Questions about KF8/AZW3/Mobi8 from/to conversion

[pirl8]

Hi,

1. I've tried to convert from AZW3 to ePub a sample downloaded from Amazon (search on amazon com kindle store "relatività esposizione divulgativa"). The sample is actually very small. It cointains some font (DejaVuSerif). When these files are estracted, they are very small too, actually much smaller than the original TTF files. And it looks like they don't work in the generates ePub. Any clue?
2. Does the AZW3 exporter embed fonts within the generated file?
3. It' very sad that AZW3 does not have hyphenation support jet. By the way it does support the soft hyphen (& shy; ) HTML entity (and the <wbr/> tag, too).
   I've manually patched the text around a shaped div and it works well (see the small attached sample).
   May be this is a stupid idea but... Would it be possible to create a filter to automatically add a & shy; in the right hyphen positions to the generated AZW3? The filter should parse the XHTML file, apply an hyphenation algorithm according to the language of the book just adding & shy; and the kindle would then be able to render it.

[kovidgoyal]

1. no idea, attach the sample

2. yes

3. I have no interest in this, patches are welcome. Though calibre will never do this by default as it can lead to problems when rendering/searching for text and so on.

[pirl8]

Originally Posted by kovidgoyal

1. no idea, attach the sample

Ok. It's here.

Though calibre will never do this by default as it can lead to problems when rendering/searching for text and so on.

Maybe not for rendering but almost certainly for searching: i didn't think about it.

[kovidgoyal]

I converted your attachment to epub and the resulting epub had the fonts declared in page_style1.css and the font files embedded correctly.

[pirl8]

Yes, and actually the Calibre reader render those fonts correctly. There must be something wrong with my PC at home...

BTW, they are actually very small (they must be small also in the original AZW3, obviously) and Windows XP cannot open them as TTF fonts. Maybe they're coded in some strange way, stripping away most of the unused characters? They could be very useful to produce downsized eBooks.

Thank you very much.

[kovidgoyal]

They are probably subsetted, do a bit of googling you will find tools that subset fonts.

[pirl8]

Did a simple search test on my x.rar sample above: it works. It seems like the kindle search algorithm is ignoring soft hyphens. Maybe soft hyphens are a bulky but viable way to have hyphenation on the kindlem, after all.

I wish I could write a plugin to implement such a feature... Where could I start?

[kovidgoyal]

You can use the tweak book framework in calibre, which can explose KF8 books into html and recompile them. Using that you can write a postprocess filetype plugin to do whatever you want to the output of the calibre conversion pipeline.

[pirl8]

Ok, following your directions i've find where to start looking on calibre site.

I've never programmed Python and this could be a good opportunity to begin. I've given a quick look around and i see there are python classes for parsing (x)html and to implement hyphenation, so it shouldn't be that hard.

My idea is to do something very simple: parsing html, adding softhypens to text according to the book language, saving.

Thank you very much for Calibre. I love it.

P.S.

Is a plugin doing something similar already there so I can learn from it?

[kovidgoyal]

It might be simpler for you to just modify the code in mobi_output.py to do what you want. You can worry about creating a plugin later. See the user manual for instruction on setting up a calibre development environment.

If you want to go the plugin route look at the modify epub plugin, that should be a good starting point.

[newman]

Kovid: I found a python-pyphen package which does what's needed for hyphenation. So far it seems to me as easy as: http://pastebin.com/3tSEsD13

Where do I start? Looking at calibre/src/calibre/gui2/convert/mobi_output.py it looks to me as a building block for GUI not a part of a code meant to fiddle with ebook text. Also I guess I have to alter the text when converting to AZW3 not MOBI, right?

[kovidgoyal]

Umm there is alreadya hyphenation plugin for calibre called Hyphenate This

[newman]

Thanks! It works.