add method to read text boxes

Feature request:

Document#getTextBoxes()

it must:

• Get all text from text boxes in the document

it should:

• Return the text boxes in the same order as they appear in the document

it could:

• Somehow try to reproduce the document text order combining the content from getBody() with the text box content. (maybe as a separate Document#getBody({includeTextBoxes: true}). I imagine this is insanely difficult, but just putting it out there 😃

[thegoatherder]

@morungos my apologies for the delay on this. Please see this test document which I hope will be useful. Note that I intentionally created the 3rd textbox after creating the 4th textbox (but positioned it above the 4th visually).

The expected test output (if we remove all blank lines from the output) would be:

This is in the header
This is the first paragraph
This is the first textbox
This is the second textbox
This is the second paragraph
This is the third paragraph
This is the third text box
This is the fourth text box
This is the last paragraph
This is in the footer

[morungos]

Hi @thegoatherder

Many many thanks for this feedback. There might be a challenge here, so here’s what’s happening inside. Word has two things going on:

1. Character positions, or CPs, which are essentially the linear character position of the text box “anchor”, where 1 is the first character, and so on. Almost all Word data is sorted by character position, and I believe (and did test) that text boxes are extracted in order of character position. So I’d be surprised, but it’s possible.
2. Visual positions, which are attached using some weird metadata, and more or less say where the X and Y coordinates of the text box are supposed to be. These matter when a text box has the “same page” type display, rather than a strictly in-line display.

The problem is that visual positions can show differently than character positions. (There is some option somewhere which allows you to see where the anchor is when you click on a text box). It’s very easy to simply drag one text box above another even if the character position order is the other way around.

This is another of the issues where if it is visual positions you want to order by, it can be tricky, because page breaks can get in the way. If there’s a page break between (say) boxes A and B, then A appears before B due to different pages, but if they’re on the same page, B appears above A. I remember this hell from trying to typeset long documents with multiple images per page. Adding code to sort the boxes might not be reliable. However, adding code to expose the coordinates of boxes will be much simpler and easier to do, if that works for you.

I mean, I might be wrong and you found a case where the code does character positions wrongly, but I suspect you might be after text boxes ordered by visual position. If so, I’d very much like a test file I can use to make sure I get it right. In fact, if you can send me a test file which shows me what you need, then I’ll be happy to add it and make it work, especially if I can add it to the open repository as a test file.

Let me know either way and I’ll see what I can do.