# Attribute lists Formatting is stored as attribute lists (on page items: chunk 0x6E03). Decoded in `src/model/attrs.rs`. The IDML name and value kind of each ID are in the tables of `src/idml/attrs.rs`; `src/idml/kind.rs` converts each kind. ## Layout u32 count, then for each attribute: | Size | Field | |---|---| | 4 | Attribute ID | | 2 | Payload size *n* | | *n* | Payload: u16 value count, then values | Each value: u32 type, u16 length, data. The first value is the attribute's value. Every attribute in the samples has a second value of type 0x6E63 with 6 bytes, meaning unknown. | Type | Length | Value | |---|---|---| | 0x6E68 | 8 | f64 | | 0x6E67 | 4 | i32 | | 0x6E65 | 2 | u16 (enumeration or boolean) | | 0x6E69 | 16 | two f64 (a point) | | 0x117 | 4 | UID reference | All 2,730 page item chunks of this layout in the samples parse exactly. ## Page item attributes The list holds the item's local values. IDML writes an attribute only when it differs from the applied object style, so a value can be in the INDD list and absent from the IDML. | ID | IDML attribute | Evidence: IDML value equals INDD value | |---|---|---| | 0x6E68 | `FillColor` (swatch UID) | 1,190 of 1,194 | | 0x6E69 | `FillTint`; 100 is written as −1 (strokes, below) | 10 of 10 | | 0x6E64 | `StrokeColor` (swatch UID) | 78 of 82 | | 0x6E65 | `StrokeWeight` | 171 of 171 | | 0x6E6D | `MiterLimit` | 778 of 778 | | 0x6E6F | `CornerOption`: code 0 = `None`, 0x5A15 = `RoundedCorner`, 0x5A16 = `InverseRoundedCorner` (the codes of object styles, `objects.md`) | 20 of 23 for 0x5A15; with all three codes, 1,876 of 1,995 rectangles, 504 of 520 text frames and 489 of 489 page item defaults in the trustworthy pairs | | 0x6E70 | `CornerRadius` | 213 of 216 | | 0x6E70, 0x6E94, 0x6E92, 0x6E93 | `TopLeftCornerRadius`, `TopRightCornerRadius`, `BottomLeftCornerRadius`, `BottomRightCornerRadius` (below) | each radius ID against its own corner 4,357, 3,200, 3,276 and 3,325 matches, no mismatch | | 0x6E6F, 0x6E91, 0x6E8F, 0x6E90 | `TopLeftCornerOption`, `TopRightCornerOption`, `BottomLeftCornerOption`, `BottomRightCornerOption`: codes as `CornerOption` (below) | 11,211 of 11,211 | | 0x551F | `GradientFillLength` | all non-zero (30) | | 0x5520 | `GradientFillStart` (point) | all non-zero (30) | | 0x5525 | `GradientStrokeLength` (assumed; equal to fill in all samples) | 30 | | 0x5526 | `GradientStrokeStart` (assumed; equal to fill in all samples) | 30 | | 0x6E6E | `StrokeType` (built-in stroke style code, below) | 51 of 51 items, 576 of 576 object styles; strokes, below | | 0x6E8C | `StrokeAlignment`: 0 `CenterAlignment`, 1 `InsideAlignment`, 2 `OutsideAlignment` (below) | strokes, below | | 0x6E66, 0x6E89, 0x6E8A, 0x6E6B, 0x6E6C, 0x6E71, 0x6E72, 0x6E84 | `StrokeTint`, `GapColor`, `GapTint`, `EndCap`, `EndJoin`, `LeftLineEnd`, `RightLineEnd`, `ArrowHeadAlignment` | strokes, below | **Corners.** The top-left corner uses the same IDs as `CornerRadius` and `CornerOption`. The corner option codes are 0 `None`, 0x5A15 `RoundedCorner`, 0x5A16 `InverseRoundedCorner`, 0x5A17 `InsetCorner`, 0x5A18 `BevelCorner` and 0x5A19 `FancyCorner`; all six occur. The evidence is over the items of the trustworthy pairs whose corners differ: each radius ID matches its own corner with no mismatch, and has 16 to 71 mismatches against each other corner. The same IDs hold the corners of object styles (chunk 0x1B92B) and of the page item defaults (class 0x6E07): all four radii and options of `PageItemDefault` match in 495 of 495 documents. **Values equal to the object style.** The style's effective value is its own (chunk 0x1B92B), else that of the style it is based on, else the root `[None]`'s (`StrokeWeight="0"`, `FillTint="-1"`, `FillColor` and `StrokeColor` `Swatch/None`, `CornerRadius="12"`, `MiterLimit="4"`, `CornerOption="None"`; `idml-values.md`). Over the 489 trustworthy pairs, the converter wrote these values where the IDML has none, and every one of them equals the style's effective value (the stroke attributes of the strokes section follow the same rule): | Attribute | Written though the IDML has none | IDML writes it though equal | |---|---:|---:| | `StrokeWeight` | 27,989 | 5 | | `CornerRadius` | 1,608 | 0 | | `FillTint` | 1,423 | 0 | | `StrokeColor` | 1,336 | 0 | | `FillColor` | 312 | 0 | | `CornerOption` | 87 | 0 | | `MiterLimit` | 26 | 0 | Of the `StrokeWeight` values, the 26,969 on items whose IDML element has the same tag all equal the style (17,501 text frames, 7,070 rectangles, 2,074 polygons, 297 ovals, 27 lines); the others are on shapes written as another element type. The converter leaves out these attributes, and the four corner radii and options, when they equal the style's value (numbers to 6 significant digits). ### Values an item does not store Every document has one object of class 0x6E02 (606 of 606 trustworthy pairs). Its chunk 0x6E03 is a page item attribute list with every fill, stroke and corner attribute. An item whose own list lacks an attribute has the value of this base list, not that of its object style. IDML writes that value where it differs from the object style's effective value, the same rule as for stored values. The base lists differ in two ways across the trustworthy pairs, always together: 481 documents have `StrokeWeight` 1 and corner radii 12, 125 have 0.2835 (0.1 mm) and 14.1732 (5 mm). All have `StrokeColor` and `FillColor` naming the swatch `None`, `FillTint` −1, `MiterLimit` 4, `StrokeType` `Solid`, `StrokeAlignment` `CenterAlignment` and corner options `None`. Evidence: an analysis script looked for the field that tells the documents whose items have `StrokeWeight="0.2834645669291339"` apart from those whose items have 1, by testing every f64 of every object whose class occurs once per document; the `StrokeWeight` of this list matched 294 of 295 documents. With the base values and the rule above, the converter reproduces, over the rectangles, ovals, polygons, graphic lines and text frames of the trustworthy pairs (anchored ones included), 8,790 of 8,794 `StrokeColor` values, 54,694 of 54,694 `StrokeWeight`, 42,500 of 42,504 `MiterLimit` and 6,383 of 6,387 `CornerRadius`, with no extra value. The values not reproduced in that measurement were on 4 text frames without an object style (no object style, below) and on items with stored values not decoded then (77 `FillTint`, stored fills such as pasted smooth shades, the stroke alignment `OutsideAlignment` and stroke types; strokes, below). Over all pairs, 4 extra `StrokeWeight` values appear, in 3 stale pairs. IDML writes no corner attributes for EPS text (class 0x660B; none in 634 elements), so only its other attributes follow the rule. **Groups.** A group (class 0x401) stores none of these attributes. IDML gives a group, for each fill, stroke and corner attribute, the value that all its child page items have in effect: the child's own value (stored, or from the base list where IDML writes it), else its object style's effective value; for a child group, its own shared value. EPS text has no corner values and is left out for the corners. IDML writes the shared value when it differs from the group's object style value, when the style turns the category off, when the group has no object style, and for a fill or stroke tint that every child stores as 100 (written −1, strokes below); it leaves the attribute out when the children differ. With this rule the converter reproduces, over the groups of the 1,251 trustworthy pairs of the corpus after 2026-10 (none was written before): | Attribute | Values | |---|---:| | `StrokeWeight` | 4,741 of 4,741 | | `FillColor` | 3,813 of 3,813 | | `MiterLimit` | 3,679 of 3,679 | | `StrokeColor` | 1,221 of 1,221 | | `CornerRadius`, `TopLeftCornerRadius` | 1,001 of 1,001 each | | `StrokeTint` | 664 of 664 | | `FillTint` | 417 of 417 | | `BottomLeftCornerRadius`, `BottomRightCornerRadius` | 326 of 326 each | | `TopRightCornerRadius` | 296 of 296 | | `GapColor` | 124 of 124 | | corner options (five attributes) | 92 to 103 of as many | | `LeftLineEnd`, `RightLineEnd`, `StrokeAlignment`, `EndJoin`, `EndCap` | 40, 9, 15, 7 and 2 of as many | | `StrokeType` | 15 of 17 | In 38 of the `FillTint` values every child stores a fill tint of 100, and the group has −1. The rule adds 27 extra values: 3 documents of versions 10 to 12 whose groups have `CornerRadius` and `TopLeftCornerRadius` but not the other three radii, although every child has all four, and one `StrokeType`. `OverprintFill` and `OverprintStroke` of groups follow the children with ink only (`objects.md`, page item overprint). **Category turned off.** When the item's object style (other than the root `[None]`) turns a category off, IDML writes the item's values of that category even where they equal the style's, stored or from the base list: | Category (ID in chunk 0x1B92E) | Attributes | |---|---| | `EnableFill` (0x1B933) | `FillColor`, `FillTint` | | `EnableStroke` (0x1B934) | `StrokeColor` | | `EnableStrokeAndCornerOptions` (0x1B935, 0x1B936) | `StrokeWeight`, `MiterLimit`, `StrokeAlignment`, `EndCap`, `EndJoin`, the line ends, `ArrowHeadAlignment`, corner radii and options | | both stroke categories | `StrokeType`, `StrokeTint`, `GapColor`, `GapTint` | Evidence over the trustworthy pairs: 66 items with styles that turn all three off and 2 with styles that turn off both stroke categories have every such attribute in IDML, with the stored or base value; 12 items whose style turns off only `EnableStrokeAndCornerOptions` have `StrokeWeight`, `MiterLimit`, `StrokeAlignment` and the corners, but no `StrokeColor` or `StrokeType`. No sample turns off only `EnableStroke`, so the category of `StrokeType` is not shown; the converter writes it when both are off. The 68 items also have `StrokeTint`, `EndCap`, `EndJoin`, `GapColor` and `GapTint`, and the 12 items `EndCap` and `EndJoin` but not `StrokeTint`, `GapColor` or `GapTint`, which gives the categories of the table (the line ends and `ArrowHeadAlignment` follow the 12 items in the corpus after 2026-10). **No object style.** An item without an applied object style has `AppliedObjectStyle="n"` in IDML, and IDML writes every fill, stroke and corner attribute of the item, stored or from the base list: 60 of 60 such items (35 rectangles, 25 text frames) of the trustworthy pairs after 2026-10, every attribute equal. Transparency attributes (IDs 0x108xx and 0x1EBxx) are described in `transparency.md`. ### Strokes Object styles (class 0x1B901) hold a full page item attribute list in chunk 0x1B92B, with the same record layout as chunk 0x6E03 but a u16 count. The converter reads them (`object style` lists in `indd audit`) and writes them with the page item table. They also add evidence: the corpus pairs have 576 object styles whose name matches an IDML `ObjectStyle`. **Stroke type (0x6E6E).** The record has three values: a reference (type 0x117), a code (type 0x6E64, 4 bytes) and the usual 6-byte value. With reference 0 the code names a built-in stroke style. IDML writes `StrokeType` as `StrokeStyle/$ID/`. | Code | IDML stroke style | Evidence | |---|---|---| | 0x5A29 | `Solid` | 574 object styles in 135 files | | 0x5A2A | `Dashed` | 2,326 items in 40 files | | 0x5A37 | `Canned Dashed 4x4` | 52 items in 20 files | | 0x5A38 | `Canned Dashed 3x2` | 160 items in 45 files | | 0x5A39 | `Canned Dotted` | 345 items in 30 files | | 0x5A3A | `Wavy` | 4 items in 1 file | | 0x5A3B | `Straight Hash` | 109 items in 43 files | | 0x5A3C | `Right Slant Hash` | 2 items in 1 file | | 0x5A3E | `White Diamond` | 33 items in 18 files | | 0x5A3F | `Japanese Dots` | 913 items in 29 files; cells (`tables.md`) | | 0xB004 | `ThinThin` | 31 items in 3 files | | 0xB005 | `ThinThick` | 10 items in 3 files | | 0xB006 | `ThickThin` | 44 items in 12 files; object styles | | 0xB007 | `ThickThick` | 35 items in 13 files; cells | | 0xB009 | `ThickThinThick` | 1 item | | 0xB01A | `Triple_Stroke` | 5 items in 3 files | Item counts are over the items of the 1,251 trustworthy pairs of the corpus after 2026-10 with reference 0, IDML value against stored code. No code has two IDML values. The codes are the same in every list that names a stroke style (items, object styles, cells, tables, paragraph rules, underlines and strike-throughs), so the converter uses one table. Codes 0x5A3D and 0xB008 occur only in files without a trustworthy IDML. **Custom stroke styles.