is it possible to iterate text and images in a top-bottom order?

[allenling]

we found that we could easily extract text content and images respectively by pypdfium2, and we now need to iterate content one by one.

suppose we have a pdf file containing text and an image like this in a page.

abcdeft. <- this some words.

img <- this is an image.

we wonder if we could just do something as below:

content_list = []
for page in pages:
    for content in page.iterate_content():
        if content is image_type:
            url = upload_image(content)
            content_list.append(url)
        else:
            content_list.append(content)       

[mara004]

Try sorting the objects by position.

[allenling]

so basically what we did was to call page.get_objects to get a list of pdf objects in that page, and then sort them by their positions (call PdfObject.get_pos).

pdf = pypdfium2.PdfDocument(file)
content = []
for page_num, each in enumerate(pdf):
    x = list(each.get_objects())  # x is the list of pdf objects.
    x = sorted(x, key=lambda n: n.get_pos())  # sort the list of objects by position.
    for pdf_obj in x:
        pass  # do something!
    #
    content.append(each.get_textpage().get_text_bounded())

as you can see, we can extract all the text content on the page by get_text_bounded, and we really want to iterate the objects list and process them one by one. (maybe we should use the term bbox or block to refer to each content chunk on a pdf page.)

but it seems that there's no such an attribute or method for us to use to get or access the content of each pdf object, the following pic shows the possible attributes.

and since we are not familiar with pypdfium C API, would you please give us some advice?

edit by mara004: syntax highlighting, formatting

[mara004]

I'm not sure using the .get_pos() 4-value tuple as-is will yield the right results. You might want to sort by just (top, left) or something.

In the debugger, you seem to be inspecting the .parent.raw attribute, which is unrelated. You won't find any pageobject methods there. Look at PdfObject layer instead.

If what you want to do is get the individual text segments, you can either do textpage.get_text_bounded(*pdf_obj.get_pos()) with the existing helper on any text objects (ideally factoring out the .get_pos() call to cache positions), or you could use the raw API FPDFTextObj_GetText() as mentioned in #325 (comment). The Raw API section has examples how to work with string output APIs. Probably we should add a helper for this in the future.

Alternatively, you could also leverage the text rects API.