@stesura/docx

Uploads an embedded image and returns its URL. When provided, every supported image is uploaded in parallel and the doc’s srcs point at the returned URLs, keeping base64 out of the doc (it bloats the collab payload). filename is the media entry name (e.g. “image1.png”). signal aborts when another upload failed or uploadTimeoutMs elapsed; wire it to the request where possible.

How long the whole image-upload batch may take before the remaining uploads are abandoned (fallback applies). Default 10s.

Whether an image may fall back to a base64 data URL when there is no uploadImage or its upload fails. Default true. Collab mounts pass false; such images are then dropped from the doc and raise a missing_image warning.

Replaces imported nodes outside the core vocabulary’s reach: called on every node, a non-null result takes its place (e.g. a picture tagged as a diagram back into the diagram node). Runs before images are uploaded, so a consumed picture is never uploaded.

Fills DocxImportResult.stats (one extra walk over the content parts).

Anything that could not be imported faithfully.

Comment threads. The doc’s comment_open/comment_close anchors carry each thread’s import-minted id: the caller creates the threads in its comments store and rewrites the anchors’ threadId to the store’s ids before mounting — or strips the anchors when there is no store.

With options.stats.

Every w:/m:/wp:/v: element in the content parts, by qualified name.

Of those, the ones no handler consumed: unwrapped to their children (w:sdt) or skipped. Silent drops inside a handler are not here; the export’s element delta catches those.

The id the document’s comment anchors carry — replace it with the store’s. The root comment’s durable id when the file has one, so the same thread gets the same id on every import.

The comment’s w:id in the file.

w16cid:durableId from commentsIds.xml: Word keeps it across saves, where w:id is renumbered — what identifies the same comment in a later import. Null when the file has no commentsIds part.

ISO date, or null when the file states none.

The body as comment-schema JSON (doc of paragraphs, shared marks, links as href).

The drawing’s r:embed relationship id.

Media entry name, e.g. “image1.png”.

Guessed from the extension; replaced by the type sniffed from the bytes before upload.

The src image nodes carry until importDocx rewrites it to the resolved one.

Uploaded URL or data URL; stays placeholderSrc when unresolved.

The editor listRef: the resolved abstractNumId (Word runs one counter per abstractNum), or ${abstractNumId}.${numId} for a w:num that redefines the abstract.

A fork counts on the counters of ref (its abstract’s config) and restarts the levels (0-based ilvls) it overrides on its first use (TC-213). ifShared: a fork that redefines a level needs the base only when another list counts on that abstract too.

w:ilvl, 0-based.

w:lvlText pattern, e.g. “%1.%2”.

w:suff. Absent = tab.

w:lvlJc. Absent = start (Word’s left).

w:isLgl: this level’s text renders every number as a decimal.

Px from the number’s start, from the level’s explicit w:tabs and w:ind. Absent without an explicit tab.

w:lvl/w:pPr/w:ind, px: the lowest tier of the indent cascade, which most numbered paragraphs in real files rely on.

w:lvl/w:pStyle: the paragraph style linked to this level. Imported as the style’s numbering binding; the style’s own w:numPr wins on conflict.

Content the serializer skipped or degraded — surface these to the user.

warnings with a code and the node each came from.

Thread structure per comment id, consumed by the post-pack stage.

Per-section text direction in document order, consumed by the post-pack stage.

Per-section watermark in document order, written by the post-pack stage.

Forks’ w:nums, repointed by the post-pack stage.

Lists’ own follow-bys, written by the post-pack stage.

Pictures’ crop, effects and DOCX passthrough, written by the post-pack stage.

Sources and our bibliography record as custom XML, written by the post-pack stage.

Each redaction written as a content control, by id: our custom XML part, written by the post-pack stage.

Pictures’ redacted areas by picture key, in the same part.

The theme part and the theme refs docx-js cannot write, for the post-pack stage.

The exported document’s attrs, after transformDoc: the post-pack stage writes its properties from these, never from the source state’s.

Display name for a mentioned user id. Optional: authors named in the same threads are the fallback, then the raw id.

True for every comment after the first in its thread — drives commentsExtended.xml reply linking.

Thread resolved state — becomes w15:done in commentsExtended.xml.

Handlers merged over the defaults (same name replaces).

Rewrites the document before anything reads it (preload, pre-passes, rendering), e.g. to resolve content that must not reach Word. Same schema.

Bytes added to the image preload pass, keyed as handlers will ask for them through state.options.getImageBuffer. Runs alongside preloadImages.

Filled with where the export’s time and bytes went (createExportTrace()).

The saving user’s name, stamped as cp:lastModifiedBy as Word does on save.

Leaves the document’s properties (title, author, custom properties) out of the file, as Word’s Inspect Document → “Document Properties” does: kept otherwise, they reach whoever receives it (SEC-98).

Labels and colors of redaction reasons (the host’s taxonomy), for the content controls’ titles and boxes.

Child indexes from the rendered root (the doc, a note body, a header) to the node.

Types from that root down to the node.

The node’s first 60 characters of text.

Each post-pack injector and validator.

The final file.

Each XML part of the final file: size and element count.

Render self-time per node type (children excluded), slowest first.

OPC part the violation was found in, e.g. word/document.xml.

Rule id, e.g. foreign-element.

What is wrong, in one clause.

Whitespace-collapsed XML around the first offending offset.