** A record with code 0x5A42 and a non-zero reference names a custom stroke style: the reference is the UID of an object of class 0xB016 (striped) or 0x5A43 (dashed). IDML defines each in `Resources/Graphic.xml` after the built-in `StrokeStyle` elements and names it in `StrokeType` as `StripedStrokeStyle/` or `DashedStrokeStyle/`. | Class | Chunk | Layout | IDML | |---|---|---|---| | 0xB016 | 0xB023 | flag byte, in-object string | `Name` | | 0xB016 | 0xB002 | u32 *n*, then *n* pairs of f64 (start, width), fractions of the stroke weight | `StripeArray`: each value × 100, in stored order | | 0x5A43 | 0x5A48 | flag byte, in-object string | `Name` | | 0x5A43 | 0x5A4F | u32 *n*, *n* f64 dash and gap lengths in points, u16 corner adjustment, u32 end cap | `DashArray` (the values as stored), `StrokeCornerAdjustment` (3 `DashesAndGaps`), `EndCap` (as 0x6E6B) | IDML writes `Self`, `StripeArray`, `Name` for a striped style and `Self`, `DashArray`, `StrokeCornerAdjustment`, `EndCap`, `Name` for a dashed one; `Self` is `StripedStrokeStyle/` or `DashedStrokeStyle/`. Every object of the two classes is written, also where nothing uses it, in UID order. Evidence over the trustworthy pairs of the corpus after 2026-10: the converter reproduces 21 of 21 `StripedStrokeStyle` elements (one definition, (0, 0.05), IDML `StripeArray="0 5"`) and 5 of 5 `DashedStrokeStyle` elements (one definition: 24, 0, corner 3, cap 1, IDML `DashArray="24 0"`, `StrokeCornerAdjustment="DashesAndGaps"`, `EndCap="RoundEndCap"`), with every attribute, and with them 21 more object style `StrokeType` values, 5 graphic line values and one of `PageItemDefault`. Four of the five files with the dashed style use it nowhere. A chunk whose length is not that of its layout leaves the style out with a warning. Over the 6,039 distinct INDD and INDT files of the corpus outside the privately held samples, 78 files have such styles: all 98 striped and 119 dashed objects have the chunk lengths of the layout (4 + 16*n* and 4 + 8*n* + 6, either byte order). Striped styles have 1, 2, 3, 5, 6 or 7 stripes; every dashed style has an even number of lengths, corner code 3 (117) or 0 (2) and end cap 0 (108) or 1 (11). Only the one-stripe form and corner code 3 occur in a pair. **Stroke alignment (0x6E8C).** 0 `CenterAlignment`, 1 `InsideAlignment` (3,238 items in 44 files), 2 `OutsideAlignment` (196 items in 50 files). Object styles use 0x1BB8 with the same codes. **More stroke attributes.** Object styles other than `[None]` were matched by `Self` with their IDML style over the 1,251 trustworthy pairs (4,305 styles; every one has the IDs below except 0x6E84), and page items by `Self` with their IDML element: | ID | IDML attribute | Values | Object styles | |---|---|---|---| | 0x6E66 | `StrokeTint` | f64; 100 is written as −1 | 4,305 of 4,305; 69 styles other than −1 (25 to 75) | | 0x6E89 | `GapColor` | swatch UID | 4,305 of 4,305; 23 `Color/Paper` | | 0x6E8A | `GapTint` | f64; 100 is written as −1 | two styles with 100 have −1 in IDML | | 0x6E6B | `EndCap` | 0 `ButtEndCap`, 1 `RoundEndCap`, 2 `ProjectingEndCap` | 4,305 of 4,305; 3 round | | 0x6E6C | `EndJoin` | 0 `MiterEndJoin`, 1 `RoundEndJoin`, 2 `BevelEndJoin` | 4,305 of 4,305; 5 round | | 0x6E71 | `LeftLineEnd` | line end code (below) | 4,305 of 4,305; 9 other than `None` | | 0x6E72 | `RightLineEnd` | line end code | 4,305 of 4,305; 14 other than `None` | | 0x6E84 | `ArrowHeadAlignment` | 0 `InsidePath`, 1 `OutsidePath` | in the 3,349 styles whose IDML has the attribute, 3,349 of 3,349; one style has 1 | Line end codes (u32): 0 `None`, 0x5A03 `SimpleArrowHead` (1 item), 0x5A04 `SimpleWideArrowHead` (110 items in 5 files), 0x5A05 `TriangleArrowHead` (44 in 8), 0x5A06 `TriangleWideArrowHead`, 0x5A07 `BarbedArrowHead` (64 in 6), 0x5A08 `CurvedArrowHead`, 0x5A09 `CircleArrowHead`, 0x5A0A `CircleSolidArrowHead`, 0x5A0C `SquareSolidArrowHead` (the page item defaults of 5 files) and 0x5A0D `BarArrowHead`. `EndCap` 2 occurs on 4 items in 3 files, `EndJoin` 2 on 6 items in 1 file. One line in the pairs before 2026-10 has 0x6E71 = 0x5A08 and 0x6E72 = 0x5A0D with `LeftLineEnd="CurvedArrowHead"` and `RightLineEnd="BarArrowHead"`, which gives 0x5A0D. `ArrowHeadAlignment` code 1 rests on one style; that style's line ends are mapped by their own IDs, which leaves 0x6E84. **On page items** IDML writes these attributes, the stroke type and the stroke alignment by the rule of the fill and stroke colour: the value the item stores (or that of the base list) where it differs from the applied object style's effective value, every value where the style turns the category off, and every value without a style (above). Groups never store these IDs; their values come from their children (values an item does not store, groups). A tint is compared as stored: an item that stores 100 differs from a style value of −1, and IDML writes it as −1. 4,508 items in 192 files store `StrokeTint` 100 and have `StrokeTint="-1"` in IDML; no IDML item has `StrokeTint="100"`. Items that store −1, or the style's own value, have no attribute; one polygon that stores 100 has none either. `GapTint` behaves the same (11 items in 2 files store 100), and so does `FillTint`: over the 1,252 trustworthy pairs, 90 items in 10 files store a fill tint of 100 (79 polygons, 76 of them in groups, and 11 text frames), and all 90 have `FillTint="-1"` in IDML. No page item or object style in these IDMLs has `FillTint="100"`. Over the rectangles, ovals, polygons, graphic lines and text frames of the trustworthy pairs the converter reproduces these values; the one extra value is a polygon of a 19.0 document that stores `StrokeTint` 100 and has no attribute in IDML: | Attribute | Values | |---|---:| | `StrokeTint` | 5,874 of 5,874 | | `EndCap` | 8,538 of 8,538 | | `EndJoin` | 8,220 of 8,220 | | `GapColor` | 435 of 435 | | `GapTint` | 181 of 181 | | `LeftLineEnd` | 453 of 453 | | `RightLineEnd` | 435 of 435 | | `ArrowHeadAlignment` | 139 of 139 | | `StrokeType` | 4,230 of 4,235 (5 custom stroke styles) | | `StrokeAlignment` | 3,572 of 3,572 | The values of `PageItemDefault` come from the page item defaults list with the same table (`preferences.md`). **Not identified.** `LeftArrowHeadScale` and `RightArrowHeadScale` are 100 in every object style and in every `PageItemDefault` but one, as are 0x6E95 and 0x6E96, so the two IDs cannot be told apart. The converter writes them from the value files on object styles, and from the IDs in their order on `PageItemDefault`, where the values are equal in every file (`preferences.md`, page item defaults). 0x6E78 is 0 in every style, 0x6E73 in all but 2; 0x6E78 is `Nonprinting` of `PageItemDefault`. 0x6E67, 0x6E6A and 0x6E8B are the overprint flags (`objects.md`, object style settings). **Dashes (0x5A35).** Page item attribute 0x5A35 (value type 0x5A36): 8 bytes (0 in every sample), u32 count *n*, *n* f64, u16 corner adjustment. IDML writes the f64 list as `StrokeDashAndGap` (numbers separated by spaces) and the code as `StrokeCornerAdjustment`: 3 `DashesAndGaps`, the only code in the samples, which the dashed stroke style's chunk 0x5A4F also stores for `DashesAndGaps`. IDML has the two attributes only on items whose stroke type in effect is `Dashed` (`StrokeStyle/$ID/Dashed`; every such item has them), and on groups by the group rule: no item with another stroke type has them, though 1,251 documents store the attribute in the page item defaults and many items of other stroke types store it. Object styles never have them. With these rules the converter reproduces 2,312 of 2,316 graphic line values of each attribute, 8 of 9 for polygons and 10 of 24 and 29 for groups, with no extra value. The values of 9 rectangles and 7 EPS text items are not written (not examined). ## Text attribute lists Text formatting uses the same record layout with a **u16** count. It appears in: - **Styles:** chunk 0x23F of a style object (u16 count, records). - **Story runs:** each paragraph-style or character-style run record (chunk 0x262, see `objects.md`) continues after the style UID with a u16 count and records. These are the run's local overrides. Value types are specific to each attribute (for example 0x1B05 for swatch references, 0x1B28 for point size), so the converter decodes a text value by the layout of its attribute. The attributes whose value is not a single number have their layout listed in `text_layout` (`src/model/attrs.rs`): string values (`FontStyle`) are a flag byte followed by an in-object string; tab lists, nested, GREP and line styles, bullet characters, points, ruby text and cell edge stroke types are described below. Other values are numbers, decoded by length: 8 bytes f64, 4 bytes u32, 2 bytes u16. A value that does not fit its attribute's layout is kept undecoded and not written. ### Mapping and evidence Found by aligning IDML `ParagraphStyleRange` and `CharacterStyleRange` elements with INDD runs at the same text offset (1,562 ranges), and IDML styles with INDD styles of the same name (486 styles), then checked with `tools/compare.py`. Counts are matches/total in the current comparison (ranges + paragraph styles + character styles where present). | ID | IDML attribute | Encoding | Evidence | |---|---|---|---| | 0x1B01 | `FillColor` | swatch UID | 354/354 ranges, 226/226 styles | | 0x1B02 | `FontStyle` | flag + string; empty: below | 474/474 ranges, 202/202 styles | | 0x1B03 | `PointSize` | f64 | 1,090/1,090 ranges | | 0x1B06 | `HorizontalScale` | fraction ×100 | 2/2 ranges, 75/75 styles | | 0x1B07 | `KerningMethod` | code: 15972 Metrics, 79875 Optical, 0x3E65 `$ID/Metrics - Roman Only` (27 styles), 0x221 `$ID/None` (21 styles and text defaults); a run with manual kerning: below | 20/20 ranges, 81/81 styles | | 0x1B08 | `Ligatures` | 1 = true | 201/201 ranges | | 0x1B0A | `StrokeWeight` | f64 | 241/241 ranges | | 0x1B0B | `Tracking` | thousandths of an em ×1000 | 49/49 ranges, 94/94 styles | | 0x1B0C | `Composer` | code: 0x2001 HL Single, 0x2002 HL Composer, 0x2078 HL Composer Optyca, 0x2079 HL Single Optyca, 0x2010 HL Composer J, 0x2011 HL Single J (runs/styles: 47/14, 0/30, 7/15) | 4/4 ranges, 114/114 styles | | 0x1B0D | `DropCapCharacters` | u16 | 79/79 styles; see below | | 0x1B0E | `DropCapLines` | u16 | 79/79 styles; see below | | 0x1B10 | `BaselineShift` | f64 | 79/79 styles; see below | | 0x1B11 | `Capitalization` | 0 Normal, 1 SmallCaps, 2 AllCaps, 3 CapToSmallCap, 4 LowerCase | 77/77 styles; codes 1, 3 and 4 below | | 0x1B12 | `StrokeColor` | swatch UID | 75/75 styles | | 0x1B13 | `KerningValue` | ems ×1000, rounded to a whole number; 1e8 = none, written `1e+11` | 1e8: 144/144 ranges, 29/29 character styles; rounding and scale (from a sample and its print PDF): below | | 0x1B15 | `VerticalScale` | fraction ×100 | 79/79 styles; see below | | 0x1B16 | `LeftIndent` | f64 | 22/22 ranges | | 0x1B17 | `RightIndent` | f64 | 86/86 styles | | 0x1B18 | `FirstLineIndent` | f64 | 12/12 ranges | | 0x1B1A | `AutoLeading` | fraction ×100 | 75/75 styles | | 0x1B1B | `Leading` (Properties) | f64; negative = Auto | 43/43 ranges, 137/137 styles | | 0x1B1D | `AppliedLanguage` | language UID; see below | 88/88 styles | | 0x1B1F | `Hyphenation` | 3 = true, 0 = false | 204/204 ranges | | 0x1B24 | `NoBreak` | 1 = true | 