A freeform gradient's image was sampled, encoded and written inline at
every fill, outside the reuse placed images have. Fields now go through
the same codes and placements: a fill whose encoded image (color setup,
image dictionary, filter, operator and samples) matches one already
kept draws from it, so its samples are written once in the setup as
VCim<i>. A field drawn once is still written where it is drawn.
A fill whose placement can't be drawn, or that arrives when the page is
already at the 256 MB limit, is no longer sampled first, and a field
image taken back with art that passed the limit is forgotten.
PDF, print, EPS and DXF wrote a symbol instance by placing its symbol's
art as stored, so live effects inside that art were missing: a square
moved 100 pt by a Transform effect stayed in place in the file while the
canvas drew it moved.
bake_document_placing_symbols replaces each instance of a symbol whose
art (or the art of a symbol nested in it) needs baking by that art
placed as the instance places it, then baked. The art is placed before
its effects apply, as on the canvas, so an instance that scales its art
does not scale what its effects do. A dynamic instance's own paint
changes apply before baking. Nesting stops at 32 levels and the placed
art at 16,000,000 objects per export; past them an instance stays an
instance. DXF places 9-slice cells unclipped, as its scene does. The
symbols' own art, SVG and the .ai editing data are unchanged, and HPGL
already baked each placed instance.
Type whose color bitmap glyphs fit under the page limit, but whose run
fill passes it, is left out together with the image placements of its
glyphs. The new test builds the color font in memory and checks that
the page then equals the page without that type: no reference or shared
resource is left for the glyph image, a glyph drawn once elsewhere is
written inline again, and an image drawn twice keeps its shared samples.
Opening an .ai file now reads Insert Aki Left and Right from its text
document through the same cascade as the other character keys: the run,
its character style and the styles that one is based on, [Normal
Character Style], then the document's defaults. The values are read in
thousandths of an em. A negative value or one that is not a number opens
as 0, and a value past 1000 opens as 1000. How the text document stores
these values is VectorCraft's own rule.
An .ai save writes each run's nonzero values in the form that reader
reads, and no longer reports Insert Aki as left out of the editing data.
A legacy .ai text object has no Insert Aki: type with it that is written
as text now reports that the setting is left out. Type written as
outlines keeps the space.
A clipping path inside art cut to another clipping object is written as
its fills and its strokes. Its text wrap options went with the fills, so
when the outer region left nothing of the fills, the strokes were written
without them. The strokes now keep them unless a fill object was written.
A symbol definition that had no room left for its group under the object
limit returned without restoring the writer's drawn-art depth. Asked for
from art drawn for an object (a stained instance's art), the depth then
dropped below zero on the way out, which panics in debug builds. That
path now restores it as the others do.
Two faults in the editing data writer's text wrap handling:
- A clipping path inside art cut to an outer clipping object is written
as its fill and its stroke. The stroke's text wrap options were
dropped whenever the path had a fill, so when the outer object cut
the fill away entirely, the stroke was written without them and the
options were lost. The stroke now drops them only after the fill
writes an object.
- When a symbol definition had no room left under the object limit,
`define_symbol` returned without restoring the drawn-art counter it
had cleared. A live instance inside drawn art (a stained instance's
art) then left the counter at zero, and the art around it subtracted
from zero: a panic in a debug build. That path now restores the
counter as the other exits do.
In a composite font the file carries (FMapType 2 or a CMap), cshow ran
its procedure once for each byte of the string, with that byte's width
in the composite font and the composite font still current. It now runs
once for each character the font maps, with the last byte of its code
and its width, and with the descendant that draws it current, at the
size the composite font sets it, while rootfont gives the composite
font (PLRM 3rd ed., cshow and rootfont).
While the procedure runs, showing the byte it was given in that
descendant shows the character the code maps to: a CMap's CID,
base-font code or glyph name. In a CIDFont any other string is a
rangecheck error.
A top object whose top and bottom sides, or left and right sides, have
different numbers of segments was written to the editing data as its
distorted art. The writer now splits the side with fewer segments into
as many as the side across from it has, each segment into equal spans of
its curve, up to the 50 columns or rows the grid holds. The reader opens
the grid's edge as the top object's outline with the same corners and the
same mapping of the contents; the outline holds the added anchors.
A split that would share out the contents another way, or a side of more
than 50 segments, still writes the envelope as its art with the warning.
A name ending in the row-and-cell CMap H or V split only when the file
defined the CIDFont or a face of that name was installed, so
/Ryumin-Light-H findfont in a file that only names it read its strings
byte by byte as a plain font: <467c> came in as "F|" instead of 日.
Ryumin-Light and GothicBBB-Medium, the Japanese CIDFonts a printer
carries (Adobe Technical Note #5014, appendix A.2), now split at H or V
as well. Latin names such as Foo-Bold-H still stay whole.
