Remove some blocks and then use standard PyMuPDF get_text functionality?

[stevesimmons]

Is your feature request related to a problem? Please describe.
I want to extract text from PDFs where some pages have text overlays that need to be ignored.

Describe the solution you'd like
I would like to scan the blocks on a page, delete (or otherwise mark as "ignore") the ones I don't want, and then use page.get_text(clip=rect) on the target regions for the remaining blocks.

Describe alternatives you've considered
Alternatively stream the elements of a page into a new empty page, filtering out the blocks I don't need. And then do the text extraction on the filtered page.

Additional context
Any suggestions would be most welcome. I would like to solve 100% of this problem with PyMuPDF because it is such a great tool.

[JorjMcKie]

The most challenging part is how you want to identify undesired text. If you can do that, no additional technique is required: simply ignore it on encounter while extracting text.
But maybe I miss some aspect of your requirement.

When you do page.get_text("dict", ...) you will be given all text with all available meta-information, from font name, ~ properties, ~ size, text color, position information, writing direction.
What else do you need to for identification?

[JorjMcKie]

Carefully looking again, I am not sure what you technically mean by text "overlays". Just text overlapping other text?

[stevesimmons]

Ah, good suggestion! I am too used to page.get_text(clip=...) magically bringing my text back in a single string.

I'll look at the dict option now and use the other properties to filter out the text I don't want. The bits I don't want are a different font size, so irrespective of the clip window I will still be able to filter them out.

Thank you