num() rounded ties to even only at 17 significant digits. IDML takes
the decimal nearest to the exact value at the shortest length, with
ties to even, at any length, and writes numbers below 1e-4 with an
exponent (1e-05). The rule gives the IDML text for all but 67 of the
1,090,937 distinct numeric texts of the reference IDMLs (those 67 are
text values).
diff_outputs.py: 2,672 outputs differ, no change in failures or
warnings.
compare.py --exclude own/:
- trustworthy 97.72% (+77 values): PDF GraphicBounds 849 -> 913 of
913, EPS 323 -> 329, SVG 140 -> 146
- all pairs +99 values: PDF, EPS and SVG GraphicBounds all equal
- extra values unchanged
Record kind 2 of chunk 0x3E05 is u8 2, u8 0, the u32 UID of another
font family and a u16 count of font entries. The converter read it as
a family record and left the fonts out with a warning ("string tag 5
at 3, expected 2" and similar), in 14 Japanese pairs and 5 other files.
IDML writes a family with a count of 0 as a copy of the family it
refers to, with FontType="Unknown", Version="" and TypekitID="$ID/"
(15 of 15 pairs). The converter now does the same. For a record with
entries (4 families in 2 files, no IDML) it keeps the name and leaves
the fonts out with a warning.
compare.py --exclude own/:
- trustworthy 97.72% (+234 values: 16 fonts, FontFamily Name 15,
AppliedFont of styles and text 11); all pairs 95.89% (+234)
- extra values unchanged (26,256; all pairs 304,533)
In InDesign 3.0 files chunk 0xCB02 starts with four zero bytes before
the flagged name (9 of 9 distinct 3.0 files, both byte orders). The
converter left the composite font out with a warning ("string tag 0
at 1, expected 2"); it now reads the name after the four bytes when it
does not parse at 0. The 3.0 files convert without that warning.
No pair is from 3.0: compare.py --exclude own/ is unchanged
(trustworthy 97.72%, all pairs 95.89%).
Decode the save history of the document (chunk 0x1D8): one record per
session, with a u16 language code. The layout parses 4,431 of the
4,439 distinct valid corpus files (not the six InDesign 2.x files and
two without the chunk) and all 803 pairs. Code 0x0101 in the last
record marks a session of a Japanese or Chinese edition: all 149
Japanese pairs have it.
The converter uses it for values that IDML writes in the exporting
InDesign's language:
- spread FlattenerPreference without chunk 0x10833: 400/400 for
0x0101, otherwise 300/150, none when the history cannot be read
(7,481 of 7,490 spreads of all pairs; the 9 others are one pair
exported in a session the INDD does not record)
- IndexOptions without chunk 0x13010: BetweenEntriesSeparator "、" or
"; " (789 of 791 pairs) and Title "索引" for 0x0101 (158 of 159);
no title for other codes, whose titles vary by language
- EndnoteOption without chunk 0x2261E: separator U+3000 and
RubyMarker for 0x0101, tab and SuperscriptMarker otherwise (140 of
141)
The endnote title and the Assignment Name do not follow the code and
stay out (objects.md, save history, has the counts).
compare.py --exclude own/:
- trustworthy coverage 97.71% -> 97.72% (+1,119 values: flattener
resolutions 5,065 -> 5,404 of 5,413 each, index title 10 -> 136,
separator 481 -> 604 of 606, endnote separator and marker position
301 -> 397 of 398)
- all pairs 95.88% -> 95.89% (+1,679)
- extra values unchanged (26,256; all pairs 304,533)
- compare.py --all with schemas: 0 conversion and 0 schema failures
on pairs
Story chunk 0x2EE also holds the text orientation (u16 at 0: 0
Horizontal, 1 Vertical) and the frame type (u16 at 14: 0
TextFrameType, 1 FrameGridType). Over the 28,669 stories of all corpus
pairs both match the IDML in every story; stories without the chunk
are all TextFrameType and Horizontal. The converter writes
StoryPreference FrameType and StoryOrientation from it; the
orientation of the story's frames is used only for an unknown code.
compare.py --exclude own/:
- trustworthy coverage 97.55% -> 97.71% (11,986,627 -> 12,006,265);
StoryPreference FrameType 0 -> 19,424 of 19,424, StoryOrientation
19,210 -> 19,424 of 19,424
- all pairs 95.72% -> 95.88%
- extra values 26,256 (unchanged); all pairs 303,866 -> 304,533: the
two attributes on the stories of stale pairs that the reference does
not have
Regenerate src/idml/element_values.xml from the 827 distinct IDML files
of the corpus (683 before). Values that now vary across the IDML files
drop out:
- Story AppliedNamedGrid, StoryPreference FrameType and
StoryOrientation (stories with named grids, frame grids or vertical
text)
- Page GridStartingPoint (28 pages of DOM 18)
- TOCStyle SetStoryDirection (3 styles are Vertical)
- 9 TextDefault, 2 CjkGridPreference and 1 IndexOptions preference values
element_values.py now collects ClippingPathSettings only for
ClippingType="None", the only case in which the converter writes the
element, and leaves user-named trap presets out of the TrapPreset list.
