# Values written from IDML observation Some values the converter writes do not come from the INDD file. They are values that every IDML exported by InDesign in the corpus has, on the root styles and in the preferences. They are facts about InDesign's IDML output, not about the INDD format, and the evidence is the corpus IDML files alone. Why they are written: IDML consumers apply their own defaults to attributes that are missing. When those defaults differ from InDesign's, the layout differs. For example, a consumer that defaults to keeping all lines of a paragraph together breaks pages differently from a document whose root style says `KeepAllLinesTogether="false"`. ## Root style values The root styles are `[No paragraph style]`, `[No character style]`, the object style `[None]`, the cell style `[None]` and the table style `[No table style]`. Every corpus IDML has all five. `tools/root_values.py` reads `Resources/Styles.xml` of every corpus IDML and collects each attribute of the root styles, each `Properties` child, and each attribute of their child elements (for example `TextFramePreference` in the object style). It keeps a value when: - every IDML from some DOM version on has it, with the same value, and - the converter does not read it from the INDD. The kept values are in `src/idml/root_values.xml`, generated by that tool. The converter writes them on its root styles. A value the converter reads from the INDD takes precedence: for the paragraph and character root styles, attributes in the text attribute table (`attributes.md`) are left out of the file, and any attribute the INDD supplies is written from the INDD. The file is generated from the reference IDML of every pair that `tools/compare.py` compares (a little-endian INDD whose IDML has the same major version, without the privately held samples): 1,585 pairs, 1,448 distinct IDML files, DOM 7 to 21. Values that first appear after DOM 7 are written only for documents of that InDesign version or later (`MinimumVersion` in the file). | Root style | Values | Same in | Written for version | |---|---|---|---| | `[No paragraph style]` | 72 | 1,448 of 1,448 files | all | | | 1 | 1,437 of 1,437 files with DOM 8 or later | 8 and later | | | 3 | 1,331 of 1,331 (DOM 10 or later) | 10 and later | | | 3 | 1,211 of 1,211 (DOM 11 or later) | 11 and later | | | 14 | 1,039 of 1,039 (DOM 13 or later) | 13 and later | | | 1 | 640 of 640 (DOM 15 or later) | 15 and later | | | 1 | 61 of 61 (DOM 21) | 21 | | `[No character style]` | 1 | 1,448 of 1,448 | all | | | 2 | 1,211 of 1,211 (DOM 11 or later) | 11 and later | | | 1 | 1,039 of 1,039 (DOM 13 or later) | 13 and later | | | 1 | 640 of 640 (DOM 15 or later) | 15 and later | | | 1 | 61 of 61 (DOM 21) | 21 | | Object style `[None]` | 63 | 1,448 of 1,448 | all | | | 9, 22, 1, 4, 7, 4, 4, 8 | all files with DOM 8, 9, 10, 11, 12, 13, 14, 15 or later | from that version | | | 9 | 61 of 61 (DOM 21) | 21 | | Cell style `[None]` | 1 | 1,448 of 1,448 | all | | Table style `[No table style]` | 89 | 1,448 of 1,448 | all | | | 6 | 1,211 of 1,211 (DOM 11 or later) | 11 and later | | | 4 | 254 of 254 (DOM 18 or later) | 18 and later | **Left out because they vary between files:** - `[No paragraph style]` (19): `HyphenQuality`, `StyleUniqueId`, the origins and some corner radii of paragraph borders and shading, and the `StyleExportTagMap` child. - `[No character style]`: `StyleUniqueId`. - Object style `[None]` (42): among them `AppliedNamedGrid`, `AppliedParagraphStyle`, `StrokeWeight`, `BaselineFrameGridColor`, the corner radii, `OpticalMarginSize`, `TextColumnGutter`, `FirstBaselineOffset`, the overprint settings, the flexible layout paddings and gaps, and some export options. - `[No table style]` (19): the border, row and column stroke weights, `SpaceBefore`, `SpaceAfter`, the end stroke tints, the text insets, `CaptionPosition`, `CaptionParagraphCount` and `HeaderColumnCellStyle`. The four region cell styles (`HeaderRegionCellStyle` and the others, `n`) are on every table root style from DOM 18 only: 24 files of DOM 10 and 17 lack them. The converter reads many of these from the INDD: the object style settings (`objects.md`, object style settings and named grids), the baseline frame grid colour (`preferences.md`), the