SD-2440 - fix: toc not loading on paragraphs

[chittolinag]

Issue

Importing documents where the TOC field sits inline inside a paragraph crashed because the converter always emitted block tableOfContents nodes that expect paragraph parents. When that type of document is imported, SuperDoc crashes completely.

Solution

Detect when the parent disallows block nodes and encode the TOC as a new inline tableOfContentsInline node with proper exporter/schema support plus translator tests locking encode/decode behavior. We now are loading that type of document properly, however the TOC is not 100% well formatted - but we'll tackle that separately.

[linear]

SD-2440 BUG: Import fails on field-based list of tables/figures TOC paragraphs

[github-actions]

Status: PASS

The OOXML field character handling in this PR is spec-compliant.

The inline TOC decodes to the standard complex field structure defined in ECMA-376 §17.16.2:

w:fldChar (begin) → w:instrText (preserve) → w:fldChar (separate) → [content] → w:fldChar (end)

All three pieces check out:

• w:fldCharType values begin, separate, end are the exact three valid values per §17.16.18
• xml:space="preserve" on w:instrText matches the spec's own examples (§17.16.23) — preserving whitespace in field codes like TOC \h is correct
• Field sequence is valid — separate is optional per spec, but including it (as the code does) is correct when there's result content following it

The only difference from the block variant is that the runs are returned as a flat array for injection into an existing paragraph rather than being wrapped in w:p elements. That's the right behavior for inline fields — the surrounding paragraph context is owned by the parent, not this node.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: ac0874bd59

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

[chatgpt-codex-connector]

Decode TOC children only in the active branch

In decode, tocContent is exported once into inlineContentNodes before we know whether the node is inline, and block TOCs then export the same children again into blockContentNodes (next line). This double pass is not just extra work: child decoders are stateful/mutating in this codebase (for example hyperlink decode removes link marks and can append relationships), so the second pass can serialize different output than the first and lose field content metadata in block TOCs. Move the export call inside each branch so each child is decoded exactly once.

Useful? React with 👍 / 👎.