The KSC-Johab CMaps had no code space listed, so type in a font over
KSC-Johab-H or KSC-Johab-V read every code as two bytes, so two ASCII
characters moved the point one em instead of two. The code
space now follows Adobe's KSC-Johab-H CMap: <00>-<80>, <8441>-<D3FE>,
<D831>-<DEFE> and <E031>-<F9FE>
(https://github.com/adobe-type-tools/cmap-resources/blob/master/Adobe-Korea1-2/CMap/KSC-Johab-H).
Paragraphs in a modern .ai text document now keep their East Asian line
breaking settings: the Kinsoku set (None, Hard or Soft), Kinsoku Shori
Type, Burasagari and Bunri-Kinshi. Each comes from the paragraph, else
its paragraph style or a style that one is based on, else the document's
defaults, as the other paragraph settings do. A value that cannot be read
takes the paragraph style's value.
A paragraph whose Kinsoku set is a custom set takes its paragraph
style's set, and the import warns that custom Kinsoku sets were left out.
.ai export does not write these settings.
A path's guide lengths matched the brush's within 0.001, so on very short art guides at 30% and 35% of the art matched. The guides now also match as fractions of the art's length.
Opening an .ai file compares at most 256 envelopes and embedded images
with a Warp with their drawn art, from one budget. Saving the editing
data counted only envelopes, so a file with more than 256 of them in all
was saved with images past the budget written live and no warning, and
it opened with those images as their drawn look.
- The writer's budget now counts every image written with its Warp as
well as every envelope, in the order they are written. An image past
it is saved as its drawn look with a warning, and an envelope past it
as its distorted art, as before.
- An image whose pixels can't be read is written as its gray outline,
which the reader does not check, so it takes no part of the budget
and keeps the warning about its pixels.
- The envelope warning names the shared budget.
An envelope whose contents hold type was written to the editing data as
its distorted art, so the type opened as outlines. The type is now
written as type in the envelope's contents, in the form the .ai reader
opens as a live envelope of type (TypedEnvelope in
crates/eps/src/import/ai/envelope.rs), with Edit Contents off or on.
An envelope whose type is threaded, has text wrap options or is written
as its outlines (mirrored type, type in a symbol's art) is still written
as its distorted art, with the warning. A record taken back for the
object limit or for outlined type also takes its stories out of the text
document. The writer charges one more object for each text object in the
dictionary's contents, where the reader reads a copy of it.
A path's Art brush settings matched the brush's within a small tolerance, and only the brush's were checked. A path with a Spacing of 0.0005, or guides that leave no art between them, opened live with the brush's settings. The path's settings now pass the same checks as the brush's, and such a path keeps its drawn look.
A copy of an image written whole already was cut to the painted bounds of
its opacity mask even when the mask was then left out, nested too deep or
past the object limit, so the file read back with a cropped image and no
mask. The copy is now cut to its mask only when the mask is written, and
otherwise written whole.
The progress-doc line now gives the measured total for the warped test
image's pieces: about 7 times the image's samples, not twice.
A linked image whose appearance is one Warp effect, with no opacity mask,
was saved to the .ai editing data as its drawn art, so its live Warp was
lost. It is now written with an art style of that Warp, as an embedded
image is: its bent pieces are its drawn look, and the image itself
follows on %_ lines as a placed file with its path, its folder relative
to the document's and its preview. In an opacity mask's art the copy is
embedded, with the existing warning for linked images in masks.
Opening such a file reads the image back linked, with its Warp live.
A hidden image with a Warp, embedded or linked, is now compared with its
drawn look shown, so it keeps its Warp too; before, the look drew
nothing and the image came in as its drawn pieces.
A .ai path with an Art brush opened with the brush on its stroke only
when the brush stretched its art at 100% in Tints, unflipped. Paths with
an Art brush in None, at another fixed width, with Stretch Between
Guides, or flipped along or across now open as the path with the brush
on its stroke too, and the brush in the Brushes panel has those options.
The guides come in as fractions of the art, from the lengths of the art
outside them. A brush with Scale Proportionately comes into the panel
with that option, and its paths keep their drawn look.
Paths keep their drawn look when their settings that change the drawing
differ from the brush's, and when the brush is in Tints and Shades or
Hue Shift, has a width that follows the pen, has guides that leave no
art between them, or has a Spacing, Adjustment or FoldAdjustment other
than 0, 1 and 1.
The .ai save writes the same options, so such brushes and the paths
that use them come back as they were; a path with a proportional brush
is written as its brush's art.