78/78 styles | | 0x1B25 | `HyphenationZone` | f64 | 155/155 ranges | | 0x1B26 | `SpaceBefore` | f64 | 13/13 ranges | | 0x1B27 | `SpaceAfter` | f64 | 42/42 ranges | | 0x1B29 | `TabList` (Properties) | list; see below | 110/110 styles | | 0x1B2A | `Underline` | 1 = true | 81/81 ranges | | 0x1B2B | `AppliedFont` (Properties) | font family UID; a composite font as an object (below) | 1,071/1,071 ranges, 236/247 styles | | 0x1B2C | `OTFFigureStyle` | 0 TabularLining (23 runs, 5 styles), 1 ProportionalOldstyle, 2 ProportionalLining, 3 TabularOldstyle (below), 4 Default | 4/4 ranges, 138/138 styles | | 0x1B2E | `MaximumWordSpacing` | fraction ×100 | 77/77 styles | | 0x1B2F | `MinimumWordSpacing` | fraction ×100 | 77/77 styles | | 0x1B31 | `MaximumLetterSpacing` | fraction ×100 | 78/78 styles; see below | | 0x1B32 | `MinimumLetterSpacing` | fraction ×100 | 78/78 styles; see below | | 0x1B33 | `DesiredGlyphScaling` | fraction ×100 | 22/22 `ParagraphStyleRange` elements in 10 documents (96 to 99), 1,251 trustworthy pairs; 1.0 in the root style and text defaults of every pair, as IDML's 100. No other text list of the pairs has it | | 0x1B37 | `StartParagraph` | 0 Anywhere, 1 NextColumn, 2 NextPage, 3 NextFrame, 4 NextOddPage | 78/78 styles; codes below | | 0x1B3C | `Position` | 0 Normal, 1 Superscript, 2 Subscript, 3 OTSuperscript, 4 OTSubscript, 5 OTNumerator, 6 OTDenominator (runs/styles: 107/23, 4/7, 1/0, 3/0, 21/0 for 1, 2, 3, 4, 6) | 20/20 ranges, 78/78 styles | | 0x1B40 | `KeepLinesTogether` | 1 = true | 78/78 styles; see below | | 0x1B42 | `FillTint` | f64 | 24/24 ranges | | 0x1B46 | `GradientFillAngle` | f64 | 81/81 styles | | 0x1B48 | `GradientFillLength` | f64 | 2/2 ranges, 81/81 styles | | 0x1B4A | `GradientFillStart` | two f64 | 2/2 ranges, 81/81 styles | | 0x1B4D | `RuleAboveLineWeight` | f64 | 155/155 ranges | | 0x1B4F | `RuleAboveOffset` | f64 | 78/78 styles | | 0x1B50 | `RuleAboveLeftIndent` | f64 | 91/91 styles | | 0x1B51 | `RuleAboveRightIndent` | f64 | 91/91 styles | | 0x1B52 | `RuleAboveWidth` | 1 ColumnWidth, 2 TextWidth | 91/91 styles | | 0x1B53 | `RuleBelowColor` (Properties) | swatch UID, 0 = "Text Color" | 6/6 ranges, 114/114 styles | | 0x1B54 | `RuleBelowLineWeight` | f64 | 155/155 ranges | | 0x1B55 | `RuleBelowTint` | f64 | 78/78 styles | | 0x1B56 | `RuleBelowOffset` | f64 | 78/78 styles | | 0x1B5D | `RuleBelow` | 1 = true | 83/83 styles | | 0x1B6A | `ParagraphBreakType` | 0 Anywhere, 1 NextColumn, 2 NextPage, 3 NextFrame, 4 NextOddPage (below) | 11/11 ranges, 78/78 styles | | 0x1B6B | `SingleWordJustification` | 0 LeftAlign, 3 FullyJustified | 80/80 styles | | 0x1B75 | `AllNestedStyles` (Properties) | list; see below | 124/124 paragraph styles, 10/10 text defaults (1,251 trustworthy pairs) | | 0x1B7E | `Justification` | 0 LeftAlign, 1 CenterAlign, 2 RightAlign, 3 FullyJustified, 4 LeftJustified, 5 CenterJustified, 6 RightJustified, 8 ToBindingSide, 9 AwayFromBindingSide (runs/styles for 3, 6, 8, 9: 34/1, 3/0, 2/1, 7/14) | 70/70 ranges | | 0x1B80 | `DropcapDetail` | u32 | 5/5 ranges, 103/103 styles | | 0x1B8C | `OTFContextualAlternate` | 1 = true | 220/220 ranges | | 0x1B8D | `UnderlineColor` (Properties) | swatch UID, 0 = "Text Color" | 81/81 ranges | | 0x1B91 | `UnderlineOffset` | f64 | 81/81 ranges | | 0x1B94 | `UnderlineWeight` | f64 | 81/81 ranges | | 0x1BB7 | `MiterLimit` | f64 | 161/161 ranges | | 0x1BB9 | `EndJoin` | 0 MiterEndJoin, 1 RoundEndJoin | 86/86 styles | | 0x1BBA | `AllGREPStyles` (Properties) | list; see below | 152/152 paragraph styles, 48/48 text defaults, 320/320 text ranges (1,251 trustworthy pairs) | | 0x1BBB | `AllLineStyles` (Properties) | list; see below | 6/6 paragraph styles, 9/9 text ranges (1,251 trustworthy pairs) | | 0x1BBD | `SpanColumnType` | 0 SingleColumn, 1 SpanColumns, 2 SplitColumns (69 runs, 10 styles) | 41/41 ranges, 94/94 styles | | 0x1BBE | `SpanSplitColumnCount` (Properties) | u16; 1 = All | 12/12 ranges, 78/78 styles | | 0x1BBF | `SplitColumnInsideGutter` | f64 | 155/155 ranges | | 0x1BC4 | `SpanColumnMinSpaceAfter` | f64 | 84/84 styles | | 0x1BD2 | `ParagraphShadingColor` (Properties) | swatch UID | 25/25 ranges | | 0x1BD3 | `ParagraphShadingTint` | f64 | 79/79 styles | | 0x1BD6 | `ParagraphShadingOn` | 1 = true | 88/88 styles | | 0x1BDB | `ParagraphShadingTopOffset` | f64 | 91/91 styles | | 0x1BDC | `ParagraphShadingBottomOffset` | f64 | 90/90 styles | | 0x1BF6 | `ParagraphBorderColor` (Properties) | swatch UID | 72/72 styles | | 0x1BFA | `ParagraphBorderGapColor` (Properties) | swatch UID; 0 is `Swatch/None` | present exactly when the list has the ID (344 of 344 styles of the corpus of 2026-10); trustworthy pairs after 2026-10: 1,147 of 1,147 paragraph styles, 918 of 918 text defaults, 2,745 of 2,745 text ranges, 642 + 84 story and cell paragraph ranges | | 0x1BF9 | `ParagraphBorderOn` | 1 = true | 77/77 styles | | 0x1DF03 | `ParagraphBorderTopOffset` | f64 | 83/83 styles | | 0x1DF04 | `ParagraphBorderBottomOffset` | f64 | 83/83 styles | | 0x1DF21 | `SameParaStyleSpacing` (Properties) | f64; −1 = SetIgnore | 15/15 ranges, 100/100 styles | | 0x4265 | `GridAlignFirstLineOnly` | 1 = true | 81/81 styles | | 0x4266 | `GridAlignment` | 0 None, 1 AlignBaseline, 2 AlignEmTop, 3 AlignEmCenter (below) | 86/86 styles | | 0x425E | `Tatechuyoko` | 1 = true | root styles; 1 from a sample typeset vertically and its print PDF, see below | | 0x4279 | `ShataiDegreeAngle` | degrees, written ×100 | root styles; see below | | 0x427A | `ShataiAdjustTsume` | 1 = true | root styles; see below | | 0x427B | `ShataiAdjustRotation` | 1 = true | root styles and a sample typeset vertically; see below | | 0x422D | `RubyFlag` | number; written when not 0 | a sample and its print PDF; see below | | 0x422E | `RubyString` | u32 length, text segments; written when not empty | a sample and its print PDF; see below | | 0x1A401 | `BulletsAndNumberingListType` | 0 NoList, 1 BulletList, 2 NumberedList (8 runs, 20 styles) | 10/12 ranges | | 0x1A406 | `BulletChar` (Properties) | u32 type, u32 value; see below | 1/1 ranges, 84/84 styles | | 0x1A413 | `BulletsFont` (Properties) | font family UID, 0 = `$ID/` | 1/1 ranges, 84/84 styles | | 0x1A414 | `BulletsFontStyle` (Properties) | flag + string; empty = `Nothing` | 1/1 ranges, 84/84 styles | | 0x1A417 | `AppliedNumberingList` (Properties) | numbering list UID (`objects.md`, numbering lists) | 519/519 styles | | 0x1A419 | `NumberingContinue` | 1 = true | 2/2 ranges, 78/78 styles | | 0x1A41F | `BulletsCharacterStyle` (Properties) | character style UID | 84/84 styles | | 0x1A420 | `NumberingCharacterStyle` (Properties) | character style UID | 79/79 styles | | 0x1A423 | `NumberingExpression` | flag + string | 79/79 styles | | 0x4221 | `Mojikumi` (Properties) | table UID; see below | 936/1,019 styles | | 0x4224 | `KinsokuSet` (Properties) | table UID; see below | 987/987 styles | | 0x42C0 | `TreatIdeographicSpaceAsSpace` | 1 = true | 1,082/1,083 styles | | 0x50F18 | `DiacriticPosition` | 4 OpentypePosition, 5 OpentypePositionFromBaseline | 1,007/1,010 styles | The counts of these four are paragraph and character styles of all 654 pairs whose IDML style has the attribute and the same name (the root paragraph style of every pair among them); they also reproduce the `TextDefault` values of the preferences (`preferences.md`). The kinsoku and mojikumi value is a UID: 0 is `Nothing`; a built-in table (`kinsoku and mojikumi tables` in `objects.md`) is written as an enumeration: `kHardKinsokuName` `HardKinsoku` (51 styles), `kSoftKinsokuName` `SoftKinsoku` (11), `kKoreanKinsokuName` `KoreanKinsoku` (12), `kSimpChineseKinsokuName` `SimplifiedChineseKinsoku` (1), `kMojikumiDefaultName1` `LineEndAllOneHalfEmEnum` (55) and `kMojikumiDefaultName16` `SimpChineseDefault` (1); a custom kinsoku table (class 0x4204) as the object `KinsokuTable/` (38). The corpus of 2026-10 adds three built-in tables, each from the `TextDefault` of every trustworthy pair that uses it: `kTradChineseKinsokuName` `TraditionalChineseKinsoku` (90 documents), `kMojikumiDefaultName15` `TradChineseDefault` (80) and `kMojikumiDefaultName2` `OneEmIndentLineEndUkeOneHalfEmEnum` (1). One paragraph range of one trustworthy pair (12.1) applies `kMojikumiDefaultName14`, `LineEndPeriodOneEmEnum`; the base table of custom tables adds `kMojikumiDefaultName10`, `LineEndAllOneEmEnum` (`objects.md`, custom mojikumi tables). With them the converter reproduces 1,251 of 1,251 `TextDefault` `KinsokuSet` and `Mojikumi` values, 1,430 of 1,430 `KinsokuSet` and 1,532 of 1,532 `Mojikumi` values of paragraph styles. Other built-in tables are not written. A custom mojikumi table (class 0x4203, `objects.md`) is the object `MojikumiTable/`. In the trustworthy pairs of the corpus of 2026-10 this gives the IDML value for 576 paragraph ranges, 105 paragraph styles, 94 ranges in table cells, 19 story ranges and 11 text defaults, with no value that differs from the IDML. Paragraph style counts include the 78 root styles `[No paragraph style]` of the compared pairs. For many attributes above the root holds the only value that is the same in every pair (for example `RightIndent="0"`), so the evidence that tells the attribute apart is in the other styles: | Attribute | Other styles or ranges with the attribute | |---|---| | `BaselineShift` | 1 style with 0; 1 range with −16 in a story that `compare.py` does not align | | `RightIndent` | 8 styles in one pair, two different values | | `AppliedLanguage` | 10 styles in 2 pairs | | `NoBreak` | 1 style with 1 (true), in a pair whose IDML is from an older version | | `TabList` | 33 styles in 5 pairs, left and right stops, with and without leaders | | `OTFFigureStyle` | 60 styles in one pair (codes 1 and 2), 4 ranges in another (code 4) | | `Position` | 20 ranges in one pair | | `GradientFill*` | 3 styles in one pair (angle −90, distinct length and start), 2 ranges in another | | `RuleAbove*Indent`, `RuleAboveWidth` | 13 styles in one pair (left −1.13, right −2.83) | | `RuleBelowColor` | 36 styles in 2 pairs, four colours | | `RuleBelow` | 5 styles in one pair (4 true, 1 false) | | `ParagraphBreakType` | 19 ranges in one pair | | `SingleWordJustification` | 2 styles in one pair with code 0 | | `EndJoin` | 8 styles in one pair | | `SpanColumnType` | 16 styles in 2 pairs, 46 ranges in one pair, both codes | | `SpanSplitColumnCount` | 13 ranges in one pair with 2 | | `SpanColumnMinSpaceAfter` | 6 styles in one pair | | `ParagraphShading*`, `ParagraphBorder*` | 2 to 14 styles in one pair, several different offsets | | `SameParaStyleSpacing` | 30 styles and 16 ranges, each in 2 pairs | | `GridAlignment`, `GridAlignFirstLineOnly` | 8 and 3 styles in 3 pairs | | `BulletChar`, `BulletsFont`, `BulletsFontStyle`, `BulletsCharacterStyle` | 6 styles in 2 pairs, 1 range in another | | `NumberingExpression`, `NumberingCharacterStyle` | 1 style | | `NumberingContinue` | 2 ranges in 2 pairs | | `VerticalScale` | 1 style; see below | For codes, the root style gives the default: for example every root style has 0x1B6B = 3 and IDML `SingleWordJustification="FullyJustified"`, and 0x1BBE = 1 with `SpanSplitColumnCount` `All`. **`VerticalScale` (0x1B15).** The value is 1 (100%) wherever it appears, like `HorizontalScale` (0x1B06), so values do not tell them apart. Over all pairs, including those with an IDML from another version, 0x1B15 is present in exactly the 5 styles where IDML writes `VerticalScale`; 0x1B06 is identified separately by values other than 100%. **Languages (0x1B1D).