Without these two rules a DOM 12 file with its own trap preset and 8
images with a clipping path would have removed the values for every
document.
root_values.py read the text attribute table from src/idml/mod.rs,
which moved to attrs.rs; it now runs again. Its files stay pinned and
are not regenerated (idml-values.md says why).
compare.py --exclude own/ (803 pairs, 606 trustworthy):
- trustworthy coverage 97.95% -> 97.55% (12,036,295 -> 11,986,627 of
12,287,602); all pairs 96.12% -> 95.72%
- the loss is the dropped Story AppliedNamedGrid (19,416), FrameType
(19,368), Page GridStartingPoint (10,052), TOCStyle
SetStoryDirection (618) and StoryOrientation (214)
- extra values 26,257 -> 26,256 (all pairs 305,366 -> 303,866)
- conversion failures 0 of 803
The last u16 of page chunk 0x528 is the column direction (1 vertical,
0 horizontal), inherited from the master page with the columns. It
replaces the constant Horizontal the writer wrote on every page.
compare.py --exclude own/:
- trustworthy value coverage 98.03% -> 98.03% (+14 values); all pairs
+14
- Page/MarginPreference ColumnDirection 9,274 of 9,274 (14 vertical)
- extra values unchanged; no key lost values
The writer wrote these as literals. element_values.py now collects
Story, XmlStory, Story/StoryPreference, the InCopyExportOption of
stories, rectangles, ovals and polygons, the ClippingPathSettings of
images, PDF and EPS graphics, and Ink; the writer reads them from
element_values.xml after the values it reads from the INDD.
- UserText (a WhenWritten value) is written from version 11.2, where
IDML starts to have it (was 12).
- Ink ConvertToProcess is read from byte 56 after the ink name; it was
the only ink value that varies (3 inks true).
- InCopyExportOption of ovals and the EPS clipping values follow the
versions the corpus shows (ovals from DOM 8, EPS up to DOM 20).
compare.py --exclude own/:
- trustworthy value coverage 98.03% -> 98.03% (+771 values: UserText
768, ConvertToProcess 3); all pairs 96.12% -> 96.13% (+1,211)
- extra values: trustworthy unchanged; all pairs +43 (UserText in
stale pairs whose IDML is older)
- compare.py --all with schemas: 0 schema failures on pairs; files
without a reference as before
New codes for Position, Justification, Composer, KerningMethod,
ParagraphBreakType, StartParagraph, OTFFigureStyle, SpanColumnType and
BulletsAndNumberingListType, each found by aligning INDD runs with IDML
ranges and styles by name; no run or style contradicts them.
compare.py --exclude own/:
- trustworthy value coverage 98.01% -> 98.03% (+1,859 values); all
pairs 96.10% -> 96.12% (+2,700)
- extra values: trustworthy unchanged; all pairs +13 (stale pairs)
- no key lost values
Hyphenation, glyph and word spacing, keeps, rules above, underline and
strikethrough types, colours and tints, numbering format, alignment and
level, paragraph and character direction, OpenType features, CJK
settings (aki, kinsoku, ruby, kenten) and paragraph border and shading
widths, offsets and weights. Each ID is present on a style exactly when
the IDML style has the attribute. Rule, underline and strikethrough
stroke types are a reference and a built-in code (new layout
RefOrCode); NumberingFormat is written as a string.
MergeConsecutiveParaBorders is written from version 13.1, as in IDML.
compare.py --exclude own/:
- trustworthy value coverage 97.66% -> 98.01% (+41,082 values); all
pairs 95.64% -> 96.10% (+74,262)
- of 121,196 values of these attributes, 120,409 reproduced and 3 wrong
(all 3 were wrong before)
- extra values: trustworthy unchanged; all pairs +218 (ranges and
styles of stale pairs that their IDML lacks)
- compare.py --all with schemas: 0 schema failures on pairs; files
without a reference as before
Preferences chunk 0x1A49B lists the default numbering list and the
others; each list (class 0x1A483) has its name and two flags in chunk
0x1A49A. The converter writes every list in designmap.xml, the default
first, in place of the constant default list, which it keeps only for a
file whose lists it cannot read. Paragraph attribute 0x1A417
(AppliedNumberingList) names a list by UID.
compare.py --exclude own/:
- trustworthy value coverage 97.65% -> 97.66% (+896 values); all pairs
95.63% -> 95.64% (+1,231)
- NumberingList 705 of 705 (216 more), AppliedNumberingList +32
- extra values unchanged; no key lost values
- diff_outputs.py: other outputs differ only in the position of the
AppliedNumberingList property
Story chunks 0xA44C (StoryTitle), 0x50F96 (StoryDirection, 1 right to
left) and 0x2EE (OpticalMarginSize and OpticalMarginAlignment) replace
the constant values the writer wrote for every story. Without a chunk
the story keeps the value of every story without it.
compare.py --exclude own/:
- trustworthy value coverage 97.64% -> 97.65% (+1,566 values); all
pairs 95.61% -> 95.63% (+3,461)
- StoryDirection 780, OpticalMarginAlignment 441, OpticalMarginSize 315
and StoryTitle 30 more stories right
- extra values unchanged; no key lost values
- Paragraph and character styles: SplitDocument, EmitCss and
IncludeClass from the u16 values after the tag map count of chunk
0x28F0 (when the style has no tag maps); false, true, true without
the chunk.
