[BUG]: Confluence import drops code blocks whose macro lists ac:macro-id before ac:name

How are you running AnythingLLM?

Local development

What happened?

The Confluence connector drops the code from a code block when the page stores the code macro with ac:macro-id before ac:name. The text around the block is kept, along with the text of any macro parameters such as a language name, so the imported page shows a bare bash where the code was, or nothing at all when the macro has no parameters. The code should come through as a fenced block, as it does when ac:name comes first.

createDocumentFromPage in collector/utils/extensions/Confluence/ConfluenceLoader/index.js converts code macros with a regex that starts with <ac:structured-macro ac:name="code" (line 111), so it only matches when ac:name is the first attribute. Any other order falls through to html-to-text, which drops the CDATA body.

On master a3557034 I ran createDocumentFromPage on this page body, then again with ac:macro-id="7f3c" moved to the front of the macro tag:

<p>Install with:</p><ac:structured-macro ac:name="code" ac:schema-version="1" ac:macro-id="7f3c"><ac:parameter ac:name="language">bash</ac:parameter><ac:plain-text-body><![CDATA[npm install example]]></ac:plain-text-body></ac:structured-macro>

ac:name first     pageContent="Install with:\n\n```bash\nnpm install example\n```\n"
ac:macro-id first pageContent="Install with:\n\nbash"

Real pages are stored this way. In a saved response from https://cwiki.apache.org/confluence/rest/api/content?spaceKey=KAFKA&expand=body.storage,version (the first 25 pages), all 120 code macros put ac:macro-id first, and replaying those page bodies through the loader on master a3557034 produces no code blocks.

This matches the lost code snippets reported in #1501. #1804 closed that issue with the code block conversion, but its regex expects ac:name first.

Are there known steps to reproduce?

1. On a self-hosted Confluence instance, find a page whose code macro puts ac:macro-id before ac:name. The content REST API with expand=body.storage shows the stored markup, which is what the connector reads.
2. In AnythingLLM, open Data Connectors, choose Confluence, select Self-hosted, enter the instance URL, space key and credentials, and submit the form.
3. Open the saved document JSON for that page in the output folder the connector reports. Its pageContent has the surrounding text but not the code. If the macro has a language parameter, its name appears where the code should be.

LLM Provider & Model (if applicable)

Not applicable.

Embedder Provider & Model (if applicable)

Not applicable.