** The value is the UID of an object of class 0x2D07. Its chunk 0x2D0F is a flag byte, then the language name as an in-object string (`English: USA`), followed by other strings. IDML writes `$ID/` and that name. Over all pairs, all 62 styles (root styles included) and 4 ranges with the attribute match, with 5 different languages. The language named `Neutral` is the one exception: IDML writes it as `[No Language]`, without `$ID/`, although its `Language` element is named `$ID/[No Language]` (`objects.md`, languages). Over the 1,252 trustworthy pairs of the corpus of 2026-10, `AppliedLanguage` has a value without `$ID/` 514 times in 43 documents, always `[No Language]`; no IDML writes `$ID/Neutral` or `$ID/[No Language]` there. Each of the 490 values matched with an INDD object (31 text defaults, 54 paragraph styles, 4 character styles, 401 ranges) names a language object called `Neutral`. **Tab stops (0x1B29).** u16 count, then for each stop: f64 position, u16 alignment (0 `LeftAlign`, 1 `CenterAlign`, 2 `RightAlign`, 3 `CharacterAlign`), for code 3 the alignment character as one UTF-16 unit, u16 leader length and the leader in UTF-16 code units. IDML writes each stop as a `ListItem` with `Alignment`, `AlignmentCharacter`, `Leader` and `Position`. A stop with another code stores no character, and IDML writes `.` for it in every sample. An empty list is a count of 0. The converter leaves out a list with another alignment code. Evidence, corpus of 2026-10, the stops of paragraph styles matched by `Self` and of story paragraph ranges: 79 `CenterAlign` stops (59 in trustworthy pairs) and 36 `CharacterAlign` stops (34; 27 with `,`, 9 with `.`) match in alignment, character, leader and position. Before code 3 was read, a list with such a stop was not read at all: the character made the record longer than the stops. **Nested styles (0x1B75).** u32 count, then for each item: u32 character style UID, u32 length, and a delimiter code of that many characters stored as text segments (see `objects.md`). The code gives `Delimiter`, `Inclusive` and `Repetition`: | Code | IDML | Evidence | |---|---|---| | `[c]` | `Delimiter` string *c*, `Inclusive="false"` | 13 items in one pair (`.` and `:`) | | `(c)` | `Delimiter` string *c*, `Inclusive="true"` | 1 item in the same pair | | `^c` | `Dropcap`, `Inclusive="true"` | 1 item in the same pair | | digits after `)` or `]` | `Repetition`; no digits = 1 | a sample and its print PDF | | `(^w)` | `AnyWord` | a sample and its print PDF | | `(^?)` | `AnyCharacter` | a sample and its print PDF | | `(^c)` | `Dropcap` | 54 `Dropcap` records in 18 pairs (with `^c`) | | `^t` in brackets | `Tabs` | 9 records in 5 pairs (`[^t]`, `(^t)`, `(^t)2` seen) | | `^L` in brackets | `Repeat`; the digits are `Repetition` | 41 records in 33 pairs (`(^L)2`, `(^L)4` seen) | | `^p` in brackets | `EndNestedStyle` | 2 records in 1 pair | | `^9` in brackets | `Digits` | 2 records in 1 pair (`[^9]`, `(^9)`) | | `^S` in brackets | `Sentence` | 1 record in 1 pair | The public pair has only single-character literals and repetition 1. The last three rows rest on a sample and its print PDF: - A sample has nested items with `(^?)` and with `(^w)` followed by a number, the latter applying a character style with Capitalization code 3 (`CapToSmallCap`, see below). - In the PDF, capitals in the first words of paragraphs in those styles are set at small-cap height (ink height measured on the rendered page), and capitals later in the same lines at full height. Small capitals reach at most one word past the number, and capitals a few words further on are at full height, so the number is the repetition. Of the counting delimiters in the IDML schema, only `AnyWord` gives a stretch of several words: `AnyCharacter`, `Letters` and `Digits` would end within the first word, and `Sentence` would run past sentence ends, which the PDF does not show. - A paragraph-level override has `(^?)` followed by a number *n*, then `(^w)`. In the PDF the first capital after the first *n* characters of that paragraph is a small cap. With a word delimiter, the first *n* words would keep full capitals, and that capital lies within them. So `^?` counts characters: `AnyCharacter`. The last six rows come from the 1,251 trustworthy pairs of the corpus after 2026-10. Their reference IDMLs have 265 nested style records in 72 documents (styles, text defaults and text ranges; the record counts above are these). The stored items map one to one onto the IDML records, with the bracket giving `Inclusive` and the digits `Repetition` as above: the converter reproduces `AllNestedStyles` of 124 of 124 paragraph styles, 10 of 10 text defaults, 274 of 274 text ranges and 7 of 7 `Story/ParagraphStyleRange` elements. Without these codes it reproduced 54 paragraph styles and 4 text defaults. The converter leaves out a list that has any other code. **GREP styles (0x1BBA).** The layout of nested styles, with a GREP expression in place of the delimiter code: u32 count, then for each item u32 character style UID (0 for none), u32 length *n* and the expression as *n* UTF-16 code units in text segments. IDML writes `AllGREPStyles` (`type="list"`), one `ListItem type="record"` per item in stored order, with `AppliedCharacterStyle` (object) and `GrepExpression` (string). Evidence over the 1,251 trustworthy pairs: every 0x1BBA value of the paragraph styles parses and ends exactly after its last item; the reference IDMLs have 494 such lists with 1,064 items, all of these two fields. The converter reproduces 152 of 152 paragraph styles, 48 of 48 text defaults, 320 of 320 text ranges (`Story/ParagraphStyleRange` 110, `Cell/ParagraphStyleRange` 94), with no extra value. A style's list is its own value; IDML does not write an inherited one. **Line styles (0x1BBB).** u32 count, then 8 bytes per item: u32 character style UID and u32 line count. IDML writes `AllLineStyles`, one record per item: `AppliedCharacterStyle` (object), `LineCount` (long) and `RepeatLast` (long). `RepeatLast` is −1 in all 9 list items of the trustworthy reference IDMLs (3 documents) and no stored field holds it, so the converter writes −1. Reproduced: 6 of 6 paragraph styles and 9 of 9 text ranges. **Empty lists.** None of the 1,251 trustworthy reference IDMLs has an `AllGREPStyles`, `AllLineStyles` or `AllNestedStyles` element without items. The converter writes no GREP or line style list without items. **Bullet character (0x1A406).** Two u32: the character type (0 `UnicodeOnly`, 1 `UnicodeWithFont`, 2 `GlyphWithFont`) and the character value. IDML writes an empty `BulletChar` element with `BulletCharacterType` and `BulletCharacterValue` attributes. In the pairs, type 0 occurs in the root styles with value 8226 (•), type 1 in one range and type 2 in 6 styles, each with its value and a `BulletsFont`. **`KeepLinesTogether` (0x1B40).** In every pair the root style has 0, and IDML writes `KeepLinesTogether="false"`, `KeepAllLinesTogether="false"`, `KeepFirstLines="2"` and `KeepLastLines="2"`. No public style sets another value. The rest of the evidence is a sample and its print PDF: - Some paragraph styles of the sample set 0x1B40 to 1. Apart from attributes identified above, they set nothing else. - In the PDF, paragraphs in those styles that break across pages always leave at least two lines on each page. A paragraph of *n* lines broken at a random line leaves a single line on one side with probability 2/(*n* − 1), so this pattern does not come from chance. - Such paragraphs do break across pages, so the attribute is not `KeepAllLinesTogether`. A boolean that keeps at least two lines together at both ends, with the defaults above, is `KeepLinesTogether`. **`Capitalization` codes 1, 3 and 4.** - Code 1 is `SmallCaps`. In one pair (InDesign 20.2, IDML from the same version), 8 of 8 paragraph styles with code 1 have `Capitalization="SmallCaps"`. The same pair confirms code 2 (4 of 4 styles) and code 0. - Code 3 is `CapToSmallCap`. No pair has it. The evidence is a sample and its print PDF. In the PDF, words set in a style with code 3 that start with a capital letter show that capital at the height of the letter after it, about two thirds of the height of word-initial capitals in ordinary text. Heights were measured on the rendered pages. So code 3 sets capital letters as small capitals too. Of the values in the IDML schema (`Normal`, `SmallCaps`, `AllCaps`, `CapToSmallCap`, `LowerCase`), only `CapToSmallCap` does that. - Code 4 is `LowerCase`. No pair of the corpus has it: no IDML outside the privately held samples has `LowerCase`, and no paired INDD stores code 4. One privately held sample (DOM 21) has 1 paragraph style with code 4, and its IDML has `Capitalization="LowerCase"` (1 of 1). The converter leaves out other codes. **Drop caps (0x1B0D, 0x1B0E).** In every pair, the root style `[No paragraph style]` has 0 for both and IDML writes `DropCapCharacters="0"` and `DropCapLines="0"`. One pair (InDesign 20.2, IDML from the same version) also has a style with 0x1B0D = 1 and 0x1B0E = 3; its IDML has `DropCapCharacters="1"` and `DropCapLines="3"`. The different values tell the two attributes apart. **Letter spacing (0x1B31, 0x1B32).** In every pair both are 0 in the root style, as are `MaximumLetterSpacing` and `MinimumLetterSpacing` in the IDML. That does not tell them apart from other attributes that are 0 there. The rest of the evidence is a sample and its print PDF: - A paragraph style of the sample sets 0x1B31 to a positive fraction and 0x1B32 to a negative one. The style sets no tracking, and its other unidentified attributes are not fractions. - In the PDF, justified lines in that style have a character spacing (`Tc` operator) that changes from line to line, both positive and negative. So the style lets letter spacing vary in both directions, which needs a negative minimum and a positive maximum. - The values are stored as fractions, like word spacing (0x1B2E, 0x1B2F), so they are written ×100. The positive one is the maximum. A layout of the converted sample with these two attributes reproduces noticeably more of the PDF's line breaks than one without them. **Ruby (0x422D, 0x422E).** IDML writes neither attribute on the root style. In every pair the root style stores 0 for 0x422D and a 4-byte 0 for 0x422E. The first evidence was a sample typeset vertically and its print PDF: - Character runs of the sample set 0x422D to 1 together with 0x422E, whose value is a u32 length in characters followed by text segments (as story text, `objects.md`). The text is a short reading of the run's characters. - In the PDF, each such text appears in small type in a narrow column right beside the run's characters and centred on them, as ruby is set beside vertical text. - The converter decodes the value of 0x422E by this layout, whatever its length. The converter writes `RubyString` and writes `RubyFlag` with the stored number (1 in every such run; the schema type is an integer). Pairs with ruby were added later: in the corpus of 2026-10 the converter reproduces `RubyFlag` and `RubyString` in 185 of 185 IDML ruby ranges (15 distinct documents). Per-character ruby stores the readings of the base characters separated by U+3000, and IDML writes the same string. The other ruby settings are in "Ruby, kenten and warichu" below. **`Tatechuyoko` (0x425E).** Every public style that stores it has 0, and IDML writes `Tatechuyoko="false"` on all 240 root paragraph styles of the corpus IDML files; no public range has the attribute. The evidence for 1 is a sample typeset vertically and its print PDF: - Runs of a few Latin characters (letters or punctuation) set 0x425E to 1. Elsewhere in the same vertical stories, Latin text without it is drawn turned sideways: the PDF text matrix is a quarter turn, and the letters follow each other down the column. - The glyphs of each run with 1 are drawn upright (identity text matrix), side by side on one baseline, and the group is centred on the column's centre line to 0.01 pt. That is horizontal-in-vertical setting. **Shatai (0x4278 to 0x427B).