1.24.6&1.24.13 get_text different

[tenghian]

Extraction of blocks in version 1.24.6 is perfect. How can I make version 1.24.13 work like 1.24.6? Thank you!

text_blocks = page.get_text("dict", flags=pymupdf.TEXTFLAGS_BLOCKS)["blocks"]

pdf1246-cut.pdf
pdf12413-cut.pdf

[JorjMcKie]

You cannot - except with your own code of course.

PyMuPDF is not deciding about block segmentation, this is a result of MuPDF's algorithms. The next MuPDF version 1.25.0 will bring significant improvements here. With a new text extraction option, MuPDF can be asked to search for recognizable page layout segments which will each be turned into a block for PyMuPDF.
That should bring back some of the earlier results.
In yet a subsequent version, MuPDF will also add guessing correct paragraph breaks and other structures.
But this is more in the future.