- Page items of master spreads: AllowOverrides from chunk 0x142D
(1 false, else true). Placed graphics and items of document spreads
have none.
- Document/Properties/Label from document chunk 0x1630B.
compare.py --exclude own/:
- trustworthy value coverage 97.48% -> 97.64% (+18,499 values); all
pairs 95.46% -> 95.61% (+22,885)
- AllowOverrides 11,791 of 11,791; paragraph style SplitDocument +3,403
and IncludeClass +2,869; document label 436 of 438
- extra values +3 (ParagraphStyle SplitDocument); no key lost values
- compare.py --all with schemas: 0 schema failures on pairs; files
without a reference as before
From the preferences object:
- TextDefault AppliedParagraphStyle and AppliedCharacterStyle (chunks
0x28D4, 0x28D5); PageItemDefault graphic, text and grid object styles
(chunk 0x1B959).
- LayoutGridDataInformation and StoryGridDataInformation (chunks 0xCD2F,
0xCD2E): font, font style, sizes, akis, scales, character count size.
- MasterTextFrame and CreatePrimaryTextFrame (chunk 0x59C),
BlendingSpace (0x1081F), AllowPageShuffle, GuidesShown, snapping,
GuidesInBack, BaselineGridRelativeOption, ZeroPoint, layout
adjustment, ShowTextThreads (from version 21.1), AbutTextToTextWrap
and HonourTextIndentsWithTextWrap.
- Interface colours of guides, grids, margins, columns and the
pasteboard, with four more named colours (LightGray, GridBlue, Peach,
Burgundy).
Node::merge no longer adds observed text or children to a typed value,
so a colour list is not merged with an observed colour name.
compare.py --exclude own/:
- trustworthy value coverage 97.42% -> 97.48% (+6,332 values); all
pairs 95.41% -> 95.46% (+8,463)
- extra values: trustworthy unchanged; all pairs +1 (ShowTextThreads in
one pair whose IDML is from an earlier version)
- compare.py --all with schemas: 0 schema failures on pairs; files
without a reference as before
- FontType codes 0 Type1, 3 ATC and 0xFFFFFFFF Unknown.
- TypekitID from INDD version 9.2 (was 12).
- The key flag of the native style name gives FontStyleNameNative a
$ID/ prefix; the key flag of the family name gives $ID/ to the family
name in FontFamily, Font Name and every AppliedFont.
- Missing-font records (kind 3) parse: no fonts, or short font records
with style, platform name and Unknown type.
- The family name in the font's own script replaces the name when the
family's writing script is the script byte of the document users.
compare.py --exclude own/:
- trustworthy value coverage 97.36% -> 97.42% (+7,627 values); all
pairs 95.35% -> 95.41% (+9,808)
- TypekitID +3,483, FontStyleNameNative +1,262, FontType +664, 196
more Font elements matched, AppliedFont and BulletsFont +235
- extra values 26,348 -> 24,144 (Font elements whose Self now matches)
- warnings 413 -> 406 (missing-font records now parse)
- compare.py --all with schemas: 0 schema failures on pairs; files
without a reference as before
Chunk 0x1307E is now read field by field: each group's name, include
byte and header variant, then its header variants and their sections.
This replaces the search for group names. objects.md gives the layout.
diff_outputs.py: 0 differing outputs, 0 changed failures and warnings
over 4,274 files.
The index object (class 0x13004) lists its sections (class 0x13005);
chunk 0x13007 of a section holds its topic records depth first, with
their level, name, sort order and page references. The converter
writes the Index element in designmap.xml with its nested Topic
elements, and gives each PageReference its ReferencedTopic and
PageReferenceType. Evidence: docs/format/index.md.
compare.py --exclude own/:
- trustworthy value coverage 97.17% -> 97.36% (+21,617 values); all
pairs 95.21% -> 95.35% (+21,808)
- Topic 2,185 of 2,185 with Name and SortOrder; Index 16 of 16;
PageReference ReferencedTopic and PageReferenceType 7,523 of 7,523
- extra values +15: the 3 topics and 3 page references one INDD has
and its earlier IDML lacks
- compare.py --all with schemas: 0 schema failures on pairs; files
without a reference as before. One of them (DOM 19) has topic
records of another length; its sections are left out with a warning
(21 warnings).