** In every public file the root paragraph style stores 0x4278 = 0, 0x4279 = 45, 0x427A = 1 and 0x427B = 0, and IDML writes `ShataiMagnification="0"`, `ShataiDegreeAngle="4500"`, `ShataiAdjustTsume="true"` and `ShataiAdjustRotation="false"` on all 240 root styles; no other public style or range sets them. Only 45 and 4500 are not 0 or a boolean, so 0x4279 is the angle, written ×100. The rest of the evidence is a sample typeset vertically and its print PDF: - Runs set 0x4278 = 20, 0x4279 = 60 and 0x427A = 0. Their glyphs have the text matrix (0.850, 0.0866, 0.0866, 0.950) × point size. That matrix scales by 1 along the direction at 60° and by 0.80 across it: a distortion of 20 % at 60°. So 0x4278 is the magnification in percent and 0x4279 the angle in degrees. - A character style sets 0x4278 = 10, 0x4279 = 60 and 0x427B = 1. Its glyphs have the matrix (0.9222, 0.0843, 0, 0.9759) × point size. This is the 10 % distortion at 60°, turned so that the glyph's vertical axis stays vertical (the third value is 0). The runs above, with 0x427B = 0, are not turned. So 0x427B is `ShataiAdjustRotation`. - That leaves 0x427A for `ShataiAdjustTsume`, which also matches the root values (1 and `true`; 0x427B 0 and `false`). With 0 the glyphs keep an advance of one em; no sample has 1 outside root styles. The converter writes the angle and the two booleans. It leaves out `ShataiMagnification`: the sample stores the percentage, but the root value 0 does not show the unit IDML uses, and the angle shows that it need not be the stored number. **`KerningValue` (0x1B13).** The value 1e8 means no manual kerning. IDML writes it `KerningValue="1e+11"` (1e8 × 1000, the scale of other values) on ranges and character styles: - In every one of the 1,252 trustworthy pairs, the number of live story run records (chunk 0x262 of class 0xCA17) that store 1e8 equals the number of reference `CharacterStyleRange` elements with `KerningValue="1e+11"`: 144 in 3 documents, 0 elsewhere. - Of the 1,796 style objects (class 0x205, chunk 0x23F) that store 1e8, 1,767 are root paragraph styles (one or two per file) and 29 are character styles. The reference has `KerningValue="1e+11"` on exactly those 29 character styles (21 documents). - The preferences object (class 0x2202) stores 1e8 in 1,251 files. The schema allows no `KerningValue` on paragraph styles or `TextDefault`, so the root paragraph styles and text defaults get none. - The corpus IDMLs have 329 `KerningValue="1e+11"` (169 ranges, 160 character styles), always with that literal text. IDML writes every other value as a whole number. Every `KerningValue` in every corpus IDML is an integer or `1e+11` (6,798 integers, none with a fraction). Stored values ×1000 such as −231.8, 26.9, 58.6, 59.9, 68.0066 and −29.9945 are −232, 27, 59, 60, 68 and −30 in the reference (18 ranges of the trustworthy pairs), each the nearest integer. Some runs store a tiny negative number (about −7e−18); IDML writes `0` (10 ranges in 4 documents), not `-0`. No sample stores a value exactly halfway between two integers, so how halves round is open; the converter rounds them away from zero. The scale rests on a sample and its print PDF: - Runs of one character set 0x1B13 to −0.1, −0.06, −0.12 and 0. In the PDF, the gap between that character and the next (start of the next glyph minus the end of the first glyph's advance width) is −0.1, −0.06, −0.12 and 0 times the point size, to 0.01 pt, once the letter spacing that every other gap in the same line shows is subtracted. - The same letter pairs elsewhere, without the attribute, have other gaps, which come from the font's kerning. So the stored value replaces the font's kerning after that character. - The value is in ems. The IDML schema types `KerningValue` as a plain number. `Tracking` (0x1B0B) is also stored in ems, and IDML writes it ×1000 (thousandths of an em, from the pairs); the converter writes `KerningValue` the same way. No sample shows this scale directly. - The schema allows `KerningValue` on character styles and ranges, not on paragraph styles, so the converter leaves it out of paragraph styles (public samples without IDML have it there). **Manual kerning method.** When a story run stores both 0x1B07 and 0x1B13, and 0x1B13 is not 1e8, IDML writes `KerningMethod="$ID/Manual"` if the written `KerningValue` is not 0 and `KerningMethod="$ID/None"` if it is 0, whatever method code is stored: - Trustworthy pairs: 37 ranges in 7 documents are `$ID/Manual` (stored 79875 Optical in 34, 15972 Metrics in 3; values from −1000 to 200). 9 ranges in 2 documents with value 0 are `$ID/None` (stored Optical in 8, Metrics in 1). Runs that store a method and 1e8 keep the stored method (6 ranges, `$ID/Optical`). No range contradicts the rule. A stale pair adds 36 `$ID/Manual` ranges and 3 `$ID/None` ranges (those 3 store code 0x221 and value 0). - Runs that store a value but no method get no `KerningMethod` in IDML (728 ranges with a non-zero value, 141 with 0). - The corpus IDMLs have `$ID/Manual` 92 times, only on `CharacterStyleRange`, never on a style. No style stores a method together with a value other than 1e8, so the converter keeps the stored method on styles. - No sample has a run with code 0x221 and a non-zero value; the rule gives `$ID/Manual` there. Code 0x221 is `$ID/None`. In the trustworthy pairs it occurs 21 times in 7 documents (style lists and text defaults), and each document's count equals its number of reference styles and defaults with `KerningMethod="$ID/None"` (11 paragraph styles, 8 character styles, 2 `TextDefault`). Over all little-endian corpus files, 0x221 is the only 0x1B07 code not in the table (26 files in style lists, 12 in runs). **`StartParagraph` (0x1B37).** In every pair the root style has 0 and IDML writes `StartParagraph="Anywhere"`; no pair has another value. Code 2 rests on a sample and its print PDF: - Every paragraph whose style has code 2 is the first text on its page in the PDF, and the page before it often ends with room for more lines. Those styles set no other attribute that is unidentified. - Such paragraphs start on both odd and even pages, so code 2 is not `NextOddPage` or `NextEvenPage`. - In the sample, `NextColumn`, `NextFrame` and `NextPage` would give the same layout, so the PDF does not tell them apart. The converter writes `NextPage`. Later samples show the other codes in pairs: 1 `NextColumn` (1 style), 3 `NextFrame` (1 run) and 4 `NextOddPage` (1 style), with no run or style that contradicts them. The new codes of the table were found by aligning INDD runs with IDML ranges at the same offset over 17,914 trustworthy stories and styles by name over all 654 pairs; no run or style contradicts them, and every value converted with a new code equals the IDML. A run with manual kerning has `KerningMethod="$ID/Manual"` whatever code it stores (manual kerning method, above). **Percentages.** IDML writes a fraction attribute (`HorizontalScale`, `VerticalScale`, `AutoLeading`, the word and letter spacing and glyph scaling attributes, `RubyXScale`, `KentenXScale`) as the stored f64 times 100 with every digit of the product, binary noise included: `112.99999999999999` for a stored 1.13. Over the paragraph and character styles and story ranges of the trustworthy pairs, 24,055 reference values of these attributes have more than 9 decimals; the full product gives 23,910 of them exactly (rounding to 9 decimals gave none), and the 140,887 exact matches among the 142,524 shorter values stay the same. ### More paragraph and character attributes Found by aligning styles by name over all pairs and runs with IDML ranges at the same offset (17,914 trustworthy stories). On styles other than the root styles, each ID is present exactly when the IDML style has the attribute, and every value matches. "Styles" counts the styles that have both, other than the root styles in brackets. Encodings: bool = u16, 1 `true`; pct = f64 fraction, written × 100; num = f64 or u32 as is; swatch = u32 swatch UID, 0 = `Text Color`; stroke type = 8 bytes, u32 0 (a built-in style) then the stroke style code (`StrokeStyle/$ID/` and the name of the code, as for page items). A code not listed leaves the attribute out. | ID | IDML | Encoding | Styles | |---|---|---|---| | 0x1B14 | `HyphenateLadderLimit` | num | 780 (285) | | 0x1B68 | `HyphenateLastWord` | bool | 760 (265) | | 0x1B85 | `HyphenateAcrossColumns` | bool | 751 (256) | | 0x1B22 | `HyphenateCapitalizedWords` | bool | 709 (214) | | 0x1B34 | `MaximumGlyphScaling` | pct | 709 (214) | | 0x1B35 | `MinimumGlyphScaling` | pct | 709 (214) | | 0x1B20 | `HyphenateAfterFirst` | num | 651 (156) | | 0x1B23 | `HyphenateWordsLongerThan` | num | 636 (141) | | 0x1B39 | `KeepWithNext` | num | 619 (124) | | 0x1B2D | `DesiredWordSpacing` | pct | 587 (92) | | 0x1B21 | `HyphenateBeforeLast` | num | 580 (85) | | 0x1B30 | `DesiredLetterSpacing` | pct | 559 (64) | | 0x1B38 | `KeepAllLinesTogether` | bool | 556 (61) | | 0x1B3E | `CharacterAlignment` | 0 `AlignBaseline`, 1 `AlignEmCenter`, 2 `AlignEmBottom`, 3 `AlignEmTop`, 4 `AlignICFTop`, 5 `AlignICFBottom` (below) | 551 (56) | | 0x1B4C | `RuleAboveColor` (Properties) | swatch | 546 (51) | | 0x1B5C | `RuleAbove` | bool | 541 (46) | | 0x1B83 | `HyphenWeight` | num | 541 (46) | | 0x1B71 | `RuleAboveType` (Properties) | stroke type | 520 (25) | | 0x1B95 | `UnderlineType` (Properties) | stroke type | 541 (46) | | 0x1B72 | `RuleBelowType` (Properties) | stroke type | 505 (10) | | 0x1B9D | `StrikeThroughType` (Properties) | stroke type | 499 (4) | | 0x1A40B | `NumberingFormat` (Properties, string) | u32: 0x1A477 `1, 2, 3, 4...`, 0x1A479 `i, ii, iii, iv...`, 0x1A47A `A, B, C, D...`, 0x1A47B `a, b, c, d...`, 0x1A497 `01,02,03...`, 0x1A478 `I, II, III, IV...`, 0x1A4A7 `$ID/Circle Numbering`, 0 `None`; the last two: below | 531 (36) | | 0x1B73 | `BalanceRaggedLines` (Properties, enumeration) | 0 `NoBalancing`, 1 `VeeShape` | 528 (33) | | 0x1A422 | `NumberingAlignment` | 0 `LeftAlign`, 2 `RightAlign` | 527 (32) | | 0x1A421 | `BulletsAlignment` | 0 `LeftAlign`, 1 `CenterAlign`, 2 `RightAlign` | 505 (10) | | 0x1B4E | `RuleAboveTint` | num (−1 kept) | 526 (31) | | 0x50F29 | `ParagraphDirection` | 0 `LeftToRightDirection`, 1 `RightToLeftDirection` | 526 (31) | | 0x50F2A | `ParagraphJustification` | 0 `DefaultJustification`, 6 `NaskhKashidaJustificationFrac` | 495 roots (4 with 6) | | 0x50F1F | `CharacterDirection` | 0 `DefaultDirection`, 1 `LeftToRightDirection`, 2 `RightToLeftDirection` | 518 (23) | | 0x50F11 | `DigitsType` | 0 `DefaultDigits`, 2 `HindiDigits` | 499 (4) | | 0x1A418 | `NumberingLevel` | num | 525 (30) | | 0x1A424 | `BulletsTextAfter` | flagged string (as `NumberingExpression`) | 512 (17) | | 0x1A41D | `NumberingApplyRestartPolicy` | bool | 497 (2) | | 0x1BB8 | `StrokeAlignment` | 0 `CenterAlignment`, 2 `OutsideAlignment` | 523 (28) | | 0x4297 | `LeadingModel` | 0 `LeadingModelRoman`, 1 `LeadingModelAkiBelow`, 2 `LeadingModelAkiAbove`, 3 `LeadingModelCenter` | 523 (28) | | 0x1BC1 | `KeepWithPrevious` | bool | 522 (27) | | 0x1B93 | `UnderlineTint` | num | 519 (24) | | 0x1B8E | `UnderlineGapColor` (Properties) | swatch | 518 (23) | | 0x1B77 | `RuleAboveGapColor` (Properties) | swatch | 506 (11) | | 0x1B96 | `StrikeThroughColor` (Properties) | swatch | 502 (7) | | 0x426C | `Rensuuji` | bool | 518 (23) | | 0x4295 | `OTFProportionalMetrics` | bool | 516 (21) | | 0x1B57 | `RuleBelowLeftIndent` | num | 513 (18) | | 0x1B58 | `RuleBelowRightIndent` | num | 504 (9) | | 0x1B59 | `RuleBelowWidth` | 1 `ColumnWidth`, 2 `TextWidth` | 507 (12) | | 0x1B86 | `KeepRuleAboveInFrame` | bool | 512 (17) | | 0x1B6F | `OTFTitling` | bool | 508 (13) | | 0x1B6E | `OTFDiscretionaryLigature` | bool | 501 (6) | | 0x1BCA | `OTFSwash` | bool | 501 (6) | | 0x1B89 | `OTFStylisticSets` | num | 501 (6) | | 0x1B44 | `OverprintFill` | bool | 506 (11) | | 0x1B92 | `UnderlineOverprint` | bool | 506 (11) | | 