mirror of
https://github.com/storytold/wordcraft.git
synced 2026-11-10 04:09:53 -06:00
Claude
c26b952e83
A comment, tracked change or index mark reaching from one paragraph into another, or a text box continuing in another's, ties paragraphs into a group: written as read while every paragraph of it is as read, else all from the model, which names what ties them afresh. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01GBVQLyDnrG8Tn1yDQcUabC
118 lines
6.5 KiB
Markdown
118 lines
6.5 KiB
Markdown
# ODT element audit
|
|
|
|
> **Last reviewed:** 2026-11-04 · **Last updated:** 2026-11-04 · **Change:** trivial (paragraphs tied together kept as read as groups); previously trivial (paragraphs with charts kept as read); previously trivial (sections in table cells and at empty paragraphs kept); previously trivial (password, number, time and grid form controls kept); previously minor (first version: every element of the ODF 1.3 schema against two corpora and a generated sample of each element never seen)
|
|
|
|
Every element the ODF 1.3 schema (OASIS, Part 3, RELAX NG) defines, checked against what WordCraft's ODT reader and writer do with it. The question for each: does a file holding it open, and does a save without edits keep it? (See [file-format-parity.md](file-format-parity.md) for what WordCraft models; markup it doesn't model is written back as read while unedited: [odf-interpretation.md](odf-interpretation.md).)
|
|
|
|
## Method
|
|
|
|
1. **The schema**: its 596 element names, by namespace.
|
|
2. **Two corpora** (kept locally, not in the repository): 133 documents (LibreOffice's templates, bug-tracker and converter files, documents made for the checks) and 232 from the ODF Toolkit's and Calligra's test data, made by Microsoft Office, OpenOffice.org, StarOffice, LibreOffice, Calligra/KOffice, Lotus Symphony and others. Each is opened and saved by WordCraft, both copies are rendered by LibreOffice and compared page by page and by text, and a loss check lists every element, attribute and value the saved file has fewer of.
|
|
3. **Elements no file held**: a small document of our own for each one a text document can hold, opened and saved by WordCraft without edits and with an edit, each saved copy checked well-formed and opened in LibreOffice.
|
|
|
|
## Namespaces
|
|
|
|
| Namespace | Elements | Seen in a file | Note |
|
|
|---|---|---|---|
|
|
| `text:` | 174 | 124 | |
|
|
| `table:` | 98 | 11 | most are a spreadsheet's (ranges, pivot tables, database ranges); a text document's tables use few |
|
|
| `draw:` | 44 | 27 | |
|
|
| `style:` | 43 | 42 | |
|
|
| `office:` | 28 | 23 | |
|
|
| `form:` | 30 | 13 | |
|
|
| `number:` | 27 | 18 | |
|
|
| `svg:` | 10 | 2 | |
|
|
| `dc:` | 6 | 6 | |
|
|
| `meta:` | 15 | 14 | |
|
|
| `config:` | 5 | 5 | settings.xml, kept as read |
|
|
| `script:` | 1 | 1 | |
|
|
| `math:` | 1 | 1 | MathML, in formula objects |
|
|
| `xforms:` | 1 | 0 | |
|
|
| `chart:` | 25 | 18 | charts, read from their own objects |
|
|
| `dr3d:` | 6 | 0 | 3D scenes, kept as read with their paragraph |
|
|
| `presentation:` | 20 | 6 | presentations (a text document holds none) |
|
|
| `anim:` | 12 | 0 | presentations |
|
|
| `db:` | 50 | 0 | databases only |
|
|
|
|
## Elements no file held, a text document can
|
|
|
|
Each in a document of our own: it opens, saves without edits keeping it, and saves with an edit; every saved copy is well-formed and LibreOffice opens it.