- IndexHeaderSetting (chunk 0x1300E): its five attributes and the whole
Properties/ListOfIndexHeaderGroup tree.
- IndexOptions (chunk 0x13010): title, title and level styles,
separators and flags. Styles are matched by name among the styles
outside groups. Without the chunk no Title is written: it follows the
exporting application's language.
- ChapterNumberPreference (chunk 0x1A4C4): format, source, number.
- DictionaryPreference Composition: UseDocument with chunk 0x2806,
Both without.
- EPubExportPreference (DOM 8 on): Version and Id from chunk 0x21A1A;
without it the observed Id, TocStyleName $ID/, and Version Epub3
from INDD version 18.1.
compare.py --exclude own/:
- trustworthy value coverage 97.15% -> 97.17% (+2,465 values); all
pairs 95.19% -> 95.21% (+3,281)
- ListOfIndexHeaderGroup 489 of 489; IndexOptions Title 10 of 10 with
the chunk; Composition wrong in 5 documents (Both with the chunk);
TocStyleName with the chunk is not written (13)
- extra values unchanged; no key lost values; 0 failures
- compare.py --all with schemas: 0 schema failures on pairs; the files
without a reference fail as before (22 conversions, 11 schemas)
XMLImportPreference AllowTransform, ImportCALSTables,
ImportTextIntoTables, ImportToSelected, RemoveUnmatchedExisting,
RepeatTextElements and DictionaryPreference MergeUserDictionary are the
same in all 654 pairs. Two of the 683 IDML files differ; both were
exported by an earlier version than their INDD was last saved with, so
they are not pairs. element_values.py now leaves those files out for
these keys (LATER_INDD_EXCEPTIONS); idml-values.md records why.
compare.py --exclude own/: trustworthy value coverage 97.12% -> 97.15%
(+3,423 values, 7 keys x 489 pairs); all pairs 95.16% -> 95.19%
(+4,578). No key lost values; extra values unchanged.
Spreads without chunk 0x10833 get LineArtAndTextResolution 300 and
GradientAndMeshResolution 150. No INDD field decides these values:
5,045 of the 5,074 trustworthy spreads without the chunk have them; the
other 29 (all spreads of 9 documents) have 400 and 400. The evidence is
in objects.md and idml-values.md.
compare.py --exclude own/:
- trustworthy value coverage 97.03% -> 97.12% (+10,090 values); all
pairs 95.07% -> 95.16% (+13,836)
- Spread/FlattenerPreference resolutions: 5,074 missing -> 29 wrong
- extra values: trustworthy unchanged; all pairs +306 (the attributes
on spreads of stale pairs that their IDML no longer has)
Chunks 0xA4C and 0xAF2 are read to their end: device type, paper
offset range, colour output, screening, scaling, tiling and thumbnails,
graphics and font download, printer marks, bleed, slug, profile, ink
screens, flattener preset and the paper size selector. AllPrinterMarks
and PreserveColorNumbers are derived from them. Printer and PPD follow
the corrected rules (built-in empty printer, device-independent PPD
only for PostScript file output). Chunk 0xAF0 gives the four
PrintBookletOption margins. Booklets leave out the attributes their
element does not have.
compare.py --exclude own/ (trustworthy pairs):
- value coverage 96.85% -> 97.03% (+21,049 values; all pairs
94.89% -> 95.07%, +28,090)
- PrintPreference and PrintBookletPrintPreference: every attribute
reproduced in all 489 pairs except PreserveColorNumbers (2) and
MarkType (1); PrintBookletOption margins 11 -> 0 documents wrong
- extra values unchanged; no key lost values; 0 failures
- Notes (class 0xA429): their text and a Note element in place of the
U+FEFF, with dates by the link time rule, user name and document user.
- Index markers (class 0x13006): a PageReference with Self and Id in
place of the U+FEFF (topics are not written yet).
- Tracked changes: deleted text (class 0xA40A) as Change DeletedText
before the following character; inserted runs (strand run kind
0xA466) wrapped in Change InsertedText inside each character range,
closed at tables, anchored items and XML markers.
- XML markers are left out of the text when the node store cannot be
read.
- Character-level sources open before an element character (note,
destination), apart from paragraph destinations, which the schema
does not allow in them.
Trustworthy pairs: reproduced 10,772,444 -> 11,266,452 (92.60 % ->
96.85 %); stories with identical text 18,574 -> 18,602 of 18,610;
PageReference 7,523 with Id; 1,709 more hyperlink sources and
hyperlinks (in deleted text). Notes 211 of 211 and changes 6,024 of
6,027 match the reference attributes (multisets per pair, no Self).
Extra values 26,107 -> 26,333: 220 from Change elements written inside
sources where InDesign writes an empty Change before them, 9 from
PageReference. HyperlinkTextSource/Note loses 3 elements for the same
reason (the source's first child is now the Change). Schema
validation: 0 failures.
