured width, and the Home
screen's links are centred (label only, no icon; About keeps its
left-aligned icon rows). The test now measures where the centred text
lies and checks the buttons, the Discord button and the links.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0113uEh3VQ6et4zDMZ9cqzkf
Part of #4684 (unify the UI with PhotoCraft). The Home screen's welcome,
New / Open buttons and Community block are centred, as PhotoCraft's
Home screen is, and the large Discord button takes the theme's accent
instead of Discord's own colour. Preset cards, Recent Files and the
wording are unchanged.
Still open from #4684: shortcut hints on the Home buttons, PhotoCraft's
PRO and STUDIO themes, the Preferences layout and the title bar.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0113uEh3VQ6et4zDMZ9cqzkf
A new test binary opens a document of 13 type objects with the plans a
font face keeps and again while every shaped segment makes its own plan.
The document, its cached bounds and baselines included, and the first
frame's pixels must be the same. `test_fonts::keep_shape_plans` switches
the plans off; it exists only in tests and with the `test-fonts` feature.
The engine's open test now builds objects of one script and direction
that differ in kerning and language, and lays each one out in a font
database of its own, so a plan used for the wrong features fails its
comparison. It no longer sets Arabic, which could fall back to a system
font in one database and not the other.
Every shaped segment built its own harfrust shaping plan, although a
plan depends only on the face, direction, script, language and features.
Opening a document lays out each type object once, so a document of many
short objects built thousands of identical plans.
A font face now keeps up to 32 plans, oldest dropped first, and shapes
with the one that matches. Feature ranges still come from the shape
call, so text with the same features over other ranges shares a plan.
The glyphs, bounds and baselines do not change.
Docs: docs/target-app-parity.md.
The clipping path around a transparent image flattened for a version 7
.ai followed 2 pt cells, a visible staircase. It now follows the rendered
pixels (in blocks only past 4 million), grown by one pixel so the whole
anti-aliased edge is inside, through the middle of each pixel edge and
simplified with Douglas-Peucker, so slopes and curves are sloped lines.
BFlat