Regular Expression Help

[smartmart]

Hi, i've converted a pdf to awz with the Amazon service, now i want convert it to mobi with Calibre (so i can add metadata and TOC).

I've a problem with chapter recognition, every chapter start with "Chapter XXX." so my regex is:
//*[re:test(., "chapter", "i")]

It works with the original pdf but not with the awz.
it matchs the word "chapter" in the text (sometimes there is the word "chapter" in the script) but it doesn't match the real chapters.

So i've saved the debug and i've seen that the chapters are not in a html tag (the text anyway is child of the body tag ofcourse):
<p> foo foo foo</p>CHAPTER 1<p> foo foo foo </p>

Is this the problem?
How can i fix it?

Thx

[desertgrandma]

Welcome to MobileRead, smartmart

Help should be arriving soon.

[BookGnome]

Originally Posted by smartmart

So i've saved the debug and i've seen that the chapters are not in a html tag (the text anyway is child of the body tag ofcourse):
<p> foo foo foo</p>CHAPTER 1<p> foo foo foo </p>

I'm not sure how you need to specify it with Calibre's custom syntax, but your regex itself is flawed. Here's a working regex in Python:

>>> import re
>>> myString = '<p> foo foo foo</p>CHAPTER 1<p> foo foo foo </p>'
>>> re.findall('Chapter \d+', myString, re.I)
['CHAPTER 1']

A lot depends on how consistent the input file is, but this should catch any instance of the word 'chapter' followed by one or more numbers, without regard to case. How to wrap that in Calibre's regex DSL is a question for the Calibre gurus.

[ldolse]

I'm terrible with xpath, but I have a hunch you're screwed trying to search for text just free floating throughout the book in the body tag.

You're best bet is to take the html from debug info and do a find replace in a text editor with regex search/replace support.

Search for this:

(Chapter\s+\d+)

and replace it with this:

<h2>\1</h2>

Depending on the editor you use it might be $(1), or $1, or whatever instead of '\1' as I used above - check the documentation for your editor.

Then import the edited html file to Calibre, and have Calibre convert using the zipped html source instead of the pdf. Calibre's default chapter detection xpath will automatically pick the chapters up if your search and replace properly wrapped the html in <h2> tags.

[smartmart]

Originally Posted by BookGnome

I'm not sure how you need to specify it with Calibre's custom syntax, but your regex itself is flawed. Here's a working regex in Python:

>>> import re
>>> myString = '<p> foo foo foo</p>CHAPTER 1<p> foo foo foo </p>'
>>> re.findall('Chapter \d+', myString, re.I)
['CHAPTER 1']

A lot depends on how consistent the input file is, but this should catch any instance of the word 'chapter' followed by one or more numbers, without regard to case. How to wrap that in Calibre's regex DSL is a question for the Calibre gurus.

I know, i've used a wide regex only for testing purpose

Thx Idosle, but i'm searching for a solution in calibre (if it's possible) so i can use the setting every time.

PS: i don't use the pdf to mobi from calibre because it fails with the wrap.
It seems that every page of the pdf is a paragraph.

[ldolse]

There is no solution in Calibre to do it the way you're trying to do it. Amazon is creating a really screwy mobi file, Calibre hasn't been programmed to handle that scenario, and It's unlikely to happen anytime soon.

If you're seeing that Calibre isn't unwrapping the lines when you use it to convert from pdf to mobi it means that your line unwrapping factor under pdf input is incorrectly set. It seems to be set to zero on a number of user's systems. Set the line unwrapping factor to 0.45, this is the default and generally provides the best results.