0x1B8F | `UnderlineGapOverprint` | bool | 500 (5) | | 0x1B3D | `StrikeThru` | bool | 505 (10) | | 0x1B9A | `StrikeThroughOffset` | num (−9999 kept) | 501 (6) | | 0x1B9E | `StrikeThroughWeight` | num (−9999 kept) | 499 (4) | | 0x1B9C | `StrikeThroughTint` | num | 497 (2) | | 0x1B78 | `RuleAboveGapTint` | num | 496 (1) | | 0x1B4B | `Skew` | num | 500 (5) | | 0x1B65 | `LastLineIndent` | num | 497 (2) | | 0x1B81 | `PositionalForm` | 0 `None`, 1 `Calculate`, 2 `Initial` | 500 (5) | | 0x1B87 | `IgnoreEdgeAlignment` | bool | 498 (3) | | 0x42AD | `GlyphForm` | 0 `None`, 2 `ExpertForm`, 5 `MonospacedHalfWidthForm`, 9 `ProportionalWidthForm`, 10 `FullWidthForm` | 501 (6); code 2 below | | 0x4222 | `LeadingAki` | num (−1 kept) | 502 (7) | | 0x4223 | `TrailingAki` | num | 504 (9) | | 0x4225 | `KinsokuType` | 0 `KinsokuPushInFirst`, 2 `KinsokuPushOutOnly`, 3 `KinsokuPrioritizeAdjustmentAmount` | 502 (7) | | 0x4226 | `KinsokuHangType` | 0 `None`, 1 `KinsokuHangRegular`, 2 `KinsokuHangForce` | 498 (3); code 1 below | | 0x4227 | `BunriKinshi` | bool | 499 (4) | | 0x421D | `Tsume` | num | 499 (4) | | 0x42A6 | `CjkGridTracking` | bool | 497 (2) | | 0x422C | `RubyOpenTypePro` | bool | 505 (10) | | 0x422F | `RubyFontSize` | num | 500 (5) | | 0x4248 | `KentenFontSize` | num | 500 (5) | | 0x4232 | `RubyType` | 0 `GroupRuby`, 1 `PerCharacterRuby` | 497 (2) | | 0x4247 | `KentenKind` | 0 `None`, 1 `KentenSesameDot`, 5 `KentenSmallBlackCircle` | 496 (1); code 5 below | | 0x1DF1F | `MergeConsecutiveParaBorders` | bool | 347 | | 0x1DF1A | `ParagraphBorderTopLineWeight` | num | 393 (67) | | 0x1DF1B | `ParagraphBorderBottomLineWeight` | num | 395 (69) | | 0x1DF1C | `ParagraphBorderLeftLineWeight` | num | 387 (61) | | 0x1DF1D | `ParagraphBorderRightLineWeight` | num | 387 (61) | | 0x1DF01 | `ParagraphBorderLeftOffset` | num | 348 (22) | | 0x1DF02 | `ParagraphBorderRightOffset` | num | 347 (21) | | 0x1BF7 | `ParagraphBorderTint` | num | 336 (10) | | 0x1DF16 | `ParagraphBorderStrokeEndCap` | 0 `ButtEndCap`, 1 `RoundEndCap` | 339 (13) | | 0x1DF17 | `ParagraphBorderWidth` | 0 `ColumnWidth`, 1 `TextWidth` | 328 (2) | | 0x1DF20 | `ProviderHyphenationStyle` | 0 `HyphAll`, 3 `HyphPreferredAesthetic` | 327 (1) | | 0x1BD9 | `ParagraphShadingLeftOffset` | num | 406 (26) | | 0x1BDA | `ParagraphShadingRightOffset` | num | 406 (26) | | 0x1BD5 | `ParagraphShadingWidth` | 0 `ColumnWidth`, 1 `TextWidth` | 386 (6) | | 0x1BD7 | `ParagraphShadingClipToFrame` | bool | 388 (8) | **`NumberingFormat` codes from paragraph runs.** Two codes occur in story paragraph runs of the trustworthy pairs, each run matching a reference `ParagraphStyleRange` with that value and no other: - 0x1A478 is `I, II, III, IV...` (between 0x1A477 `1, 2, 3, 4...` and 0x1A479 `i, ii, iii, iv...`): 2 runs in one document, the only 2 ranges with that value in the corpus IDMLs. - 0x1A4A7 is `$ID/Circle Numbering`: one run in each of 10 documents; a stale pair adds an eleventh. The corpus IDMLs have 11 such values. Other codes occur only in files without IDML and are not decoded: 0xFFFFFFFF (11 files, styles), 0x1A473 (2), 0x1A472 (1), 0x1A4AA (1). **Found in the corpus of 2026-10.** These IDs are in every text default list (chunk 0x23F of the preferences, 1,460 of 1,460 files, no contradiction) and in styles and runs. Counts are the converter's output over the 1,251 trustworthy pairs, all reproduced, with no extra value: | ID | IDML | Encoding | Text defaults; styles; ranges | |---|---|---|---| | 0x1B3A | `KeepFirstLines` | num | 1,251; 1,258 paragraph styles | | 0x1B3B | `KeepLastLines` | num | 1,251; 1,258 paragraph styles | | 0x42C1 | `AllowArbitraryHyphenation` | bool | 1,251; 1,270; 30 | | 0x4264 | `GridGyoudori` | num | 1,251; 1,323; 80 | | 0x4268 | `RotateSingleByteCharacters` | bool | 1,251; 1,277; 329 | | 0x4261 | `AutoTcy` | num | 1,251; 1,285; 241 | | 0x42A3 | `ScaleAffectsLineHeight` | bool | 1,251; 1,359 paragraph and character styles; 879 | | 0x1BD4 | `ParagraphShadingOverprint` | bool | 1,032; 1,033 | | 0x1BFD | `ParagraphBorderType` (Properties) | stroke type | 918; 919 | | 0x4233 | `RubyFont` (Properties) | u32 font family UID; 0 is `$ID/` (as `BulletsFont`) | 1,251; 1,255; 23 | | 0x4257 | `KentenFillColor` (Properties) | swatch | 1,251; 1,262; 23 | `KeepFirstLines` and `KeepLastLines` are equal in every list (2 in all but one file, 1 in that one), also in every paragraph style that has them, so which ID is which is assumed from their order. 0x1BD4 is told from 0x1BD8 by one document whose shading is overprinted (0x1BD4 1, 0x1BD8 0). 0x4233 and 0x4257 are told apart in the one document that sets them by the class of the stored UID: a font family (class 0x3E03) and a colour. `ParagraphBorderType` uses the stroke type codes of page items: 0x5A29 `Solid` in 1,042 files, 0x5A3F `Japanese Dots` in one. 0x1BD4 and 0x1BFD are absent exactly where IDML has no attribute. `MergeConsecutiveParaBorders` is in IDML from version 13.1 (in none of the 3 trustworthy pairs of version 13.0, in all 46 of 13.1); the converter writes it when the header or the last session of the save history is 13.1 or later (`objects.md`, save history). Evidence over the converter's output (trustworthy pairs): 121,196 values of these attributes in paragraph and character styles, text defaults and text ranges; 120,409 reproduced, 3 wrong (`TextDefault` `DesiredLetterSpacing`, `DesiredWordSpacing` and `UnderlineType` in one document each), the rest in stories whose text differs. **More codes, corpus of 2026-10.** Over the 1,251 trustworthy pairs, the converter's output with these codes reproduces the IDML values below and adds no extra value: | ID | Code | IDML | Evidence (converter's output, trustworthy pairs) | |---|---|---|---| | 0x4266 | 2 | `GridAlignment="AlignEmTop"` | with code 3: 11,024 of 11,024 range values (2,333 more than before), 1,826 of 1,826 paragraph style values | | 0x4266 | 3 | `GridAlignment="AlignEmCenter"` | | | 0x1B3E | 3 | `CharacterAlignment="AlignEmTop"` | with codes 4 and 5: 9,603 of 9,605 range values (1,699 more) | | 0x1B3E | 4, 5 | `AlignICFTop`, `AlignICFBottom` | one run each, one document | | 0xB676 (cells) | 3 | `FirstBaselineOffset="EmboxHeight"` | 6,668 of 6,668 cell values (2,143 more) | **Page number type (0x1B09).** A u16 enumeration: 0 `AutoPageNumber`, 2 `NextPageNumber`, 3 `TextVariable`, written as the range's `PageNumberType`. The root paragraph style has 0, the `AutoPageNumber` of IDML's root style. In the trustworthy pairs the ranges with `TextVariable` are those that hold a text variable instance; the converter reproduces 278 of 278 values with no extra value. **Custom glyph (0x1B5E).** Value type 0x1B26: u16 length *n*, *n* bytes of a glyph name, u32 glyph ID. IDML writes `Properties/CustomGlyph` (type `string`) as `$ID/` and the name, for example `$ID/ampersand.smcp`. The converter leaves out records with an empty name. It reproduces 1,499 of 1,499 values in the trustworthy pairs, with no extra value. The glyph ID is not used. **Alternate glyphs (0x42AE).** A glyph chosen from the alternates of an OpenType feature (for example in the Glyphs panel) is stored on its run as attribute 0x42AE, value type 0x42DF: u16 count *n*, then per feature four bytes of the feature tag in reading order and a u32 number, in the file's byte order. IDML writes `Properties/OpenTypeFeatures` (type `list`) on the character range, after `AppliedFont`: one `ListItem` of type `list` per feature, holding a `ListItem` of type `string` with `$ID/` and the tag and a `ListItem` of type `long` with the number: ```xml $ID/aalt 1 ``` Evidence, in the corpus of 2026-10: 127 of 127 runs in the trustworthy pairs (27 files) have the IDML value at the same story offset, with 11 tags (`ss01` to `ss07`, `aalt`, `ccmp`, `liga`, `nalt`) and numbers 1, 2 and 8; no IDML range has `OpenTypeFeatures` without the attribute. Every run in the corpus stores one feature (*n* = 1). The tag order in both byte orders comes from 7 big-endian files (29 runs, no IDML): the count and number are big-endian there, the tag bytes are not reversed. Styles store the attribute with *n* = 0 (2,230 styles of the paired files); no IDML style has `OpenTypeFeatures`, and the converter writes nothing for them. The IDML form of several features is not shown, so the converter leaves out values with *n* other than 1, and values whose tag is not printable ASCII or that do not fill the value exactly. ### Ruby, kenten and warichu These attributes are in story runs (character and paragraph runs, chunk 0x262) and in style chunk 0x23F. They were found in the corpus of 2026-10 (803 pairs, 606 trustworthy) by aligning IDML character and paragraph ranges with the INDD run at the same text offset, and styles by kind and name. The evidence: - 24 character ranges in 4 pairs store a full block of CJK attributes (150 to 152 attributes each), all with default values, and their IDML ranges write every ruby, kenten and warichu attribute. Values that occur only once in the block identify the field: 0.66 ↔ 66, 0.5 ↔ 50, 4 ↔ `RubyJIS`, 7 ↔ `Auto`. - One paragraph style (a pair of version 13) sets eight ruby attributes to other values, and one ruby range (another pair of version 13) sets five others. - 326 ranges and 9 styles in 6 pairs set `RubyAutoAlign="false"` and `WarichuAlignment="LeftAlign"`. - 205 trustworthy ranges in 13 documents have `RubyXScale` and `KentenXScale`, 44 have `RubyAutoScaling`, and 573 ranges and styles have an empty ruby and kenten font style. Over all these items, each ID below is present exactly when the IDML element has the attribute, and every value matches. Encodings: bool = u16, 1 `true`; pct = f64 fraction, written ×100; code = u16, by the table; string = flag byte and in-object string. | ID | IDML | Encoding | Values seen (items) | |---|---|---|---| | 0x4231 | `RubyAlignment` | code: 1 `RubyCenter`, 2 `RubyRight`, 4 `RubyJIS` | 4: 24 ranges; 2: 1 range; 1: 1 style | | 0x4235 | `RubyParentSpacing` | code: 0 `RubyParentNoAdjustment`, 1 `RubyParentBothSides`, 2 `RubyParent121Aki` | 2: 24; 0: 1; 1: 1 | | 0x423C | `RubyParentOverhangAmount` | code: 0 `None`, 1 `RubyOverhangOneRuby`, 5 `RubyOverhangNoLimit` | 1: 24; 5: 1; 0: 1 | | 0x423A | `RubyPosition` | code: 0 `AboveRight`, 1 `BelowLeft` | 0: 24; 1: 1 | | 0x4239 | `RubyYOffset` | f64, points | 0: 24; −1.4173228: 1 | | 0x423F | `RubyParentScalingPercent` | pct | 0.66: 24; 0.4: 1 | | 0x423E | `RubyAutoScaling` | bool | true: 21 ranges and 1 style; false: 24 | | 0x423B | `RubyAutoAlign` | bool | false: 326 ranges and 9 styles; true: 24 | | 0x42B1 | `RubyAutoTcyDigits` | u16 number | 0: 24; 2: 3 styles; 5: 1 | | 0x42B2 | `RubyAutoTcyIncludeRoman` | bool | false: 24; true: 1 | | 0x4236 | `RubyXScale` | pct | 222 ranges and 1 style, 80.9 to 113.1 | | 0x424C | `KentenXScale` | pct | the same items; always equal to 0x4236 | | 0x4234 | `RubyFontStyle` (Properties) | string; empty = enumeration `Nothing` | 512 ranges and 61 styles, all empty | | 0x424B | `KentenFontStyle` (Properties) | as 0x4234 | the same items | | 0x4281 | `WarichuAlignment` | code: 0 `LeftAlign`, 7 `Auto` | 0: 326 ranges and 9 styles; 7: 24 | | 0x427D | `Warichu` | bool | false: 24 ranges; true: print PDF, below | | 0x427E | `WarichuLines` | u16 number | 2: 24; see below | | 0x427F | `WarichuSize` | pct | 0.5 ↔ 50: 24; print PDF, below | | 0x4280 | `WarichuLineSpacing` | f64, points | 0: 24; print PDF, below | `KentenKind` (0x4247) code 5 is `KentenSmallBlackCircle` (1 range). `KinsokuHangType` (0x4226) code 1 is `KinsokuHangRegular`: 31 styles and 120 ranges over all pairs (5 styles, 8 paragraph ranges and 6 story ranges in trustworthy pairs), and no other value goes with code 1. `GlyphForm` (0x42AD) code 2 is `ExpertForm`. Over the pairs of the corpus of 2026-10 that have the value, the converted output of each pair has `GlyphForm="ExpertForm"` on as many elements of each kind as its IDML: 41 character ranges in 13 pairs, 45 paragraph styles in 31 pairs and 4 text defaults (34 pairs with it in styles or text defaults). Codes 1, 3, 4, 11 and 12 occur only in samples without IDML; the converter leaves them out. With both codes, `compare.py` reports no wrong, missing or extra `GlyphForm` or `OTFFigureStyle` value in the pairs of the corpus of 2026-10, in styles, text defaults or ranges. **How IDs with equal values were told apart.