A path's own lines can be followed by samples of the pen
(`(…) Xt n XD`). The reader took them as the end of the object, so the
style marker after them was not matched: the path came in as its drawn
look plus a second, unpainted copy of the path, and no note said the
look was kept. The samples now stay with the path. They aren't read,
so a path that has them keeps its drawn look alone, with the note, when
a setting of its Calligraphic brush varies with the pen (Pressure,
Tilt, Bearing or Rotation, with a variation above 0). With any other
style, an Art brush or a Calligraphic brush that doesn't follow the
pen included, it takes the style as it would without them.
A Calligraphic brush's filter with an entry the reader doesn't know, or
with a variable width, keeps the drawn look too, as an Art brush's
filter does.
Docs: docs/file-format-parity.md, docs/gaps.md.
A straight `(N) *` guide path that runs across the canvas or an artboard
opens as a ruler guide, but a hidden one (`1 Xw`) stayed a guide path.
It now opens as a ruler guide too, and every ruler guide made from such a
path keeps the path's hide, lock and name, so its Layers row shows them.
Before, a locked guide opened unlocked and a named one lost its name.
Hidden or locked guide paths that are short, slanted, grouped or `(S) *`
stay guide paths, and commented copies (`%_`) are still left out.
Docs: docs/target-app-parity.md, docs/file-format-parity.md.
A symbol instance colored with the Symbol Stainer lost its stain in
export whenever the canvas's stained art was not used:
- An instance with geometry effects was reshaped for export from the
symbol's art with only its dynamic changes, so SVG, PDF, EPS, .ai,
DXF, EMF/WMF and HPGL wrote the unstained colors. A color adjustment
on an instance recolored the unstained art the same way.
- PDF, EPS, legacy .ai, DXF and EMF/WMF wrote an instance without
effects from the unstained art.
tint_node, stain and instance_art now live in
vectorcraft_doc::symbol_overrides, which both the effects and the
exporters can use; vectorcraft_brush re-exports them. Baking, color
adjustments and the writers above now use instance_art: dynamic changes
first, then the stain.
A straight compound path with Twist was read live up to 180° whatever
guides it held, and the live Twist was measured from bounds that take in
the guide members. A straight compound path with Twist whose guide
members reach past its other members now opens as its drawn look
instead of live, and so does a compound path made only of guides.
Compound paths whose guides lie inside their other members, a Twist of
0°, and single paths keep opening live.
The `.ai` writer keeps a graphic-style link only where the reader reads
the style live, so such an object is now written without the link, with
the existing style-link warning.
A raster mesh warp with Preserve Shape Using Transparency cuts an image
into pieces with opacity masks and puts them in a knockout group, each
piece knocking out its triangle. The reader gives the pieces no such
mark, so each knocked out all that its image paints without its mask,
and the file opened with most of the image hidden.
The group of such pieces is now written without knockout. An image
whose pixels are partly transparent draws its alpha twice only where
neighboring pieces overlap.
A PDF-compatible .ai saved with Include Editing Data for Other Apps
wrote no text wrap options, so a wrap object opened from the file no
longer wrapped type. An object with only wrap options got no
dictionary at all.
The writer now writes each wrap object's options in its dictionary as
`AI11TextWrapOffset` (`/Real`) and `AI11TextWrapInvert` (`/Bool`, 1 or
0), the entries `text_wrap` in crates/eps/src/import/ai/read/document.rs
reads, beside the object's name, variables and opacity mask.
- The offset is the one in the art's space, so on a large canvas it is
written at 1/10 and read back at its size. An offset that is not
finite is written as the default 6 pt, and one past the -1000 to
1000 pt the reader keeps as the nearer end of that range.
- Art the writer draws for an object (a blend's steps, a symbol
instance written as its art, a brushed stroke's pieces, pattern
tiles, an object's drawn look) is written without the options of the
objects it is made from. The object itself and a symbol's own art
keep theirs.
- A clipping path written as its fills and its strokes has its options
once.
Docs: docs/file-format-parity.md, docs/target-app-parity.md.
A legacy .ai graph (%_Gs ... %_GS) opened as plain art, with its table,
type, axes and settings left out. A grouped column, stacked column, line
or area graph now keeps them on its group when every record has the
shape and values the version 7 specification's appendix A gives, the
table holds at least one number, the drawn art strokes the plot's left
and bottom edges, and each series group is tagged once as a whole
column, holds only paths and paints them with one fill and one stroke.
The saved art stays as it is on open; Graph Data redraws the graph from
its data afterward, with each series' fill and stroke kept.
Any other graph (pie and scatter graphs, designs, customizations, record
shapes of later versions, tables past 10,000 categories, 256 series or
2,600,000 cells) keeps its art as plain objects with the existing note.
Saving a graph to .ai, in either form, now warns that its data and
settings are not written.