Ordering annot.vertices naturally

[Don-Yin]

Hi all,

Is there a native way to modify annot.vertices for highlights in documents with two columns? I found that annot.vertices seem to return a top-down ordered list of coordinates by default, which is not great for cross-column highlights, e.g.:

Is there a native way to modify annot.vertices for highlights in two-column documents? annot.vertices appears to return a top-down ordered list of coordinates by default, which is inconvenient for cross-column highlights, this one for example:

In the image above, area 1 and 2 belong to the same highlight.

While I am trying to get the rects for each individual 'highlight bars', by doing this:

points = annot.vertices[i * 4 : i * 4 + 4]
quad_count = int(len(annot.vertices) / 4)
for i in range(quad_count):
        points = annot.vertices[i * 4 : i * 4 + 4] 
rect = fitz.Quad(points).rect

These bars look like this (using get_pixmap):
It first returns area 2:

And area 1:

This order is the opposite of what I had hoped for. Again, the issue is that the returning order of vertices appears to have ignored the columns and worked from top to bottom - which is understandable. I understand that this could simply be how highlight is coded.

But still, let me ask: is there a fitz native way to take PDF columns into account for the vertices? I'm considering writing a clustering algorithm, but that seems like a lot of work which I'd like to avoid if there is a easier way.

I'd appreciate any advice.
Don

[JorjMcKie]

No there is not. Consult the PDF spec to see how this information is stored in a PDF file.

[JorjMcKie]

Also note that you cannot rely on any order for annotations when retrieving them from page.annots().
Or rather: they are retrieved in the same order as they have been stored by the creator, which is the order as the annotation xrefs are stored in the array /Annots present in the page definition. You can access this array via doc.xref_get_key(page.xref, "Annots"), or, more reliably(*), via page.get_annot_xrefs().

(*) the second method automatically resolves any indirect references, whereas the first one may return another separate xref, which then contains the array.