|
|
|
|
| Element | Unedited save |
|
|
|---|---|
|
|
| `draw:contour-path` | kept |
|
|
| `draw:glue-point` | kept |
|
|
| `draw:image-map` | kept |
|
|
| `draw:plugin` | kept |
|
|
| `form:connection-resource` | kept |
|
|
| `form:grid` | kept (as read, with its columns; LibreOffice opens the saved file, though its PDF export of a grid control fails whenever a document has automatic styles, ours or its own) |
|
|
| `form:list-value` | kept (the sample held none; a list box's values are kept with the form) |
|
|
| `form:password` | kept (as read) |
|
|
| `number:boolean-style` | kept |
|
|
| `number:embedded-text` | kept |
|
|
| `number:era` | kept |
|
|
| `number:fill-character` | kept |
|
|
| `number:fraction` | kept |
|
|
| `number:percentage-style` | kept |
|
|
| `number:quarter` | kept |
|
|
| `number:week-of-year` | kept |
|
|
| `style:region-center` | kept |
|
|
| `text:alphabetical-index-auto-mark-file` | kept with its index (as read) |
|
|
| `text:author-initials` | kept |
|
|
| `text:creation-time` | kept |
|
|
| `text:drop-down` | kept |
|
|
| `text:editing-cycles` | kept |
|
|
| `text:editing-duration` | kept |
|
|
| `text:expression` | kept |
|
|
| `text:hidden-paragraph` | kept |
|
|
| `text:linenumbering-separator` | kept |
|
|
| `text:meta-field` | kept |
|
|
| `text:modification-time` | kept |
|
|
| `text:note-continuation-notice-forward` | kept |
|
|
| `text:object-index` | kept |
|
|
| `text:page-variable-get` | kept |
|
|
| `text:page-variable-set` | kept |
|
|
| `text:print-time` | kept |
|
|
| `text:printed-by` | kept |
|
|
| `text:sender-country` | kept |
|
|
| `text:sender-initials` | kept |
|
|
| `text:sender-phone-private` | kept |
|
|
| `text:sender-title` | kept |
|
|
| `text:sort-key` | kept |
|
|
| `text:table-formula` | kept |
|
|
| `text:table-index` | kept |
|
|
| `text:toc-mark-start` | kept |
|
|
| `text:toc-mark` | kept |
|
|
| `text:user-field-input` | kept |
|
|
| `text:user-index-mark-start` | kept |
|
|
| `text:user-index-mark` | kept |
|
|
| `text:user-index` | kept |
|
|
|
|
## Found, fixed
|
|
|
|
The check found these, each fixed with a test (see [file-format-parity.md](file-format-parity.md), 2026-11-03 and 2026-11-04):
|
|
|
|
- A file the package held without a media type (another application's, LibreOffice's empty accelerator configuration) was written without its manifest entry, and LibreOffice refused the saved file.
|
|
- A section hidden with `text:display="none"` showed its text once written without the section.
|
|
- A reference mark was written as a bookmark only, so references to it found nothing.
|
|
- Lists lost what a paragraph's markup can't say (a label followed by a space, picture bullets, an unnumbered header): unedited lists are now kept as read.
|
|
- Styles of different families, and list styles, sharing a name overwrote each other among the definitions kept.
|
|
- A form's data source (`form:connection-resource`) was taken for a control; index marks over text, plugins and floating frames weren't kept.
|
|
- Password, number, time and grid form controls (ODF 1.3 §13.5), of kinds WordCraft doesn't model, were dropped; they are kept as read, in their place in the text, as buttons and image controls were.
|
|
|
|
## Left
|
|
|
|
- **A paragraph showing a form control**: written from the model, which keeps what WordCraft models of it.
|
|
- **Calligra's page sequence** (`text:page-sequence`, a layout of fixed pages) and **numbered paragraphs** (`text:numbered-paragraph`): read as text, written from the model.
|
|
- **Soft page breaks between paragraphs** (a layout hint LibreOffice recomputes) and **unused automatic styles**: not written.
|
|
|
|
## Revision history
|
|
|
|
| Date | Change | What |
|
|
|---|---|---|
|
|
| 2026-11-04 | minor | First version |
|
|
| 2026-11-04 | trivial | Password, number, time and grid form controls kept |
|
|
| 2026-11-04 | trivial | Sections in table cells and at empty paragraphs kept |
|
|
| 2026-11-04 | trivial | Paragraphs with charts kept as read |
|
|
| 2026-11-04 | trivial | Paragraphs tied together (a comment, a change or a mark reaching into another, linked text boxes) kept as read as groups |
|