From DOM 14 IDML repeats a frame's footnote options as
TextFrameFootnoteOptionsObject on the frames whose TextFramePreference
has FootnotesEnableOverrides. The converter writes it on those frames,
with the values of that preference.
Trustworthy pairs: reproduced 10,650,374 -> 10,669,264 (92.54 % ->
92.71 %); 3,778 elements with all four values, no extra value (26,107).
No key reproduces fewer values. Schema validation: 0 failures.
- Footnotes: text owned by a footnote (class 0x24F) is read into the
footnote and written as a Footnote element in place of its U+0004;
hyperlink sources inside footnotes are placed like the others.
- FootnoteOption from preferences chunk 0x2820 (styles, numbering,
marker, spacing, strings, splitting, straddling, both rules); without
the chunk, the text style and straddling every IDML has. The value
file reader decodes numeric character references, which fixes the
SeparatorText tab.
- EndnoteOption from chunk 0x2261E; default marker and text styles
without it. IsEndnoteStory from DOM 13 instead of 15.
- Endnotes: the endnote story (class 0x2801) is converted, references
(class 0x2805) become Endnote elements, ranges (class 0x2804, in the
range tree) EndnoteRange, and its frames EndnoteTextFrame, as InDesign
writes them.
- compare.py accepts the schema errors of that endnote markup when the
reference IDML of the pair has the same errors (docs/measurement.md).
Trustworthy pairs: reproduced 10,579,111 -> 10,650,374 (91.92 % ->
92.54 %); stories with identical text 18,565 -> 18,574 (0 missing);
documents >= 99 % 20 -> 21; TextRange +62,059, FootnoteOption and
IsEndnoteStory complete, EndnoteOption +1,356, hyperlinks in footnotes
+271. Extra values 26,179 -> 26,107. No key reproduces fewer values.
Schema validation: 0 failures; 1 pair accepted by the endnote rule; 11
files without a reference fail only on endnote markup.
InDesign 5.0 destinations have no key, hyperlinks name their destination
by UID, text destination owners use chunk 0x1351E, and text sources are
placed by start and end markers owned by the story (chunk 0x13524)
instead of a range tree. These are read by version; keys are left out
where there are none.
No 5.0 file has an IDML. Over the 17 5.0 files with hyperlinks: 121
hyperlinks, 118 text sources, 7 text destinations and 45 bookmarks are
now written, the 116 warnings about their destination objects are gone,
and every output validates. Pair results unchanged (91.92 %, extra
values 26,516). Schema validation: 0 failures.
Class 0x13552 objects are written as HyperlinkExternalPageDestination:
hidden and key from chunk 0x13580, page index from chunk 0x1B8, view
from chunk 0x13527, and the name IDML derives from the linked
document's file name. DocumentPath is left out (it starts with a volume
name the INDD does not hold). Hyperlinks of kind 2004 and bookmarks now
find them.
Trustworthy pairs: reproduced 10,566,138 -> 10,578,767 (91.81 % ->
91.92 %); 643 of 643 external page destinations with all values but
DocumentPath, 576 more hyperlinks, bookmarks 452 -> 1,027 of 1,027.
Extra values unchanged (26,516); no key reproduces fewer values. Schema
validation: 0 failures.
- A text source over several character ranges of one paragraph range is
written as a child of the ParagraphStyleRange: the ranges at its ends
are split and the parts inside become its CharacterStyleRange
children. Sources across paragraph ranges, or holding an XML marker,
are left out with a warning.
- Text sources with chunk 0x135A0 are written as CrossReferenceSource
with AppliedFormat. One that holds more than a text variable instance
is written at paragraph level, since the schema allows no Content in
it.
Trustworthy pairs: reproduced 10,530,141 -> 10,566,138 (91.50 % ->
91.81 %); HyperlinkTextSource 3,449 -> 5,475, CrossReferenceSource 116,
hyperlinks 3,097 -> 5,149, all with their values. Extra values 26,634 ->
26,516: the 332 values of cross-reference sources written as hyperlink
sources go, and cross-reference sources written around a character
range add their inner ranges (one pair). No key reproduces fewer values.
"spans several text ranges" warnings over all files 9,703 -> 3,277.
Schema validation: 0 failures.
Class 0x13504 objects are read as text or paragraph destinations (kind
in chunk 0x13526). The story's owned items of class 0x1353C name them at
their U+FEFF, which is now written as an empty HyperlinkTextDestination
or ParagraphDestination element instead of the character. Hyperlinks and
bookmarks can point at them.
Trustworthy pairs: reproduced 10,513,964 -> 10,530,141 (91.36 % ->
91.50 %); stories with identical text 18,517 -> 18,565; documents at
>= 99 % 18 -> 20; HyperlinkTextDestination 685 of 685,
ParagraphDestination 119 of 119, bookmarks 373 -> 452. Extra values
unchanged (26,634); no key reproduces fewer values. Schema validation: 0
failures. The 7 text destinations of InDesign 5.0 files (no key) are
left out with a warning, like their page and URL destinations.
