Abysmal performance with a particular file

[Fish-Face]

Hi all,

I wanted to use Calibre to convert a dictionary I have in .mobi format to a .epub, so that I could write a program to display entries in the dictionary on my computer. Unfortunately, Calibre completely chokes on the file. It's 6.3MB in size, and loading it in the viewer takes an age and consumes about 1.4GB of memory. Pressing "next page" takes memory consumption up to 1.75GB and spins a CPU core for an indeterminate length of time (I've been waiting about 20 minutes.) Obviously this is pointless: there is no way to use an ebook in this fashion. Conversion is a worse story: Calibre consumes all available memory, almost all available swap (I have 8GB between them) and sits, thrashing my disk, for ages. I have no idea if it would eventually finish, or eventually display the first page of the dictionary after the cover (you know, it's probably blank...)

Meanwhile the Kindle viewer, under wine no less, is capable of displaying the file while using a mere 200MB. Clearly there is something very wrong with Calibre's mobi reading algorithm that is making it choke. I imagine, since it is a dictionary, it must be either all the formatting (lots of bold and italics) or the links and index used for searching.

Does anyone know any workarounds for this kind of situation? I mainly want to convert it, so a lightweight tool of some kind for this purpose would be ideal - preferably on Linux. I've tried the online converters but they're even less patient than I am, and just give up after about 10 minutes.

The book in question is not free, but it is DRM-free. I could upload it and give the link to people via PM if that would help (and they should probably promise to delete it after examining it.)

[eschwartz]

calibre does not actually have a MOBI reading algorithm.
The Viewer only supports EPUB, and all other formats are converted on the fly to EPUB for viewing (well, technically I believe just the first step of a conversion, which is unpacking it and transforming it into an OEB model).

It is for this reason that EPUB is the preferred Viewer format -- no need to pre-convert the book. And books which are very slow are recommended to be converted to EPUB to speed up future viewing.


And dictionaries contain a lot of complex internal markup, which makes the conversion have to work a lot harder.
And dictionaries aren't supported in calibre anyway, so all that markup gets dropped. If you leave it for long enough (and complex conversions, which any dictionary certainly qualifies as, have been known to take up to 48 hours to finish) it will result in an EPUB that reads like an ordinary book, no special dictionary formatting at all so no semantically defined entries.