The horizon test compares a point's signed w with the largest absolute
value in the bottom row, and the invertibility test compares the
absolute value of the determinant with the largest absolute value among
its six products, so a negative determinant counts as invertible. A
stored projection that overflows is divided by the largest absolute
value in its bottom row.
The power of two that brings a small map's entries up before a point is mapped is now reduced by the power of two of the point's larger coordinate, so no product of an entry and a coordinate reaches 1. A map that mapped a point before the scaling step does so again, and affine_at uses the same entries.
apply and affine_at multiply every entry by the power of two that brings the
largest to at least 1/2 when all entries are below 1/2, so products that would
be subnormal keep their precision. The identity times 5e-324 mapped (0.5, 0.5)
to (0, 0) before. Maps with an entry of 1/2 or more are unchanged.
adjugate_inverse no longer applies the relative determinant test to its result. That test is not symmetric between a map and its inverse: a map whose determinant is 1e-8 of its largest product can have a finite, accurate inverse whose own ratio is near 1e-16, so inverse() was None while is_invertible() was true. The result is now kept when every entry is finite and its determinant, summed from the six products in split form, is not 0.
The inverse holds every cofactor and the determinant as a mantissa and a
power of two, and forms each entry from their quotient. A map such as
diag(1e-300, 1e300, 1e-300), which counts as invertible, now gets its
inverse diag(1e300, 1e-300, 1e300) instead of none.
When a cofactor of a map with a normal but very small determinant underflows to 0, the
inverse formed from the map itself can have a w row of 0, so every point mapped through it
is refused. The inverse is now kept only when it is itself invertible, and is otherwise
formed from the map divided by a power of two that brings its determinant's products near 1.
The scale-free determinant test no longer builds a rescaled copy of the map. It
forms each of the six determinant products as a mantissa product and an
exponent sum and compares their sum with the largest of them, so a map such as
diag(1e307, 1e-157, 1e-157), whose rescaled copy overflows, is still invertible
and loads from a document. The rescaled copy is built only to invert a map whose
determinant is not finite or is below the normal range.
A homography means the same map at any nonzero uniform scale of its
entries, but `apply`, `is_invertible` and `inverse` compared w and the
determinant with absolute thresholds. A map stored with small entries
refused every point and could not be inverted, and a map with large
entries drew points past its horizon.
- `apply` refuses a point when w is at most 1e-9 times the largest
magnitude in the w row.
- A map counts as invertible when its determinant is more than 1e-15
times the largest of the six products the determinant sums, taken on
a copy rescaled by a power of two. An inverse whose determinant
overflows or underflows is taken on that copy.
With Envelope Options > Distort Appearance on, a path whose Transform
effect has Scale Strokes & Effects or Transform Patterns on, or copies it
with a gradient, was baked with the effect as geometry only: every copy
kept the original's stroke weight, dashes and pattern placement before
the envelope bent it. bake_appearance now makes such a path into the
per-copy art that drawing, export and Expand Appearance already use, and
bakes each copy's strokes into filled outlines from there.
The menu bar takes ⌘V's key-down as Edit › Paste's key equivalent, and
`native_menu::key_events` turns it back into egui input. With text on the
clipboard that was a Paste event; with only a picture (a screenshot) it
was nothing, leaving the paste to the window's own key-up. egui's window
integration only releases a key whose press it saw, so that release never
came and ⌘V pasted nothing, while Edit › Paste and ⌘V with the menu open
worked.
`key_events` now ends ⌘V with the key's release, as it already did for
⌘X and ⌘C: without a Paste event that release is the picture paste
`PasteChord` looks for, and after one it ends that paste so the next
picture paste still fires.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The geometry effects after a fill's or stroke's own Transform effect with
Scale Strokes & Effects or Transform Patterns now take the bounds of the
transformed path and all its copies as their reference box, as an
object's own effects do. A later Rotate turns the copies about their
combined center instead of turning each copy about its own, and a later
Offset Path or plain Transform applies to the paths themselves. Each copy
keeps its own scaled distances.
Part of #885
transform_item_leaf now holds back the leaf's geometry for each plain or transformed piece that follows a transformed fill or stroke, so the copies of an earlier item no longer use up the budget that a later fill or stroke still needs. The nested pass over a transformed piece keeps to the same reduced limit.
Each fill or stroke without such a Transform effect is drawn on the object's geometry again, so its cost now counts toward the budget that limits the Transform effect's copies. The object's own geometry is no longer counted a second time next to the pieces that replace it. The Transform effect options entry in docs/target-app-parity.md links the current help pages.
Part of #885
A Transform effect in a fill's or stroke's own effect list now applies its
Scale Strokes & Effects and Transform Patterns options as the same effect on
the object does: the stroke's weight and dashes, the distances of the fill's
or stroke's later effects and its pattern follow the transformation, each
copy once more than the last, and with Transform Objects off the geometry
stays and the options apply once. The fill or stroke becomes an object of
its own in the art the canvas, export and Expand Appearance draw, and Use
Preview Bounds measures that art too (also for the object-level options).
Part of #885
The type objects in the selection are deduplicated with a set, each frame's
thread is looked up in a map built once, the locked artboards are measured
once per call, and the menu check stops at the first selected type object.
Type selected by id on a hidden, locked or template layer, or outside
the isolated group, no longer counts as selected type: the command stays
off and does not select that frame's thread. The check reads the
object's ancestry, so the Select menu does not walk the whole document.
Select › Same › Link Block Series no longer counts hidden or locked type
inside a selected group, or type inside a hidden or locked object there,
as selected. A group that holds only such type leaves the command off, and
the command does not select the other frames of that type's thread.
Select › Same › Link Block Series (select.same.linkBlockSeries) selects every
frame of the threads that the selected type objects, and the type inside
selected groups, are in, in thread order. Type in no thread selects itself,
other selected objects are left out, and with frames of several threads
selected it selects all of them. Hidden, locked and out-of-isolation frames
are left out, Reselect repeats it, and it records no undo step. The item is
off while no type is selected.
Menu row in every catalog, docs/mcp.md and a docs/target-app-parity.md line.
File > Open and File > Place read BMP files named .dib or .rle (run-length
encoded bitmaps included), Photoshop files named .pdd and PNG files named
.pns, in any letter case. The Open and Place dialogs list them in "All
readable files", "All placeable files" and their format's own filter, and
Glass's Load Texture lists them too. Linked files with these names update
and relink, the Links panel shows their format, and Edit Original opens
them.
The extensions are read-only: exports still write .bmp, .psd and .png, and
an export to a path with one of the new extensions is refused as before.
The desktop packaging does not associate them with the app.
dotCooCoo