For non-page-based formats like epub, how to get the text by chapters?

[workflowsguy]

I tried to get the text of a epub book by chapter, but could not find any information on how to do that.
PyMuPdf splits chapters into pages according to a rule that is not obvious to me.

What I would like to get in the end is a list of dicts like [{'Chaptername': chaptertext}].

Is this possible?

Thanks!

[JorjMcKie]

You probably just have to iterate over the EPUB chapters:

for i in range(doc.chapter_count):
    chapter_page_count = doc.chapter_page_count(i)
    chapter_text = ""
    for j in range(chapter_page_count):
        page = doc[[(i, j)]
        chapter_text += page.get_text()
    print(f"Text of chapter {i}")
    print(chapter_text)

[workflowsguy]

Thank you for the example, @JorjMcKie . I had overlooked that the doc class provides support for chapters.

I tested this with two epub files. Unfortunately, this does not work as expected.
In PyMuPDF 1.21.1 chapter_page_countreturns an incorrect value (it seems like for chapter "n", it actually gets the page count for chapter "n+1"...)

[JorjMcKie]

If you were right, then the follow test must fail, right?

chapter_sum = 0
for i in range(doc.chapter_count):
    chapter_sum += doc.chapter_page_count(i)
assert chapter_sum == doc.page_count

[JorjMcKie]

it doesn't in my test cases.

[workflowsguy]

I think that from my tests, I have made an incorrect assumption about the relationship between "chapters" and "toc" entries.

I thought I could get the chapter text like in your example and the name of the chapter from the toc. But as I currently understand, those two items do not map to one another (i.e. chapter 1 <> toc entry 1 of level 1)

[JorjMcKie]

You are right. Chapter is a fully technical term in the EPUB's file structure. There is no reason to assume / rely on any relationship with the items in the TOC. The latter may be created independently, following other considerations.