- Text source chunk 0x135B7 is decoded (type, anchor name, index marker,
page number string, level) and written as AlternativeDestination;
control characters in the page number become <?AID 00xx?> text.
- Page item sources (class 0x13503) are written as
HyperlinkPageItemSource when their page item is written, and their
hyperlinks with them.
Trustworthy pairs: reproduced 10,511,339 -> 10,513,964 (91.36 %);
AlternativeDestination 463 -> 1,048 of 1,048, HyperlinkPageItemSource
136 of 136, 136 more hyperlinks. Extra values unchanged (26,634); no key
reproduces fewer values. Schema validation: 0 failures.
- Hyperlink: Visible, Width, BorderStyle and BorderColor whenever chunk
0x13553 has the corpus pattern, Highlight from its byte 10, no
BorderColor for hyperlinks to a page of another document.
- Hyperlink destination: left out for hyperlinks into another document
(chunk 0x1359F) instead of a wrong local destination; `n` for kind
2000 without a destination; hyperlinks without a source are left out.
- Page destination: ViewBounds from chunk 0x13527 and NameManually.
Trustworthy pairs: reproduced 10,505,081 -> 10,511,339 (91.28 % ->
91.33 %); 22 more hyperlinks, Hyperlink appearance keys 1,982 -> 2,878
of 2,878, page destination ViewBounds and NameManually 831 of 831,
Destination wrong 0. Extra values unchanged (26,634); no key reproduces
fewer values. Schema validation: 0 failures.
Groups (class 0xCA8B) listed by the root cell or table style group are
written as CellStyleGroup or TableStyleGroup elements holding their
styles, in the order of the group's children (objects.md).
compare.py --exclude own/, trustworthy pairs: CellStyleGroup elements
0 -> 2 of 2 (4 values); no other change, no new extras, no conversion
or schema failures.
New table and table style attributes: stroke and fill counts, end row
and start/end column stroke colours, weights, types and tints, column
fill colours and tints, skip counts, graphic insets and
ClipContentToGraphicCell; on tables only LeftInset etc. and
BreakHeaders; on styles only the footer region style and the region
flags. Groups whose IDs are always equal (border weights, colours,
types, tints and gap colours; StartColumnStrokeType/EndColumnLineStyle;
SkipFirstHeader/SkipLastFooter) are written only when all IDs agree.
Table stroke types are decoded as two words. The root table style takes
its INDD values where the observed root values have the attribute, and
from DOM 11 its text and graphic cell values. A cell style without its
own paragraph style gets [No paragraph style] (tables.md).
compare.py --exclude own/, trustworthy pairs: value coverage 94.23 % ->
94.24 %; every Table key above is now right on every table, TableStyle
Text*/Graphic*Inset and clip flags 341-348 -> 375 of 375, border and
region keys up, CellStyle AppliedParagraphStyle 507 -> 547 of 547. No
new extras in trustworthy pairs; 98 in a stale pair's table that its
reference lacks.
An edge value is written when every grid position along that edge has
it with the same value; the priority comes from the cell's own set, or
is 1 for the right (bottom) edge of a merged cell whose edge positions
have one. New edge groups: gap tint, gap colour, overprint and gap
overprint. Cell edge tints -1 are written as 100 and a weight is 0 when
the edge colour is None. Stroke codes 0x5A3F (Japanese Dots) and 0xB007
(ThickThick) for cells and tables. New cell attributes: graphic insets,
OverprintFill, VerticalJustification 0 and 3, FirstBaselineOffset,
WritingDirection (not on cell styles: the schema lacks it),
RotationAngle, DiagonalLineStrokeOverprint. Text*Inset only from DOM 11.
Cell style edge IDs are rotated against the cell IDs (tables.md).
compare.py --exclude own/, trustworthy pairs: value coverage 93.13 % ->
94.23 %, median 96.7 % -> 96.8 %, documents >= 99 % 14 -> 18; extra
values 30,379 -> 30,237. Every Cell edge key is now right on every cell
except one bottom weight (tables.md); CellStyle edge keys rise too.
One unnamed colour (14 values) is no longer written: its only reference
was a wrong BottomEdgeStrokeColor that is now right, and IDML writes the
colour though nothing refers to it (objects.md). In stale pairs the
extras of a table missing from its reference grow by 2,590.