** - `RubyAutoAlign` and `WarichuAlignment` are always present together, with (1, 7) for (`true`, `Auto`) and (0, 0) for (`false`, `LeftAlign`). 0x423B has value type 0x1B03, which 17 attributes use (booleans and `RubyType`), and over all styles the values of that type are 0 or 1 (9,186 of 9,186). 0x4281 has its own value type and holds 7. So 0x423B is the boolean. - `RubyYOffset`, `RubyPosition` and `RubyParentScalingPercent` are in the same 25 items. The style's values −1.4173 (f64), 1 (u16) and 0.4 (fraction) fit only one attribute each. - `RubyAlignment`, `RubyParentSpacing`, `RubyParentOverhangAmount`, `RubyAutoTcyDigits` and `RubyAutoTcyIncludeRoman`: the range sets (2, 0, 5, 5, 1) and the style (1, 1, 0, none, none). 0x423C and 0x42B1 are both 5 in the range; the 24 default ranges (1 ↔ `RubyOverhangOneRuby`, 0 ↔ `RubyAutoTcyDigits="0"`) and the style (0x423C = 0 ↔ `None`, no 0x42B1) separate them. - Ruby or kenten for 0x4234/0x424B and 0x4236/0x424C: the two values are equal in every pair item. The ruby strand (`objects.md`) decides it: its records hold only ruby attributes, including 0x4234 (5,145 runs) and 0x4236 (5,297 runs), and never 0x424B, 0x424C or a kenten ID (0x4247, 0x4248). So 0x4234 and 0x4236 are the ruby attributes. **Values without an IDML pair.** 0x423C = 2 (6 styles), 0x4281 = 2 (5 styles), 0x423F = 0.75 (5 runs) and non-empty font styles (`Regular` in 104 runs) occur only in files without IDML. The converter leaves them out. **Warichu.** No pair has warichu switched on. The evidence for `Warichu`, `WarichuSize` and `WarichuLineSpacing` is a third-party sample typeset vertically (version 21, no IDML) and its print PDF: - A run of 7 characters sets only 0x427D = 1. In the PDF the paragraph is 8.504 pt and the run is set at 4.252 pt (50 %) in two rows of 3 and 4 characters, with baselines 4.252 pt apart (row pitch = size + 0), between full-size characters. The PDFs of three saved revisions show the same. - A run of 36 characters sets 0x427D = 1, 0x427F = 0.6, 0x4280 = 0.708661, 0x4281 = 0, and 0x428E = 0x428F = 1. In the PDF the paragraph is 6.378 pt and the run is 3.827 pt (0.600 × 6.378) in two rows that start at the same position, with baselines 4.535 pt apart (3.827 + 0.708), and it continues on the next line in two rows again. So 0x427D switches warichu on, 0x427F is the size as a fraction of the point size, and 0x4280 is the extra space between the rows in points. Both runs have two rows, the second with 0x428E and 0x428F set to 1, and the root style stores 0x427E = 2. So neither 0x428E nor 0x428F is the number of rows. 0x427E is `WarichuLines`: value type 0x1B10 is used only by 0x427E, 0x428E and 0x428F, and 0x427E is the remaining one. **Other IDs of the block.** The 24 default ranges also hold IDs for the ruby and kenten fonts, colours, tints, overprint, weights, Y scales and offsets, kenten placement and characters, and two warichu IDs (0x428E, 0x428F; 2 and 2 in every IDML range). Each holds the default value in every pair, and several IDs hold the same value, so their values cannot tell them apart. Their order in IDML places them (attributes placed by order, below). ### Attributes placed by order IDML writes a style's text attributes in ascending order of their attribute ID: the XML attributes in that order, and the `Properties` children in that order among themselves. The style's own attributes (`Self`, `Name`, `Imported`, `NextStyle`, `KeyboardShortcut` and the like) are not part of this order. Evidence: the 11,983 paragraph and character styles of the trustworthy pairs that match an INDD style by kind and name (2,504 root styles, 6,671 other paragraph styles and 2,808 other character styles). Every one of their 0x23F lists is in ascending ID order, and the IDs of the other tables in this section appear in IDML in the same order: | Sequence | Adjacent pairs of IDs | Out of order | |---|---:|---:| | XML attributes | 270,429 | 0 | | `Properties` children | 44,683 | 0 | | XML attributes, with the table below | 358,891 | 0 | | `Properties` children, with the table below | 52,244 | 0 | The IDs below were placed by this order. For each IDML attribute that no known ID explains, the candidates are the stored IDs without a mapping that lie between the nearest known IDs before and after it. Intersected over every trustworthy style that has the attribute, and assigned in order within each gap, this gives one assignment in every gap but four. The payload type has to fit: a `true` or `false` value needs a 2-byte value of type 0x1B03, 0x1B04, 0x1B09 or 0x1B1F, a number an f64 equal to the IDML value (or to it divided by 100 for a percentage), and a colour or font in `Properties` a UID (type 0x1B05). In the four gaps the type decides: - 0x1B79–0x1B7D, five IDs for four attributes: 0x1B7A has type 0x1B0D (0 in 1,252 of 1,252 roots), not a boolean type, so `RuleAboveGapOverprint` is 0x1B79. 0x1B7B is the only UID: `RuleBelowGapColor`. - 0x1BFB, 0x1BFC and 0x1BFE for `ParagraphBorderGapTint` (`-1`) and `ParagraphBorderGapOverprint` (`false`): 0x1BFE is an f64 of 1.0 in 919 of the 919 roots that store it, which fits neither. - the ruby IDs 0x4237–0x4246, ten IDs for eight XML attributes and two `Properties` children: 0x4240 and 0x4243 are the only UIDs. - the kenten, tatechuyoko, warichu and OpenType IDs 0x424D–0x42B3, 23 IDs for 22 XML attributes and one `Properties` child: 0x425A is the only UID. The table counts the trustworthy pairs: styles (the root paragraph style and others) and character or paragraph ranges whose stored value the converter writes equal to the reference IDML's, and those it writes with another value or where the IDML has none (Differ). Every value of these attributes that the IDML has on a matched style or range is written. Encodings: bool = u16, 0 `false`, 1 `true`; pct = f64 fraction, written ×100; colour = swatch UID, 0 `Text Color`; code = u16 or u32, by the listed codes. The second part of the encoding column is the value type. "Order only" marks an ID whose value is the same in every trustworthy item, so only the order and the payload type place it. The four OpenType alternates vary, but always together, so the order alone separates them too. | ID | IDML | Encoding, type | Equal: root styles | other styles | ranges | Differ | Values seen | Placed by | |---|---|---|---:|---:|---:|---:|---|---| | 0x1B41 | `StrokeTint` | f64, 0x1BB1 | 1,252 | 9 | 23 | 0 | `-1` (1,282), `100` (2) | order, values | | 0x1B43 | `OverprintStroke` | bool, 0x1B04 | 1,252 | 12 | 23 | 0 | `false` (1,287) | order only | | 0x1B45 | `GradientStrokeAngle` | f64, 0x1B2C | 1,252 | 7 | 23 | 0 | `0` (1,282) | order only | | 0x1B47 | `GradientStrokeLength` | f64, 0x1B13 | 1,252 | 7 | 23 | 0 | `-1` (1,282) | order only | | 0x1B49 | `GradientStrokeStart` | two f64, 0x1B22 | 1,252 | 7 | 23 | 0 | `0 0` (1,282) | order only | | 0x1B5A | `RuleAboveOverprint` | bool, 0x1B04 | 1,252 | – | – | 0 | `false` (1,252) | order only | | 0x1B5B | `RuleBelowOverprint` | bool, 0x1B04 | 1,252 | – | – | 0 | `false` (1,252) | order only | | 0x1B6C | `OTFOrdinal` | bool, 0x1B04 | 1,252 | 8 | 23 | 0 | `false` (1,283) | order only | | 0x1B6D | `OTFFraction` | bool, 0x1B04 | 1,252 | 8 | 56 | 0 | `false` (1,283), `true` (33) | order, values | | 0x1B79 | `RuleAboveGapOverprint` | bool, 0x1B04 | 1,252 | – | – | 0 | `false` (1,252) | order only | | 0x1B7B | `RuleBelowGapColor` (Properties) | colour, 0x1B05 | 1,252 | 10 | 357 | 0 | `Swatch/None` (1,252), `Text Color` (367) | order, values | | 0x1B7C | `RuleBelowGapTint` | f64, 0x1BB1 | 1,252 | – | – | 0 | `-1` (1,252) | order only | | 0x1B7D | `RuleBelowGapOverprint` | bool, 0x1B04 | 1,252 | – | – | 0 | `false` (1,252) | order only | | 0x1B82 | `OTFMark` | bool, 0x1B03 | 1,252 | – | 23 | 0 | `true` (1,275) | order only | | 0x1B84 | `OTFLocale` | bool, 0x1B03 | 1,252 | – | 23 | 0 | `true` (1,275) | order only | | 0x1B88 | `OTFSlashedZero` | bool, 0x1B04 | 1,252 | 7 | 32 | 0 | `false` (1,282), `true` (9) | order, values | | 0x1B8A | `OTFHistorical` | bool, 0x1B04 | 1,252 | 7 | 23 | 0 | `false` (1,282) | order only | | 0x1B90 | `UnderlineGapTint` | f64, 0x1BB1 | 1,252 | 7 | 23 | 0 | `-1` (1,282) | order only | | 0x1B97 | `StrikeThroughGapColor` (Properties) | colour, 0x1B05 | 1,252 | 17 | 729 | 0 | `Swatch/None` (1,282), `Text Color` (716) | order, values | | 0x1B98 | `StrikeThroughGapOverprint` | bool, 0x1B04 | 1,252 | 7 | 23 | 0 | `false` (1,282) | order only | | 0x1B99 | `StrikeThroughGapTint` | f64, 0x1BB1 | 1,252 | 7 | 23 | 0 | `-1` (1,282) | order only | | 0x1B9B | `StrikeThroughOverprint` | bool, 0x1B04 | 1,252 | 7 | 23 | 0 | `false` (1,282) | order only | | 0x1BC0 | `SplitColumnOutsideGutter` | f64, 0x1BA8 | 1,252 | – | – | 0 | `0` (1,252) | order only | | 0x1BC3 | `SpanColumnMinSpaceBefore` | f64, 0x1BA8 | 1,252 | – | – | 0 | `0` (1,252) | order only | | 0x1BD8 | `ParagraphShadingSuppressPrinting` | bool, 0x1B04 | 1,033 | – | – | 0 | `false` (1,033) | order only | | 0x1BF8 | `ParagraphBorderOverprint` | bool, 0x1B04 | 919 | – | – | 0 | `false` (919) | order only | | 0x1BFB | `ParagraphBorderGapTint` | f64, 0x1BB1 | 919 | – | – | 0 | `-1` (919) | order only | | 0x1BFC | `ParagraphBorderGapOverprint` | bool, 0x1B04 | 919 | – | – | 0 | `false` (919) | order only | | 0x4237 | `RubyYScale` | pct, 0x1B16 | 1,252 | 7 | 23 | 0 | `100` (1,282) | order only | | 0x4238 | `RubyXOffset` | f64, 0x1B14 | 1,252 | 7 | 23 | 0 | `0` (1,282) | order only | | 0x423D | `RubyOverhang` | bool, 0x1B04 | 1,252 | 7 | 23 | 0 | `false` (1,282) | order only | | 0x4240 | `RubyFill` (Properties) | colour, 0x1B05 | 1,252 | 7 | 23 | 0 | `Text Color` (1,282) | order only | | 0x4241 | `RubyTint` | f64, 0x1BB1 | 1,252 | 7 | 23 | 0 | `-1` (1,282) | order only | | 0x4242 | `RubyOverprintFill` | u16 code 0xFFFF `Auto`; 0x1BBB | 1,252 | – | 23 | 0 | `Auto` (1,275) | order only | | 0x4243 | `RubyStroke` (Properties) | colour, 0x1B05 | 1,252 | 7 | 23 | 0 | `Text Color` (1,282) | order only | | 0x4244 | `RubyStrokeTint` | f64, 0x1BB1 | 1,252 | 7 | 23 | 0 | `-1` (1,282) | order only | | 0x4245 | `RubyOverprintStroke` | u16 code 0xFFFF `Auto`; 0x1BBB | 1,252 | – | 23 | 0 | `Auto` (1,275) | order only | | 0x4246 | `RubyWeight` | f64, 0x1BB6 | 1,252 | 7 | 23 | 0 | `-1` (1,282) | order only | | 0x424A | `KentenFont` (Properties) | font family UID, 0 `$ID/`; 0x1B05 | 1,252 | 1 | 23 | 0 | `$ID/` (1,275), a composite font (1) | order, values | | 0x424D | `KentenYScale` | pct, 0x1B16 | 1,252 | 7 | 23 | 0 | `100` (1,282) | order only | | 0x424F | `KentenPlacement` | f64, 0x1B14 | 1,252 | 10 | 23 | 0 | `0` (1,285) | order only | | 0x4250 | `KentenAlignment` | u16 code 1 `AlignKentenCenter`; 0x4238 | 1,252 | 10 | 23 | 0 | `AlignKentenCenter` (1,285) | order only | | 0x4251 | `KentenPosition` | u16 code 0 `AboveRight`; 0x1B04 | 1,252 | 10 | 23 | 0 | `AboveRight` (1,285) | order only | | 0x4252 | `KentenCustomCharacter` | u16 code 0 empty; 0x1B0D | 1,252 | 7 | 23 | 0 | empty (1,282) | order only | | 0x4253 | `KentenCharacterSet` | u16 code 0 `CharacterInput`; 0x1B0D | 1,252 | 7 | 23 | 0 | `CharacterInput` (1,282) | order only | | 0x4258 | `KentenTint` | f64, 0x1BB1 | 1,252 | 7 | 23 | 0 | `-1` (1,282) | order only | | 0x4259 | `KentenOverprintFill` | u16 code 0xFFFF `Auto`; 0x1BBB | 1,252 | – | 23 | 0 | `Auto` (1,275) | order