feat(odt): nested containers and unique object names after cloning (#138)

[vaceslav]

Summary

This is step 4 of OpenDocument Text support. It covers two areas:

• Template constructs inside nested containers: text boxes, footnotes and endnotes, sections, nested lists, and header or footer tables.
• Unique object names after loop cloning. This is the ODT counterpart of DrawingIdAllocator (bug(loops): cloned loop content duplicates drawing/shape ids (docPr id, v:shape id); edge cases in table/cell detection #178).

Part of #138

Design

See docs/superpowers/specs/2026-09-26-odt-support-design.md, phase 4. This PR updates the spec.

Nested containers already went through the same ProcessContainer path in PRs 1–3, and this PR adds tests for them. A text box or note inside a loop iteration is processed with that iteration's context. An emptied text box or header gets an empty paragraph.

The new work is OdtUniqueNames. Loop cloning copies these identifiers:

• draw:name (frames and shapes)
• table:name
• the text:name of sections
• the text:id of notes
• xml:id

ODF requires these to be unique. LibreOffice renames duplicate frames and tables on load, which breaks references to them, and duplicate xml:ids make the document invalid.

A single pass runs after processing, over the body plus headers and footers, which share one name space:

• The first occurrence keeps its name. Later duplicates get _2, _3, and so on, avoiding names that already exist.
• Duplicate xml:ids are removed from the later copies. They only link RDF metadata, which belongs to the original.
• styles.xml stays byte-for-byte unchanged unless it actually contains a duplicate.

draw:name handling was planned for PR 5 and has been moved into this PR, because all the name fixes belong in one pass.

Conservative decision, documented in the spec: bookmark and annotation names are not renamed. They come in start/end pairs, LibreOffice tolerates duplicates, and DOCX does not rename loop-cloned bookmarks either.

Changes

• OpenDocument/OdtUniqueNames (new, internal).
• OdtTemplateEngine.Process runs the uniqueness pass after the body and the headers/footers.
• Spec: phase table and component table updated, plus a new known limitation (bookmarks).

The DOCX code is unchanged.

Tests

• Odt/OdtContainerTests (10 tests):
  • A loop that clones frames with text boxes, footnotes, tables and sections gets unique names and ids, and the nested content is processed per item.
  • New names avoid names that already exist.
  • Header and body share one name space.
  • Loops inside a text box and inside an endnote.
  • An empty loop inside a text box leaves one empty paragraph.
  • A loop inside a section.
  • Nested list-item loops inside a list-item loop, including access to the parent variable.
  • A table-row loop inside a footer table.
  • A conditional that removes the whole header content leaves an empty paragraph.
• New LibreOffice round trip ClonedFramesTablesAndSections_KeepTheirUniqueNamesInLibreOffice. The processed document is loaded and saved again by LibreOffice. Afterwards Box/Box_2, Prices/Prices_2 and Details/Details_2 are still present unchanged, so LibreOffice accepted them without renaming, and the content converts to text.
• All 6 LibreOffice tests passed locally on net8.0, net9.0 and net10.0 (LibreOffice 25, macOS).
• The full suite is green: 2008 tests on each of net8, net9 and net10, plus the Tools and Converter tests. CI-mode build and dotnet format pass.

Public API impact: additive

No new public symbols. PublicAPI.*.txt is unchanged. The only behavior change is in the unreleased OdtTemplateProcessor: duplicate object names in its output are made unique.

[codecov-commenter]

⚠️ Please install the to ensure uploads and comments are reliably processed by Codecov.

Codecov Report

❌ Patch coverage is 97.36842% with 1 line in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
...riasDev.Templify/OpenDocument/OdtTemplateEngine.cs	87.50%	0 Missing and 1 partial ⚠️

📢 Thoughts on this report? Let us know!