Cells start at positions with attribute 0xB666 (text) or 0x10469
(graphic); positions past a short row record are covered. The layout
record is parsed by its flags, so the width is read where it is, and
cells never laid out are kept. Spans follow a two-pass rule: the column
span whose widths best match the layout width plus the insets in effect,
then the free covered rows below. From DOM 11 every Table, Row, Column
and Cell gets TextTop/Left/Bottom/RightInset and ClipContentToTextCell
in effect: cell set, cell style chain, table-style region cell style,
table, table style chain, root table style; rows and columns copy their
first cell (tables.md).
compare.py --exclude own/, trustworthy pairs: value coverage 92.51 % ->
93.13 %, median 96.5 % -> 96.7 %; story text 18,509 -> 18,517. All
Cell, Row, Column and Table text insets and ClipContentToTextCell are
now right (Cell 11,097 of 11,097 each), Cell ColumnSpan 11,304 -> 11,550
and RowSpan 11,389 -> 11,550 of 11,550; 144 overset cells are no longer
dropped. No new extras in trustworthy pairs; the extras of one stale
pair's table that its reference lacks grow by 1,052.
Chunk 0x50F65 gives TableDirection (written on every table); the row
group attributes 0xB69F, 0x10407 and 0xB6A1 give AutoGrow (now also
true), StartRow and KeepWithNextRow (tables.md).
compare.py --exclude own/, trustworthy pairs: value coverage 92.50 % ->
92.51 %; Table TableDirection 0 -> 326 of 326, Row AutoGrow 63 -> 1,128
of 1,128, StartRow 10 of 10, KeepWithNextRow 8 of 8. No new extras in
trustworthy pairs; 7 in one stale pair's table that the reference lacks.
A style's IDML Name is its group path and its own name joined by ":",
with a ":" inside a name written "\:" and "$ID/" before the whole path
of a built-in style; Self is that name escaped (objects.md, style
groups). Object and cell styles now include their group path too, and
every reference uses the same name. A font with an empty style name is
named "<family> Regular" with the family alone in Self (fonts.md).
compare.py, trustworthy pairs: extra values 33,181 -> 30,379 (0.27 %);
value coverage 92.45 % -> 92.50 % (10,307,358 values; grouped style
names, styles now matched); no value lost; 14 documents >= 99 %,
median 96.5 %. No conversion or schema failures; story text
18,509/18,610.
IDML writes a colour or gradient without a name only if a value in the
package refers to it, a gradient stop or tint uses the colour, or the
page item defaults name it in one of the two entry tables of chunk
0x6E07 (objects.md, colours; preferences.md). The graphic part is now
written last, from the references in the other parts; the second table
of chunk 0x6E07 is read.
compare.py, trustworthy pairs: extra values 155,956 -> 33,181 (unnamed
Color -114,734, Gradient -3,498, GradientStop -4,543). 463 values are
no longer reproduced: 10 unnamed colours and 22 unnamed gradients with
their stops that IDML writes though nothing refers to them (open, see
objects.md). Value coverage 92.46 % -> 92.45 %. No conversion or schema
failures; story text 18,509/18,610.
The previous commit (09e87fa), measured alone: extra values 635,511 ->
155,956, value coverage 92.41 % -> 92.46 %; 39 values lost (7 opacity
stops and 8 drop shadows IDML writes with values equal to the style,
5 StrokeWeight, 1 TextColumnFixedWidth).
IDML leaves out the values of a page item that equal those of its
applied object style. The writer now compares with the style's
effective values (its own, then the styles it is based on, then [None]):
- TextFramePreference: an attribute is written when its category is off
in the style's category list (chunk 0x1B92E, also for [None]) or its
value differs; TextColumnMaxWidth (DOM 8), TextColumnCount (DOM 10)
and InsetSpacing (DOM 11) are always written; footnote values from
13.1 (objects.md, text frame preferences).
- Fill, stroke, miter and corner attributes (attributes.md).
- Transparency effects: only differing values, and an effect element
only when it has one or its opacity stops differ (transparency.md).
New decoding: object style chunk 0x1B924 fields for balance, fixed
width, justification, first baseline offset (code 4 = XHeight, also for
frames) and auto-sizing; chunk 0x1B92C (style transparency). Object
styles get the DOM 12 text frame values from DOM 8 and
FootnotesEnableOverrides from 13.1 (idml-values.md).
Rules from the corpus, each documented in docs/format/:
- ItemLayer only on items directly under a spread (objects.md).
- The root object style [None] gets no Enable… attributes and no effects
category settings; EnableTextFrameColumnRuleOptions from 15.1, and the
footnote values of an object style's TextFramePreference from 13.1.
- The root paragraph style has no NextStyle and no AllNestedStyles;
TextDefault has AllNestedStyles only when the list has items; the
empty TabList is written from DOM 8 (attributes.md).
- DOM 7 has no EPubExportPreference or HTMLExportPreference.
- Row and column Self use a hexadecimal index (tables.md).
- Cell CellType from DOM 11; Document Name from DOM 13; ColorGroup Self
is the UID before 11.3.
- The internal layer is the first of the document's layer list, whatever
its name.
- A CR in a name is %0d in Self (colours, gradients, tints, hyperlink
destinations, styles).
compare.py, trustworthy pairs (after the compare fix of 48d184f):
extra values 705,523 -> 635,511; value coverage 92.25 % -> 92.41 %
(10,296,599 values); 2 ItemLayer values lost (the 2 grouped items whose
IDML has one); 12 documents >= 99 %. No conversion or schema failures.
