fromArchive crashes with Error: XPDY0002: context is absent, it needs to be present to use the "." operator

[pbuszka]

I triaged the problem to the following summary but not sure what is the correct way to fix it.

If the document.xml contains image that has an anchor it crashes

        <w:drawing>
          <wp:anchor allowOverlap="1" behindDoc="0" distB="0" distT="0" distL="114300"

if everything stays the same and anchor is changed to inline it passes

        <w:drawing>
          <wp:inline distB="114300" distT="114300" distL="114300" distR="114300">

it comes down to this line in fromNode for Image class

const inlineNode = evaluateXPathToFirstNode(`./${QNS.wp}inline`, node);

if I change it to

const inlineNode = evaluateXPathToFirstNode(`./${QNS.wp}anchor`, node);

Brute force "solution" doesn't crash but I presume it doesn't work as intended for anchor use case

    let inlineNode
    try { 
      inlineNode = evaluateXPathToFirstNode(`./${QNS.wp}inline`, node);
    } catch (error) {
      inlineNode = evaluateXPathToFirstNode(`./${QNS.wp}anchor`, node);
    }

Below two DOCX files that reproduce the problem (anchor and inline version of the same image), document.xml is the sme except this single change.

describe('docxml visitor debug', () => {
  it('debug docx', async () => {
    const archive_inline = await fs.readFile(path.join(__dirname, '.data/debug_min_inline.docx'))
    const docx_inline = await Docx.fromArchive(archive_inline)

    // below it crashes
    const archive_anchor = await fs.readFile(path.join(__dirname, '.data/debug_min_anchor.docx'))
    const docx_anchor = await Docx.fromArchive(archive_anchor)
  })
})

debug_min_anchor.docx
debug_min_inline.docx

[harvestcore]

Hi @pbuszka!

Apologies for the late reply, we somehow missed this issue.

Thanks for reporting it and for the debugging. We're taking a deeper look at this and will update this thread as soon as we know more.

Regards,
Ángel

[harvestcore]

After looking into it more closely, here’s what’s going on:

• The error XPDY0002: context is absent, it needs to be present to use the "." operator is triggered (as @pbuszka noted) by the call const inlineNode = evaluateXPathToFirstNode(./${QNS.wp}inline, node). Since that element doesn’t exist in .docx files that use this type of image, the following XPath queries that rely on inlineNode fail as well, because inlineNode becomes the context for those calls.
• Docxml currently doesn’t support floating images. The <w:drawing> node can contain two possible children (wp:inline and wp:anchor), and we only support wp:inline at the moment.
• Your fix is on the right track, but still incomplete. Even with a try–catch block, the next queries are likely to fail because the child nodes under wp:inline differ from those under wp:anchor.
• Each image type needs to be handled and processed separately.

We plan to add support for floating images in the short to mid term, so you can wait for the feature or open a PR to help us bring it in sooner!