Docx RTF image parsing must not match shp inside shppict

Problem

getRtfImagesMap(...) could invent duplicate image entries while parsing RTF image data.

A \shppict block was being parsed once as a shppict image and then again as a shp image, which produced bogus s... image ids for content that only existed as i... images.

Root Cause

getRtfImagesByType(...) used rtf.split(type).

That is too loose for RTF control words because \shp is a prefix of \shppict. Splitting on \shp matched both the real \shp token and the start of every \shppict token.

Solution

Split on the exact RTF control word, not a raw substring:

const escapedType = type.replace(/[.*+?^${}()|[\]\\]/g, String.raw`\$&`);
const [, ...images] = rtf.split(new RegExp(`${escapedType}(?![a-zA-Z])`));

That negative lookahead keeps \shp from matching \shppict, while still matching the real standalone control word.

Prevention

• When parsing RTF control words, match whole tokens instead of raw prefixes.
• If one control word is a prefix of another, add one test that proves the shorter token does not steal the longer one.
• For image-map helpers, always include at least one mixed \shppict plus \shp fixture so duplicate parsing cannot hide.