attributes of the root table style (`tables.md`), the paragraph border and shading values (`attributes.md`), and the style unique IDs and export settings (`objects.md`, styles). Two kinds of IDML files are not evidence for the root styles, and the pair rule leaves them out: an IDML (DOM 7) beside an INDD of version 17.2 that has no root styles at all, and a big-endian pair of version 11.2 whose table style lacks four region cell style values that every other file has. **Left out because the converter reads them from the INDD:** 217 attributes of `[No paragraph style]`, listed in `attributes.md`. Some of these are also the same in every IDML; the INDD value is written. The object styles other than `[None]` and the preferences (below) are generated from the same pairs. Values that vary in any of them are left out, even where one document is the exception: the converter writes nothing for such a key unless it reads the value from the INDD. The cell and table root styles are written as new elements in `RootCellStyleGroup` and `RootTableStyleGroup`. ## Document attributes `AccurateLABSpots="false"` is on the `Document` element of all 654 corpus pairs' IDML files. The converter writes it. The colour settings next to it are read from the INDD (`preferences.md`). `Name` (the package's file name) is on the `Document` element from DOM 13 only: 321 of 321 trustworthy references from DOM 13, 0 of 168 before. ## Default numbering list The root paragraph style refers to `NumberingList/$ID/[Default]`. Every corpus IDML file has this element in `designmap.xml`, with the same attributes: `Name="$ID/[Default]"`, `ContinueNumbersAcrossStories="false"` and `ContinueNumbersAcrossDocuments="false"`. The converter reads the numbering lists from the INDD (`objects.md`, numbering lists) and writes them after `idPkg:Styles`, where the IDML schema puts them. It writes this default list with these values only for a file whose lists it cannot read. ## Built-in stroke styles All 240 corpus IDML files list the same 18 `StrokeStyle` elements in `Resources/Graphic.xml`, in the same order, each with only `Self` and `Name` (`StrokeStyle/$ID/Solid`, `$ID/Solid`, and so on). The converter writes the same list, so every `StrokeType` it writes refers to an element in the package. ## Story values `tools/element_values.py` collects the values of `Story`, `XmlStory` and their `StoryPreference` and `InCopyExportOption` children from the 1,495 corpus IDML files (`element_values.xml`); the converter writes them where the INDD gives no value: | Element | Values kept | Versions | |---|---|---| | `Story` | `TrackChanges="false"` | 7 on | | `Story`, `XmlStory` | `UserText="true"` where written | 11 on; IDML has it from version 11.2 (none of the 3,184 stories of DOM 10.0 to 11.0, every story from 11.2), and the converter writes it from INDD version 11.2 | | `InCopyExportOption` of stories | `IncludeGraphicProxies="true"`, `IncludeAllResources="false"` | 7 on | The other story values are read from the INDD: `StoryTitle`, `StoryDirection` and the optical margin from story chunks, with the value of every story without the chunk when it is absent (`objects.md`, story settings); `IsEndnoteStory` from the story class (`footnotes.md`); `AppliedTOCStyle` from chunk 0x8C40. `StoryPreference` `FrameType` and `StoryOrientation` are read from story chunk 0x2EE (`objects.md`, story settings). Left out because it varies: `Story` `AppliedNamedGrid` (106 of 27,816 stories name a grid, all in DOM 12 files). Before the corpus had samples typeset in Japanese, every IDML had `n`. ## Graphic settings without an INDD field `element_values.xml` holds the `ClippingPathSettings` values of every `Image`, `PDF` and `EPS` element with `ClippingType="None"` in the 1,495 corpus IDML files: `ClippingType="None"`, `InvertPath="false"`, `IncludeInsideEdges="false"`, `RestrictToFrame="false"` and `AppliedPathName="$ID/"` (and for PDF and EPS also the values the converter reads for images). The converter writes the element only for clipping type 0 in chunk 0x2C1A, which IDML writes as `None` (`objects.md`), so `tools/element_values.py` collects only those elements (`WRITTEN_FOR`). The 8 images with another `ClippingType` (7 with `AppliedPathName` set, 1 with `InvertPath` and `RestrictToFrame` `true`) are the only elements with other values in the corpus before 2026-10; the 12,280 elements with `None` in the current corpus (8,993 images, 2,213 PDF, 1,074 EPS) have the values above. Every `ImageIOPreference` has `AllowAutoEmbedding="true"`. The converter writes the values read from the INDD first (`objects.md`, clipping path settings). Over the 5,131 graphics of the trustworthy pairs, every graphic has `LocalDisplaySetting="Default"`; every image (3,607) `ImageRenderingIntent="UseColorSettings"`; every PDF and EPS (1,339) `GrayVectorPolicy="IgnoreAll"`; every SVG (146) `UseSVGAs="EmbedCode"`. From DOM 21 every image, PDF and SVG has `FlexItemWidthMode="FlexFixed"` and `FlexItemHeightMode="FlexFixed"` (81 of 81), and SVG graphics from 21.2 have `IsMathMLObject="false"` (4 of 4). The constant values of graphic layers are in `objects.md` (graphic layers). The converter writes all of these. ## Links These `Link` values are the same in all 4,699 links of the trustworthy pairs, and the converter writes them on every link: `AssetURL="$ID/"`, `AssetID="$ID/"`, `LinkClassID="35906"` (0x8C42, the class of links), `LinkClientID="257"` (the u32 at offset 4 of chunk 0x8C9B is 0x101 in every link), `LinkObjectModified="false"`, `CanEmbed="true"`, `CanUnembed="true"`, `CanPackage="true"`, `ImportPolicy="NoAutoImport"` and `ExportPolicy="NoAutoExport"`. `RenditionData="Actual"` is on every link from DOM 13 on (3,043 of 3,043) and on none before (1,656 of 1,656); the converter writes it from version 13. ## Anchored object settings of page items Where IDML writes every anchored object setting of an item (its object style turns the category off; `objects.md`, anchored object settings), the element also has `LockPosition="false"` and `AnchorSpaceAbove="0"`. `tools/element_values.py` collects them as values of the `AnchoredObjectSetting` of text frames, rectangles, ovals, polygons, graphic lines and groups that have them (`WHEN_WRITTEN`): every such element of the corpus IDML files has these two values. The value file has them for the DOM versions whose files show them (text frames and ovals 8 to 20, rectangles 12 to 20, graphic lines 13); the converter writes them from there. ## Object styles other than the root `tools/root_values.py` also collects the object styles other than `[None]`: 4,955 styles in the 1,448 distinct IDML files, every file having at least one. A value is kept when every such style of every IDML from some DOM version on has it, with the same value, and at least 10 files show it. The kept values are in `src/idml/object_style_values.xml`. The converter writes them on every object style other than `[None]`, and writes the values it reads from the INDD (`objects.md`, object style settings) in their place. | Values | Same in | Written for version | |---|---|---| | 49 | 1,448 of 1,448 files | all | | 2, 22, 1, 5, 5, 4, 6, 11 | all files with DOM 8, 9, 10, 11, 12, 13, 14, 15 or later | from that version | | 5, 3, 8 | 254, 157 and 93 files with DOM 18, 19, 20 or later | from that version | | 41 | 61 of 61 files with DOM 21 | 21 | Among them are the nine `Enable…` effect flags other than `EnableTransparency` in the four `ObjectStyle…EffectsCategorySettings` elements (all `true`), most of `ObjectExportOption`, `TransformAttributeOption` from DOM 13, and `UseFlexibleColumnWidth="false"` and `TextColumnMaxWidth="0"` of `TextFramePreference` from DOM 8. **Left out because they vary or are not in every style** (73): the fill, stroke, line end, gap and corner values, the `Enable…` attributes, `ApplyNextParagraphStyle`, `KeyboardShortcut`, `AnchoredObjectSetting` `PinPosition`, `AnchorPoint`, `VerticalAlignment`, `SpineRelative` and `HorizontalReferencePoint`, the frame fitting values, the story frame type, orientation and direction, the text wrap mode, contour type and `ApplyToMasterPageOnly`, the column, inset, footnote, auto-sizing and column rule values of `TextFramePreference`, the baseline frame grid values before DOM 18 and `EnableTransparency` before DOM 20. Most of these vary in a few styles of a few documents. The converter reads many of them from the INDD. `TextFramePreference` `FootnotesEnableOverrides="false"` is on every object style, `[None]` included, from InDesign 13.1: 1,392 