Customizing a PDF to EPUB

[jwa]

I am using a Sony PRS-T1 to read Bruce Eckel's Thinking in Java which is a PDF and is a pain in the butt to read on the Sony. I would like to read it as an EPUB but the standard conversion doesn't handle the embedded Java samples. These are nicely delimited by lines containing '//:' and '//:~'. Several of Bruce's books are available in PDF or HTML but EPUB would be nice.

This should be easily implemented in Calibre but the documentation is a bit confusing. I can do this myself if I can find my way around Calibre. Is there anything more informative than the 'manual'? I have found the code in the 'debug' data written to the 'input' directory. I have done a lot of XSLT (some years ago) and can this for myself but I want to minimize the 1st derivative of the learning curve. (I accept that it must be positive).

Where are the XSLT transformations that are applied and where can I insert some customization?

I know that PDF is nontrivial but most of the work is done. All I need is to tag certain parts of the XHTML for a monospaced font. I see how the existing stuff is marked for CSS, I just need to tailor my own.

Thanks in advance.

I have all ready cropped the PDF but it is still a pain to use in PDF.

[kovidgoyal]

calibre does not use XSLT.

[jackie_w]

If you can find your book as an html source I think it should convert to epub quite well. It can be perfected in Sigil afterwards if necessary.

I've done quite a lot of PDF to epub conversions, using various techniques, but I draw the line at technical manuals. Just too much manual effort IMO.

[jwa]

This is not a technical manual. It is a programming book and it is a very good one. It is readable as a PDF just painfully slow on the Sony reader. Bruce has several other books that contain samples of code and I would like to contribute by simplifying conversion of the books to EPUB.

I suppose I could save the INPUT files and fix them with a Perl program and feed them back into Calibre but that is not a solution that I would give back to the author. Hopefully, he might sell a few more books if they are available in EPUB format. He has given much away and I intend to buy some stuff from him but saving labour is also a way to say thanks.

[JSWolf]

It cannot be converted without errors. You would have to pick a program to use to convert and then A/B compare to make sure it's correct. A single error in the code samples in the book and they won't work. To be honest, it's not worth the time/effort to do this.

[jwa]

The code fragments in these books are delimited with //: and //:- and there are no errors because these markers are used to verify that the code works using other production tools. I suspect that the C++ books use a similar technique.

There are other code fragments in-line but these do not concern me for now. I don't want to discuss the merits, I just want to be able to grok the calibre design enough to tinker on my own.

I thought that the use of XPath implied that there would be XSLT in the black box. I guess I was mistaken to assume that.

I am reading platform.py at the moment as the existing manual seems to suffer from being written by someone who understands the entire context all ready. That is not intended in a negative way as I hope to be constructive here. It is hard to distance yourself enough from internals to explain things succinctly. Good prose is very hard to write. Good technical prose is almost impossible. That is the thing I like most about Eckel's writing.

[jwa]

So I read through that code and established that calibre doesn't do the conversion. The conversion is done by pdftohtml.

[PeterT]

I did find an HTML version of the book at the authors website, albeit not the current version.