Page items, object styles and PageItemDefault: each corner radius and
corner option from its own attribute ID (top left 0x6E70/0x6E6F, top
right 0x6E94/0x6E91, bottom left 0x6E92/0x6E8F, bottom right
0x6E93/0x6E90), with the corner codes InsetCorner, BevelCorner and
FancyCorner. Object styles no longer need all four corners equal.
Paragraph styles and text: ParagraphShadingTopLeft/BottomLeftCornerRadius
(0x1DF12, 0x1DF14); the tied groups of border corner radii, shading
right corner radii and shading corner options when their IDs agree; the
shading and border origin pairs for the known code pairs.
compare.py --all --exclude own/ (trustworthy pairs): value coverage
91.61% -> 92.25% (+71,545 values); documents >= 99 %: 5 -> 11; median
95.7% -> 96.3%; no key lower. Extra values 709,654 -> 718,266
(+8,612): the per-corner attributes follow the presence of CornerRadius
and CornerOption in the item's list, so each of the existing 2,055
extra CornerRadius values (items whose IDML leaves the value to the
object style) now has four per-corner extras, and likewise the
CornerOption extras. 0 conversion and 0 schema failures.
Links: LinkResourceModified, ShowInUI, LinkImportStamp and the two link
times from chunk 0x8C9B, PDFIdentifier from chunk 0x1B6, the size and
format name from the resource's chunk 0x8C92, and the values every link
has. Link times are written in the time zone of the XMP date nearest in
day of the year (the INDD has no time zone; compare.py does not score
them): 3,886 of 4,446 import times and 3,316 of 4,434 modification
times right.
Images, PDF, EPS and SVG graphics: the page item settings and object
style, Space, ActualPpi and EffectivePpi (chunk 0x1708 and the scale on
the spread), ImageTypeName, Profile (chunk 0x7C0F), fill colour and
gradients, vector colour policies (chunk 0x7C42 or the document's
policies), PDFAttribute (chunk 0x251B), GraphicLayerOption with the
GraphicLayer tree (chunk 0x177A) and LayerCompOption (chunk 0x9209),
and ImageIOPreference from chunk 0x1714 in place of chunk 0x8C39.
The placed graphic writer moves to idml/graphic.rs.
compare.py --all --exclude own/ (trustworthy pairs): value coverage
90.00% -> 91.61% (+179,834 values); documents >= 99 %: 1 -> 5; median
94.2% -> 95.7%; no key lower. Extra values 709,643 -> 709,654 (+11 in
SVG elements that are not in the reference). New warnings: 178 layer
chunks in 3 InDesign 3.0 files, whose layout differs, are left out.
0 conversion and 0 schema failures.
ObjectExportOption now takes its alternative text, actual text and
tagging values from chunk 0x1E206 of the item, and its other attributes
from a table by INDD major and minor version (8.0-9.1, 9.2, 10.0, 10.1
on, 21.0, 21.1, 21.4), for page items from version 8 and for every
object style, [None] included, from version 9. Page items of version 7
get no ObjectExportOption, as in their IDML.
compare.py --all --exclude own/ (trustworthy pairs): value coverage
88.72% -> 90.00% (+142,324 values); documents >= 99 %: 0 -> 1; median
92.4% -> 94.2%; no key lower. Extra values 713,688 -> 709,643: -4,319
from the DOM 7 rectangles that no longer get the element; +274 more
values in GraphicLine, Oval and TextFrame elements that are not in the
reference (more attributes per element). 0 conversion and 0 schema
failures.
IDML writes the shortest decimal that reads back as the same value; when
that needs 17 digits it takes the 17-digit decimal nearest to the exact
value, with ties to even (150.23599243164062). Rust's shortest form can
end in the other digit (…063). num() now re-rounds 17-digit output with
{:.16e}.
compare.py --all --exclude own/ (trustworthy pairs): value coverage
88.72% (9,885,604 -> 9,885,761 values); PDF P.GraphicBounds +148,
EPS P.GraphicBounds +9; no key lower; extra values unchanged (713,688);
0 conversion and 0 schema failures.
Page items and page item defaults knew only code 0x5A15; object styles
also knew 0 (None) and 0x5A16 (InverseRoundedCorner). Using the object
style codes everywhere reproduces 486 more reference values in the
trustworthy pairs (489 of 489 page item defaults, 4 rectangles, 2 text
frames) and adds no extra values. Object styles write CornerOption in
the page item attribute order.
Groups that contain themselves or nest over 100 deep, style group cycles,
table counts larger than the cell data, slot numbers outside their page,
cyclic slotted-record continuations and non-finite numbers now give a
warning or an error instead of a crash or invalid output. Corpus output
is unchanged (tools/diff_outputs.py).