of 1,392 styles of the trustworthy pairs before 2026-10, none before. The value file has it only from DOM 14, since it is generated per major version; the converter writes it when the header or the last session of the save history is 13.1 or later (`objects.md`, save history). ## Inks Every `Ink` element of the 1,495 corpus IDML files has `InkType="Normal"` and `PrintInk="true"` (`element_values.xml`); the converter writes them on every ink. `ConvertToProcess` is read from the INDD (`objects.md`, inks). ## Values of other elements `tools/element_values.py` collects the values of other elements the converter writes from every distinct IDML of the corpus (1,495 files exported by InDesign; IDML files that lack the `product` of InDesign's `aid` processing instruction were written by a script, so they are left out). A value is kept for a range of DOM versions when every element of that kind in every file of those versions has it with the same value, and it is left out entirely when some version shows two values or elements without it. The values may differ between versions; each range needs 10 files. The kept values are in `src/idml/element_values.xml`, in blocks with `MinimumVersion` and `MaximumVersion`. The converter writes them after the values it reads from the INDD and only where the element lacks the attribute. Files per DOM version: 7: 11, 8: 92, 9: 15, 10: 123, 11: 70, 12: 112, 13: 329, 14: 81, 15: 87, 16: 42, 17: 266, 18: 100, 19: 71, 20: 33, 21: 63. ### Page items | Element | Elements | Values kept | |---|---|---| | `TextFrame` | 58,521 | `LocalDisplaySetting="Default"`, `GradientStrokeAngle`, `GradientStrokeHiliteLength`, `GradientStrokeHiliteAngle` (`0`) from DOM 7; `FlexItemWidthMode` and `FlexItemHeightMode` (`FlexFixed`) from 21 | | `Rectangle` | 53,413 | `LocalDisplaySetting`, `StoryTitle="$ID/"` from 7 | | `Oval` | 10,851 | `LocalDisplaySetting`, `StoryTitle` and the gradient highlights from 8 | | `Polygon` | 134,377 | `LocalDisplaySetting`, `StoryTitle` and the fill gradient highlights from 7 | | `GraphicLine` | 19,100 | `LocalDisplaySetting`, `LockState="None"` and the gradient highlights from 7 | | `Group` | 15,857 | `LocalDisplaySetting` and the gradient highlights from 7 | Each kind also has `FlexItemWidthMode` and `FlexItemHeightMode` from DOM 21, and values that the converter reads from the INDD (`objects.md`, page item settings), which take precedence. Values such as `Name`, `Visible`, `Locked`, the gradient angles and the layout constraints are not kept: they differ in some elements, and those values come from the INDD. ### Form fields `tools/element_values.py` also collects the form field elements (`objects.md`, form fields). Only buttons and multi-state objects are in enough files (282 buttons in 43 files, 18 multi-state objects in 16): | Element | Values kept | Versions | |---|---|---| | `Button` | `VisibilityInPdf="VisibleInPdf"`, `HiddenUntilTriggered="false"` and the page item values below | DOM 8 to 18 (42 files) | | `Button` | `PrintableInPdf="true"` | DOM 10 to 18 (37 files) | | `Button` | `AppliedObjectStyle="ObjectStyle/$ID/[None]"` | DOM 8 to 13 (40 files) | | `MultiStateObject` | `InitiallyHidden="false"` and page item values | DOM 8 to 13 (16 files) | The page item values are those also kept for groups: the gradient angles and highlights, `LocalDisplaySetting`, the layout constraints and the change counts, which the converter writes from the INDD where it has them. `Button` `PrintableInPdf` is on 216 of 284 buttons: the later versions have too few files to keep it. Check boxes, radio buttons, text boxes, combo boxes and signature fields are in fewer than 10 files each, so no value of theirs is kept; their constant flags (`ReadOnly`, `PrintableInPdf` and others) are not written. ### Spreads, pages and layers | Element | Elements | Values kept | |---|---|---| | `Spread` | 11,914 | `PageTransitionDuration="Medium"`, `ShowMasterItems="true"`, `FlattenerOverride="Default"` from DOM 7; `SpreadHidden="false"` from 19 | | `FlattenerPreference` of spreads | 11,914 | `ClipComplexRegions`, `ConvertAllStrokesToOutlines`, `ConvertAllTextToOutlines` (`false`) and `RasterVectorBalance` 50 from DOM 7 | | `MasterSpread` | 2,461 | `OverriddenPageItemProps=""` from DOM 7 | | `Page` | 