Detect chapter headings with capitalized words

[fiendmish]

I have a file i am trying to convert from PDF to epub, I am having trouble with the chapter detection as the chapter headings in the PDF read something like 1 CHAPTER NAME, 2 CHAPTER NAME etc ...

When i export to epub i end up with a continuos stream of text as it doesn't pick up those as chapter breaks.

Can anyone suggest as way i can get calibre to detect chapters based on that convention of number then capitalized text?

Thanks

Sam

[Stinger]

Someone might suggest a fancy regex expression that might work, but filtering only by numbers is pretty broad and could yield a bunch of entries within the body text.

How I've handled this situation is opening the resulting epub file in Sigil, and manually going to each chapter and changing each chapter heading to 'Heading #' style. Sigil will update the TOC automatically if you do this.

[fiendmish]

Thanks Stinger, there are 108 of them so i was hoping there may be an automated way but will check out your suggestion in sigil

Cheers

[Stinger]

Yeah, manually changing 108 headings is definitely something I would want to avoid too...

You might be able to do this with the Calibre TOC autocreate system using a nice XPATH expression, but again, that is easier if you have a book with 'Chapter X' style so you can tell it to look for the word "chapter" within a certain tag. The way you described your book made me think this any XPATH you used would yield lots of junk entries.

On the other hand, it could take less time to take this route with the large number of chapters you have, and then just delete the junk entries with Sigil, rather then what I suggested before.

Can you post the epub here without violating copyright so I can take a look at the HTML structure?

[wallcraft]

I have used a modified version of the standard Chapters and page breaks expression:

//*[((name()='h1' or name()='h2') and re:test(., 'chapter|book|section|part|0|1|2|3|4|5|6|7|8|9\s+', 'i')) or @class = 'chapter']

This just keys on any number in an h1 or h2, and, as Stinger suggests, this may not be best for all ebooks. So I only use the above when the standard version fails. This probably isn't the cleanest way to test for digits, but it works.

[fiendmish]

Thanks wallcraft will give that a go now.

Cheers
Sam

[pietvo]

You could try something like:

//*[re:test(., '^\s*[0-9]+\s+[A-Z ]+\s*$')]

but we would need to see the actual HTML around the chapter titles to be sure. Notice the space after the Z.