only | | 0x425A | `KentenStrokeColor` (Properties) | colour, 0x1B05 | 1,252 | 7 | 23 | 0 | `Text Color` (1,282) | order only | | 0x425B | `KentenStrokeTint` | f64, 0x1BB1 | 1,252 | 7 | 23 | 0 | `-1` (1,282) | order only | | 0x425C | `KentenOverprintStroke` | u16 code 0xFFFF `Auto`; 0x1BBB | 1,252 | – | 23 | 0 | `Auto` (1,275) | order only | | 0x425D | `KentenWeight` | f64, 0x1BB6 | 1,252 | 7 | 23 | 0 | `-1` (1,282) | order only | | 0x425F | `TatechuyokoXOffset` | f64, 0x1B14 | 1,252 | 7 | 23 | 0 | `0` (1,282) | order only | | 0x4260 | `TatechuyokoYOffset` | f64, 0x1B14 | 1,252 | 7 | 23 | 0 | `0` (1,282) | order only | | 0x4262 | `AutoTcyIncludeRoman` | bool, 0x1B04 | 1,252 | – | – | 0 | `false` (1,252) | order only | | 0x4263 | `Jidori` | u16, 0x1B0D | 1,252 | 7 | 23 | 0 | `0` (1,282) | order only | | 0x4267 | `CharacterRotation` | f64, 0x1B14 | 1,252 | 7 | 25 | 0 | `0` (1,282), `3` (2) | order, values | | 0x4278 | `ShataiMagnification` | f64, 0x42A4 | 1,252 | 7 | 23 | 0 | `0` (1,282) | order only | | 0x428E | `WarichuCharsBeforeBreak` | u16, 0x1B10 | 1,252 | 7 | 23 | 0 | `2` (1,282) | order only | | 0x428F | `WarichuCharsAfterBreak` | u16, 0x1B10 | 1,252 | 7 | 23 | 0 | `2` (1,282) | order only | | 0x4294 | `OTFHVKana` | bool, 0x1B04 | 1,252 | 8 | 23 | 0 | `false` (1,283) | order only | | 0x4296 | `OTFRomanItalics` | bool, 0x1B04 | 1,252 | 8 | 23 | 0 | `false` (1,283) | order only | | 0x42A4 | `ParagraphGyoudori` | bool, 0x1B04 | 1,252 | 12 | 1 | 0 | `false` (1,252), `true` (13) | order, values | | 0x42B3 | `RubyAutoTcyAutoScale` | bool, 0x1B03 | 1,252 | 7 | 23 | 0 | `true` (1,282) | order only | | 0x1A407 | `NumberingStartAt` | u32, 0x1B00 | 1,252 | – | 7 | 0 | `1` (1,252), `2` (5), `3` (2) | order, values | | 0x1A41E | `NumberingRestartPolicies` (Properties) | three u32 (below), 0x1A4A6 | 1,252 | 1 | – | 0 | `AnyPreviousLevel` 0 0 (1,252), `AfterSpecificLevel` 1 1 (1) | order, values | | 0x1DF05 | `ParagraphBorderStrokeEndJoin` | u32 code 0 `MiterEndJoin`; 0x1DF3B | 919 | – | – | 0 | `MiterEndJoin` (919) | order only | | 0x1DF06 | `ParagraphBorderTopLeftCornerOption` | u32 code 0 `None`; 0x1DF51 | 919 | – | – | 0 | `None` (919) | order only | | 0x1DF07 | `ParagraphBorderTopRightCornerOption` | u32 code 0 `None`; 0x1DF51 | 919 | – | – | 0 | `None` (919) | order only | | 0x1DF08 | `ParagraphBorderBottomLeftCornerOption` | u32 code 0 `None`; 0x1DF51 | 919 | – | – | 0 | `None` (919) | order only | | 0x1DF09 | `ParagraphBorderBottomRightCornerOption` | u32 code 0 `None`; 0x1DF51 | 919 | – | – | 0 | `None` (919) | order only | | 0x1DF1E | `ParagraphBorderDisplayIfSplits` | bool, 0x1B04 | 919 | – | – | 0 | `false` (919) | order only | | 0x50F15 | `Kashidas` | u32 code 0 `DefaultKashidas`; 0x50F1B | 1,252 | – | 23 | 0 | `DefaultKashidas` (1,275) | order only | | 0x50F2B | `ParagraphKashidaWidth` | f64, 0x1DF21 | 1,243 | – | – | 0 | `2` (1,243) | order only | | 0x50F2F | `XOffsetDiacritic` | f64, 0x1B14 | 1,252 | – | – | 0 | `0` (1,252) | order only | | 0x50F30 | `YOffsetDiacritic` | f64, 0x1B14 | 1,252 | – | – | 0 | `0` (1,252) | order only | | 0x50F3C | `OTFOverlapSwash` | bool, 0x1B04 | 1,252 | – | 25 | 0 | `false` (1,275), `true` (2) | order only (the four vary together) | | 0x50F3D | `OTFStylisticAlternate` | bool, 0x1B04 | 1,252 | – | 25 | 0 | `false` (1,275), `true` (2) | order only (the four vary together) | | 0x50F3E | `OTFJustificationAlternate` | bool, 0x1B04 | 1,252 | – | 25 | 0 | `false` (1,275), `true` (2) | order only (the four vary together) | | 0x50F3F | `OTFStretchedAlternate` | bool, 0x1B04 | 1,252 | – | 25 | 0 | `false` (1,275), `true` (2) | order only (the four vary together) | | 0x50F40 | `KeyboardDirection` | u32 code 0 `DefaultDirection`, 1 `LeftToRightDirection`; 0x50F2E | 1,252 | – | 34 | 0 | `DefaultDirection` (1,252), `LeftToRightDirection` (34) | order, values | Notes: - Tints, weights and lengths are written as stored (`-1`, and `100` for the 2 `StrokeTint` values that store 100), not by the page item tint rule. - `KentenPosition` has a boolean value type but an enumeration value (0 = `AboveRight`). - The enumerations list only the codes seen. Other codes are left out and reported by `indd audit`. Each has one code in the trustworthy pairs, except `KeyboardDirection`: code 1 is `LeftToRightDirection` in 34 of 34 character ranges of one pair. It has the value type of `CharacterDirection` (0x50F2E), whose code 1 is the same value. - `NumberingRestartPolicies` is a `Properties` child with three XML attributes and no text. Its value is three u32: (0, 0, 0) in the 1,252 roots is `RestartPolicy="AnyPreviousLevel" LowerLevel="0" UpperLevel="0"`, and (1, 1, 1) in one other paragraph style is `RestartPolicy="AfterSpecificLevel" LowerLevel="1" UpperLevel="1"`. The second triple does not show which word is the lower level and which the upper, so the converter writes these two triples only. **Stored but never written.** The root paragraph styles also store IDs that no IDML attribute matches: after this table, every text attribute of the trustworthy IDML styles is explained by a stored ID except `Label` (6 styles) and `HyphenQuality` (7 styles). Such IDs, with the number of the 1,252 trustworthy roots that store each: 0x1B36, 0x1B3F, 0x1B5F–0x1B63, 0x1B7A, 0x4228, 0x4230, 0x4249, 0x4269–0x426B, 0x4290, 0x429F, 0x1300F, 0x13516, 0x13549 and 0x20714 (1,252 each), 0x1A405 and 0x1A415 (1,228), 0x1BFE (919), and in later versions 0xBA429, 0xBA42A and eight IDs from 0x16923C to 0x16924A (24 each) and 18 IDs from 0x5351D to 0x58820 (2 each). The converter does not write them. ### Root styles and text defaults The root paragraph style `[No paragraph style]` and the document's text defaults (`TextDefault` in `Resources/Preferences.xml`) take their values from attribute lists like other styles, but IDML leaves some of them out. Over the 489 trustworthy pairs (`measurement.md`): | Value | IDML | |---|---| | `NextStyle` of the root paragraph style | never (0 of 489) | | `AllNestedStyles` of the root paragraph style | never (0 of 489) | | `AllNestedStyles` of `TextDefault` | only when the list has items: absent in 485, written in the 4 with items | | `AllNestedStyles` of other paragraph styles | only when the list has items: written on 115 of 115 trustworthy styles with items, on 0 of the 2 that store a count of 0 | | `AllGREPStyles` and `AllLineStyles` of the root paragraph style | never (0 of 1,251 trustworthy pairs after 2026-10); the root style stores a count of 0 in all 1,251 | | `AllGREPStyles` of `TextDefault` | only when the list has items (48 of 48 such lists written, 1,251 trustworthy pairs) | | Empty `TabList` of the root paragraph style | DOM 8 on: 482 of 482; DOM 7: 0 of 7 | | `TabList` of `TextDefault` | DOM 8 on: in 482 of 482 (459 empty); DOM 7: 0 of 7 | The converter leaves out these values accordingly; it writes a non-empty `TabList` in any version. ### Paragraph borders and shading: corners and origins | ID | IDML | Styles with both (other than the root) | |---|---|---| | 0x1DF12 | `ParagraphShadingTopLeftCornerRadius` | 452 (126) | | 0x1DF14 | `ParagraphShadingBottomLeftCornerRadius` | 452 (126) | On every style other than the root the ID is present exactly when the IDML style has the attribute, and every value matches. The two are told apart by 8 styles (4.5 against 0). **Tied groups.** The IDs of each group below have the same value in every style, so which ID belongs to which attribute is not known. The converter writes a group only when all its IDs are present and equal: | IDs | IDML | Styles | |---|---|---| | 0x1DF0A–0x1DF0D | the four `ParagraphBorder…CornerRadius` | 446 (120) | | 0x1DF13, 0x1DF15 | `ParagraphShadingTopRightCornerRadius`, `ParagraphShadingBottomRightCornerRadius` | 444 (118) | | 0x1DF0E, 0x1DF10 | `ParagraphShadingTopLeftCornerOption`, `ParagraphShadingBottomLeftCornerOption` | 339 (13) | | 0x1DF0F, 0x1DF11 | `ParagraphShadingTopRightCornerOption`, `ParagraphShadingBottomRightCornerOption` | 331 (5) | The shading corner option codes are 0 `None`, 0x5A15 `RoundedCorner` and 0x5A18 `BevelCorner`. **Origins.** The two IDs of each pair always change together; the converter writes a pair only for these combinations of codes: | IDs | IDML | Codes → values | Styles | |---|---|---|---| | 0x1BDD, 0x1BDE | `ParagraphShadingTopOrigin`, `ParagraphShadingBottomOrigin` | 0, 0 → `AscentTopOrigin`, `DescentBottomOrigin`; 3, 2 → `EmBoxTopOrigin`, `EmBoxBottomOrigin`; 1, 1 → `BaselineTopOrigin`, `BaselineBottomOrigin` | 378, 14, 4 | | 0x1DF18, 0x1DF19 | `ParagraphBorderTopOrigin`, `ParagraphBorderBottomOrigin` | 0, 0 → `AscentTopOrigin`, `DescentBottomOrigin`; 3, 2 → `EmBoxTopOrigin`, `EmBoxBottomOrigin` | 325, 13 | Attributes whose value never varies in the corpus cannot be located this way. The converter writes the root styles' values for them from IDML observation (`idml-values.md`); styles and ranges inherit them. Many others are placed by their order in IDML (attributes placed by order, above). The attributes left include `KeepFirstLines`, `KeepLastLines`, `KeepAllLinesTogether`, `KeepWithNext`, `KeepWithPrevious`, `Skew`, `StrikeThru`, `LastLineIndent`, `RuleAbove`, `RuleAboveColor`, the `RuleBelow` indents and width, and the hyphenation and word-spacing settings other than those in the table. Some attributes vary but cannot be told apart from another attribute, because both always have the same value or appear in the same styles: - left and right shading offsets, left and right border offsets, and the four border line weights (equal in every sample); - `KeepRuleAboveInFrame` and `RuleAboveType` (always set together, with other attributes, in one pair); - `TreatIdeographicSpaceAsSpace` and `DiacriticPosition`. `OTFFigureStyle` code 3 is `TabularOldstyle`. One pair has it, on one character range, and its IDML has `TabularOldstyle` on that range. The value occurs in 3 more IDMLs of the corpus of 2026-10, none with an INDD. **`ParagraphBreakType` (0x1B6A) codes.** The codes are those of `StartParagraph`. The evidence is the 112 pairs of the corpus of 2026-10 whose IDML has a `CharacterStyleRange` with a value other than `Anywhere` (no paragraph range has one). In each story the runs with a code other than 0 were aligned in order with those ranges; every story of these pairs aligns: | Code | IDML value | Trustworthy pairs | Stale pairs | |---|---|---|---| | 1 | `NextColumn` | 642 ranges in 74 documents | 80 in 11 | | 2 | `NextPage` | 159 in 19 | 16 in 8 | | 3 | `NextFrame` | 76 in 19 | 0 | | 4 | `NextOddPage` | 5 in 1 (version 20.1) | 0 | `NextEvenPage` occurs in no IDML. Its code is not known, and the converter leaves other codes out. **Empty font styles.** A list can store an empty `FontStyle` (0x1B02). IDML then writes the style of a font of the list's `AppliedFont` family: the font whose style is `Regular`, or the family's first font if none is. Over the 1,251 trustworthy pairs, writing the stored empty string gave 294 wrong `FontStyle` values (ParagraphStyle 69, CharacterStyle 35, TextDefault 11, text ranges 179); with this rule 243 of them match and no value that matched before changes. 20 of the 243 come from the first-font case (a family of `Light` and `LightItalic`, IDML `Light`). So every empty value that IDML also writes empty is in a list without an `AppliedFont`, or in a family without a `Regular` font whose first font has an empty style. Not explained: 22 empty values that IDML writes as `Nothing`, and lists without an `AppliedFont`, which keep the empty string. **Font family names.** `AppliedFont` and `BulletsFont` name a font family (class 0x3E03), and IDML writes the family's name. An `AppliedFont` whose family stands for a composite font is written as a reference to the composite font, except in character styles. See `fonts.md`, composite fonts as applied fonts.