22,877 | `AppliedTrapPreset="TrapPreset/$ID/kDefaultTrapStyleName"` from DOM 7; `SnapshotBlendingMode="IgnoreLayoutSnapshots"`, `OptionalPage="false"` from 8 | | `Layer` | 2,723 | `ShowGuides` and `Expendable` (`true`) from DOM 7 | `Spread` `PageTransitionType` and `PageTransitionDirection` are left out: spreads of 6 files have page transitions. `Page` `GridStartingPoint` is left out: 28 pages of DOM 18 files have another value than `TopOutside`. Every spread has one `FlattenerPreference`, so the converter writes the element on every spread. For a spread without the flattener chunk (`objects.md`, flattener settings) it also writes `LineArtAndTextResolution` and `GradientAndMeshResolution`: 300 and 150, or 400 and 400 after a session of a Japanese or Chinese edition, by the language code of the save history (`objects.md`, flattener settings). The other attributes of these elements come from the INDD (`objects.md`). ### Document settings in designmap.xml These elements are children of `Document` without `Self`. The tool keeps an element when every IDML from some version on has it, with the values all of them have; the converter writes it in the order of the IDML files. | Element | From DOM | Values kept | Left out (vary) | |---|---|---|---| | `ConditionalTextPreference` | 7 | both | | | `EndnoteOption` | 13 | 9 | `EndnoteTitle`, `EndnoteSeparatorText`, the marker and text styles, `EndnoteMarkerPositioning` | | `TextFrameFootnoteOptionsObject` | 12 | all 4 | | | `LinkedStoryOption` | 8 | all 4 | | | `LinkedPageItemOption` | 8 | all 7 | | | `WatermarkPreference` | 7 | 11 | the font family, style and colour | | `TaggedPDFPreference` | 8 | (element only) | `StructureOrder` | | `AdjustLayoutPreference` | 14 | 4 | `EnableAdjustLayout`, `EnableAutoAdjustMargins`, `AllowFontSizeAndLeadingAdjustment` | | `HTMLFXLExportPreference` | 11 | both | | | `PublishExportPreference` | 11 | 7 | `ImageExportResolution`, `PublishPdf`, `CoverPage`, `PublishDescription`, `PublishFileName`, `PublishPageRange` | The values left out here are read from the INDD or follow the rules of `preferences.md` (adjust layout; EPUB, HTML and Publish Online export; values of the exporting edition). The converter writes `TaggedPDFPreference` from version 7.5, which the per-major blocks cannot express. ### DOM version Every package part has the `DOMVersion` of the application that exported the IDML. Several releases keep the DOM version of an earlier one. Over the 1,460 pairs of the corpus of 2026-10, by the version of the last session of the save history (`objects.md`, save history): | Last session | `DOMVersion` | |---|---| | 9.1, 9.2 | 9.0 | | 11.1 | 11.0 | | 14.3 | 14.0 | | 16.3, 16.4 | 16.2 | | 17.1 to 17.4 | 17.0 | | 18.1 to 18.4 | 18.0 | | 19.1, 19.2 | 19.0 | | 20.1 | 20.0 | | 20.5 | 20.4 | | every other version seen (7.0 to 21.5) | its major and minor version | The converter writes this value, from the header's version where the save history has none or names another major version. It matches 1,164 of 1,251 trustworthy pairs (the header's major version with `.0` matched 668). The others are files that a later release exported without saving them: 68 sessions of 10.1 whose IDML has 10.2, 16 of 13.0 with 13.1, and 3 more; their IDML `product` names that release. This is a rule of the exporting edition (`preferences.md`, values of the exporting edition). `Assignment` (1,495 elements, one per file): `UserName="$ID/"`, `ExportOptions="AssignedSpreads"`, `IncludeLinksWhenPackage="true"`, `FilePath="$ID/"` and `FrameColor` `Nothing`, from DOM 7. ### Language quotes A language object stores its quotes only in chunk 0x2D26, which 34 language objects of the corpus have (`objects.md`, languages); the converter writes those. For the others, for each language name the tool keeps `SingleQuotes` and `DoubleQuotes` when every IDML `Language` element with that name has the same value, in at least 3 files (a `Keyed` block of the values file): 64 languages. The tool reads IDML only, so it counts the languages with chunk 0x2D26 too. Over the 3,139 IDML languages of the corpus pairs matched with an INDD language, only Romanian differs because of them: every Romanian language without the chunk has `‚’` and `„”` (21 of 21 files), and the one with the chunk `‘’` and `»«`. The converter writes these quotes for Romanian without the chunk. Three names differ without the chunk too, and are left out: Japanese (`''` and `""` in 162 files, `‘’` and `“”` in 2), Korean (457 and 7) and Hebrew (74 and 1, `‘’` and `“”`). No stored value tells the exceptions apart; they follow the dictionary preferences of the exporting computer. Pasted smooth shades are keyed the same way by `ContentsType`: every `ConstantShade` (and every `AxialShade`) has `ContentsVersion="0"`, `SpotColorList=""`, `ContentsEncoding="Ascii64Encoding"`, `ColorEditable` and `ColorRemovable` `true`, `Visible="false"` and `SwatchCreatorID="7937"`; constant shades also `ContentsMatrix="1 0 0 1 0 0"`. ### Paragraph and character styles The styles other than the root styles (8,603 paragraph and 3,666 character styles in the IDML files): | Values kept | Styles | Versions | |---|---|---| | `PreviewColor` `Nothing` | paragraph | 7 on | | `EpubAriaRole=""` | paragraph and character | 21 on | | `IncludeClass="true"` | character | 13 on | Character styles' `PreviewColor` and `ExtendedKeyboardShortcut` are left out: a few styles have other values. TOC styles (1,556 in the IDML files) all have `RunIn="false"` and `IncludeHidden="false"`. `SetStoryDirection` varies and is read from the INDD (`objects.md`, table of contents styles). `SplitDocument`, `EmitCss` and `IncludeClass` are read from the INDD (`objects.md`, styles, export settings), as are keyboard shortcuts. Table and cell styles other than the root styles (1,587 and 138 in the IDML files) all have `KeyboardShortcut="0 0"` (cell styles from DOM 8, with 44 files) and `ExtendedKeyboardShortcut="0 0 0"` from DOM 15; the converter writes them. `Imported` and `StyleUniqueId` come from the INDD (`objects.md`, styles). ### Swatches | Element | Values kept | |---|---| | `Tint` | `ColorOverride="Normal"`, `ColorEditable`, `ColorRemovable`, `Visible` (`true`) and `SwatchCreatorID="7937"` from DOM 8 (270 tints in 142 files); `ConvertToHsb="false"` from 17 | | `Gradient` | `SwatchCreatorID="7937"` (and the flags, which the converter reads) from DOM 7 | | `Swatch` (`None`) | `SwatchCreatorID="7937"` from DOM 7 | Colours have their creator and alternate colour in the INDD (`objects.md`, colours). ### Trap presets All 1,495 IDML files list the same two `TrapPreset` elements at the end of `Resources/Styles.xml`, with the same attributes: `$ID/k[No Trap Preset]` and `$ID/kDefaultTrapStyleName`, which every page applies. The converter writes the list (a `List` block of the values file). One DOM 12 file has a third preset, named by the user, after these two. Its INDD object is not decoded, so the tool leaves presets with such names out of the list. ### Values written where the converter writes the attribute Some attributes are only on some elements; IDML writes them where a frame's value differs from its object style, or from some version on. For these the tool keeps a value when every element that has the attribute has the same value (`WhenWritten` blocks). The converter writes them on every text frame of those versions: | `TextFramePreference` values | Versions | |---|---| | `VerticalThreshold="0"` | 7 on | | `UseFlexibleColumnWidth="false"` | 8 on | | `FootnotesEnableOverrides`, `FootnotesSpanAcrossColumns` (`false`) | 13 on | | `ColumnRuleOffset`, `ColumnRuleTopInset`, `ColumnRuleBottomInset` (`0`), `ColumnRuleInsetChainOverride="true"`, `ColumnRuleStrokeTint="100"`, `ColumnRuleStrokeType="StrokeStyle/$ID/Solid"`, `ColumnRuleOverprintOverride="false"` | 15 on | The first version of the column rule and footnote values is also where the converter starts writing the column rule and footnote values it reads from the INDD (`objects.md`, text frame preferences). ### Export options of page items `ObjectExportOption` of page items has different attributes in different INDD versions, major and minor. Over the 92,409 page items of the trustworthy pairs there is no exception inside a version: | INDD version | Attributes in addition to the common set | |---|---| | 7 | no `ObjectExportOption` | | 8.0–9.1 | `CustomImageConversion="false"`, `CustomImageSizeOption="SizeRelativeToPageWidth"` | | 9.2 | also `UseOriginalImage="false"` | | 10.0 | `EpubType="$ID/"`, `UseExistingImage="false"`, `CustomHeightType="DefaultHeight"`, `CustomHeight="$ID/"`, `CustomWidthType="DefaultWidth"`, `CustomWidth="$ID/"`, `CustomImageConversion="false"`, `UseOriginalImage="false"` | | 10.1 on | `EpubType="$ID/"`, `SizeType="DefaultSize"`, `CustomSize="$ID/"`, `PreserveAppearanceFromLayout="PreserveAppearanceDefault"` | | 21.0 on | also `EpubAriaRole="$ID/"` | | 21.1 on | also `EpubAriaLabel="$ID/"`, `EpubAriaLabelSourceType="AutomaticARIALabel"` | | 21.4 on | also `AIGeneratedAltText="false"`, `AltTextGenerationError="false"`, `AltTextCropSyncRect=""` | The common set, on every page item from version 8 on: the alternative text and tagging values (`AltTextSourceType`, `ActualTextSourceType`, `CustomAltText`, `CustomActualText`, `ApplyTagType`, read from chunk 0x1E206, `objects.md`), `ImageConversionType="JPEG"`, `ImageExportResolution="Ppi300"`, `GIFOptionsPalette="AdaptivePalette"`, `GIFOptionsInterlaced="true"`, `JPEGOptionsQuality="High"`, `JPEGOptionsFormat="BaselineEncoding"`, `ImageAlignment="AlignLeft"`, `ImageSpaceBefore="0"`, `ImageSpaceAfter="0"`, `UseImagePageBreak="false"`, `ImagePageBreak="PageBreakBefore"`, `CustomImageAlignment="false"`, `SpaceUnit="CssPixel"`, `CustomLayout="false"`, `CustomLayoutType="AlignmentAndSpacing"`, and the `Properties` children `AltMetadataProperty` and `ActualMetadataProperty` (their values also from chunk 0x1E206). These are constant in all 93,701 elements except `ImageAlignment` (3 `AlignCenter`) and `SizeType` (3 `RelativeToTextFlow`), all in the same 3 items of one document; their INDD field is not identified. Object styles, `[None]` included, have `ObjectExportOption` from version 9 on (all 1,882 styles of those versions), with the common set (all values constant) and the same version-dependent attributes, except that 21.4 on adds only `AltTextCropSyncRect=""`. The converter writes these values by the version of the INDD file. It leaves out `CustomImageSizeOption`: the IDML schema the output is checked against does not allow that value. The `ObjectExportOption` blocks of `element_values.xml`, `root_values.xml` and `object_style_values.xml` are not used, because those files are keyed by major version only. Every `InCopyExportOption` of a rectangle, oval or polygon in the 1,495 corpus IDML files has `IncludeGraphicProxies="true"` and `IncludeAllResources="false"` (`element_values.xml`; ovals from DOM 8, as no DOM 7 file has one). The converter writes the element with these values on those items. ## Column direction of pages The `ColumnDirection` of pages is read from the INDD (`objects.md`, page settings). The `MarginPreference` and `DocumentPreference` of the preferences keep the observed `Horizontal` of `preference_values.xml`; 8 of the 489 trustworthy pairs have `Vertical` there, and no field was found for it (chunk 0x555 does not separate them, and 4 of the 8 lack it). ## Font platform name `PlatformName` is `$ID/` in every `Font` element of the corpus IDML files except the fonts of missing-font records, where it is read from the record (`fonts.md`, missing fonts). The converter writes `$ID/` on every other font. ## Number format IDML writes a number as the shortest decimal that reads back as the same 64-bit value. When two decimals of that length read back as the value, IDML takes the one nearest to the exact binary value, and of two equally near (the value lies exactly halfway, common for values that came from 32-bit floats) the one with the even last digit. For example the value 150.235992431640625 is written `150.23599243164062`, not `…063`. Numbers below 1e-4 are written with an exponent of at least two digits (`5.684341886080802e-14`, `1.8897456811828306e-05`); all others are written in plain notation. Evidence: of the 1,090,937 distinct numeric texts in the attributes of the 654 reference IDMLs of the corpus before 2026-10, the rule gives the same text for all but 67, which are text values rather than numbers (`DOMVersion="17.0"`, `2.000`, `0.80`). Rust's shortest form differs on 7,674 of them: ties written with the odd digit, and small numbers written without an exponent. The INDD values of `GraphicBounds` (chunk 0x1633), written this way, equal the IDML text in all 1,774 PDF, 407 EPS and 233 SVG graphics of the 803 pairs (`compare.py` compares `Properties` children as text). The converter writes every number this way. ## Preferences `Resources/Preferences.xml` of every corpus IDML has the same 41 top-level elements (`TextDefault`, `TextPreference`, `TextFramePreference`, `PageItemDefault`, `StoryPreference`, `FootnoteOption`, `TransparencyDefaultContainerObject` and others; the DOM 7 file lacks `EPubFixedLayoutExportPreference`). `tools/root_values.py` collects each attribute, `Properties` child and child element attribute of these elements with the same rule as for the root styles, and writes the kept values to `src/idml/preference_values.xml`. The converter writes those elements in the order the IDML files list them. | Values | Same in | Written for version | |---|---|---| | 697 | 1,448 of 1,448 files | all | | 73, 2, 14, 1, 1, 21, 4, 10 | all files with DOM 8, 9, 10, 11, 12, 13, 14, 15 or later | from that version | | 2 | 254 of 254 files with DOM 18 or later | 18 and later | | 19 | 93 of 93 files with DOM 20 or later | 20 and later | For example `TextDefault` keeps 140 of its 318 values, `TransparencyDefaultContainerObject` all 352 and `TextFramePreference` 29 of 34. **Left out:** - 593 values that vary between files, for example the page size, margins and grid settings, 178 values of `TextDefault` (among them the font and point size, `KeepFirstLines` and `KeepLastLines`) and 30 of `TextPreference`. Most of these differ in one to five documents; the converter does not read them from the INDD yet. - The document setup the converter reads from the INDD (`objects.md`): `PageHeight`, `PageWidth`, `FacingPages`, `Intent`, `PageBinding` and the four bleed offsets of `DocumentPreference`. The converter writes its own values for these and adds the observed values for the rest of the element. `ViewPreference` gets `RulerOrigin="SpreadOrigin"`, as before. - `AppliedGridObjectStyle` of `PageItemDefault`, which names a style other than a root style that the package need not contain. - Values first seen in a later version that fewer than 10 files show. `tools/element_values.py` also collects the preference elements from all 1,495 IDML files of the corpus (paths `Preferences/` in `src/idml/element_values.xml`), with its rule for other elements (per range of DOM versions, at least 10 files). The converter adds these values where `preference_values.xml` has none. They add, for example, the auto-sizing values of `TextFramePreference` for DOM 8 to 11 and the values of `EPubExportPreference` and `HTMLExportPreference` that all files from DOM 8 on have. `PageItemDefault` values that name an object style are written only when the package has that style. **Values kept despite unpaired files.** `XMLImportPreference` `AllowTransform="false"`, `ImportCALSTables="true"`, `ImportTextIntoTables="true"`, `ImportToSelected="true"`, `RemoveUnmatchedExisting="false"`, `RepeatTextElements="true"` and `DictionaryPreference` `MergeUserDictionary="false"` are the same in all 803 pairs of the corpus before 2026-10. Two IDML files of the 827 it had differ: one in the six `XMLImportPreference` values, one in `MergeUserDictionary`. Both were exported by an earlier version than their INDD was last saved with, so they are not pairs, and the INDD cannot show which field holds the value. The values are probably stored in the INDD; with no pair that varies, the field cannot be found. `element_values.py` keeps these values from the other files (`LATER_INDD_EXCEPTIONS`). **DOM 7.** DOM 7 references have no `EPubExportPreference` and no `HTMLExportPreference` (7 of 7 trustworthy pairs before 2026-10; `preference_values.xml` lists their values from DOM 8). The converter leaves both out before DOM 8. Some of these values are probably stored in the INDD and may differ in a document outside the corpus; none of them is read from the INDD yet. A sample typeset vertically shows this: two 2-byte fields of the preferences object (chunks 0xCD14 and 0xCD1C) are 1 there and 0 in all 250 distinct little-endian public files, and the values most likely to depend on vertical layout (`StoryPreference` `StoryOrientation`, `DocumentPreference` and `MarginPreference` `ColumnDirection`) are written as observed. The print PDF cannot show which preference each field holds, so they are not mapped.