mysterious case of lost double Ls - and other letters

[cybmole]

I have a pdf which is not converting properly to epub.

e.g. PDF source has the name Ellen,but in th epub it has become El en.

I cannot see why

 <p class="calibre3">One flight down, on the fourth landing, ear pressed against the door of 4B, El en Swenson clasped a hand over her mouth, suppressing the urge to cal out to her husband, Mike, who dozed sporadically in their apartment, behind their currently unlocked door. El en had left

some instances of Ellen are fine, but there are several like the above.

I cannot directly paste source text as it appears as an image on clip board.

is calibre attempting OCR and glitching on the double L's ? is there a workaround ?

another example - from the calibre epub viewer - see all the double L issues, whih are not in the pdf source:

Safely on the other side of the stairwel housing, Ruth tilted her head up and let the cataract wash over her cataracts. She’d been scheduled to have phacoemulsification the week after martial law was declared. Now she was stuck with cloudy vision of a cloudy sky. She pul ed some matted strands of hair away from her eyes, her fingers straying up her forehead, which seemed to go al the way to the back of her head. Maybe it was better she couldn’t see that wel . In her mind she could stil picture herself as she was. Abe, too.

fixing this with regex would be a nightmare

[Manichean]

This is called ligatures. It's a typographical feature. A simple forum search should turn up multiple threads offering solutions on how to deal with them.

[cybmole]

Originally Posted by Manichean

This is called ligatures. It's a typographical feature. A simple forum search should turn up multiple threads offering solutions on how to deal with them.

Does not compute - the name Ellen occurs hundreds of times in the text, only about 1 % of the time is is transformed to El en by the conversion. surely the word is typeset or whatever the same way every time.
I will go google & search on ligatures as you are often annoyingly correct. meantime I've sent the file to amazon to send to my kindle to see how their conversion software gets on

I am also running it thru pdftoepub as I type - that program shows a table of "pdf glyphs" and what it plans to do with them, before it gets going, I did not see any funnies in there

PS - quote - Koval - last May - "the next release of calibre will automatically convert ligatures to normal characters."

[cybmole]

OK - so I ran it through pfdtoepub, opened in sigil , searched for El en - no hits, searched for text from the above samples - they have converted OK.


do I file a calibre bug ? the pdf book is 1.5MB -not sure if that will attach to bug report

[Manichean]

Originally Posted by cybmole

Does not compute - the name Ellen occurs hundreds of times in the text, only about 1 % of the time is is transformed to El en by the conversion. surely the word is typeset or whatever the same way every time.
I will go google & search on ligatures as you are often annoyingly correct. meantime I've sent the file to amazon to send to my kindle to see how their conversion software gets on

Well, excuse me for saying so, but those are, as I said, ligatures. Look it up. I'm not that firm on typography, but I imagine your document is either badly laid out or it is not usual to contract every possible occurence into a ligature.
Also, only because Kovid said that Calibre would automatically convert those, doesn't mean it does that absolutely correct in every case- maybe you found a bug. Or, maybe, you ticked "keep ligatures" in the layout options.
Anyway, I'm going to shut up now before I annoy you further. Good luck in solving your issue.

[DoctorOhh]

Originally Posted by cybmole

PS - quote - Koval - last May - "the next release of calibre will automatically convert ligatures to normal characters."

Did you check the box marked Keep ligatures under look and feel during conversion? This doesn't always work but it is worth a shot.

Originally Posted by cybmole

do I file a calibre bug ? the pdf book is 1.5MB -not sure if that will attach to bug report

You can file a bug report and attach the file, otherwise it will get lost and never have a chance of being corrected.

[cybmole]

Originally Posted by Manichean

Well, excuse me for saying so, but those are, as I said, ligatures. Look it up. I'm not that firm on typography, but I imagine your document is either badly laid out or it is not usual to contract every possible occurence into a ligature.
Also, only because Kovid said that Calibre would automatically convert those, doesn't mean it does that absolutely correct in every case- maybe you found a bug. Or, maybe, you ticked "keep ligatures" in the layout options.
Anyway, I'm going to shut up now before I annoy you further. Good luck in solving your issue.

OK
i did not tick keep ligatures -
i did look stuff up
- I do have a good conversion via the other program.
your advice was helpful, if brusque.
I now know what a ligature is.

suspect it's combination of poor quality source + maybe a residual bug in calibre. I'll file a bug. - ticket 8459

[cybmole]

ps "keep ligatures" box

the default is not ticked
I take that to mean " replace ligatures with normal character pairs".

that seems the right choice for maximum compatibility with other ereader hardware & software ?

I'm done experimenting for now , I want to read go read the book on Kindle
but googling ligature ll
suggests a wrong diagnosis above

[Jabby]

Originally Posted by cybmole

ps "keep ligatures" box

the default is not ticked
I take that to mean " replace ligatures with normal character pairs".

that seems the right choice for maximum compatibility with other ereader hardware & software ?

I'm done experimenting for now , I want to read go read the book on Kindle
but googling ligature ll
suggests a wrong diagnosis above

There might be a problem with the calibre conversion to mobi. I just had to drop back to 7.40 from 7.42 to get a clean epub to mobi conversion.

In my case the header wasn't centered and the font size was out of whack. As soon as I dropped back, all was well.

Suggest you drop back to 7.40 and try converting again to see what happens.

Regards -John

[DoctorOhh]

Originally Posted by Jabby

There might be a problem with the calibre conversion to mobi. I just had to drop back to 7.40 from 7.42 to get a clean epub to mobi conversion.

It is always nice to reach out and lend a hand and I hope you keep on helping others. This thread though was specifically about a known problem that has been around forever when converting some PDFs to anything else. This thread hasn't even mentioned converting books to Mobi.

I would suggest that you please create a ticket with the info that you have gathered so this mobi conversion bug can be corrected. Please attach to the ticket the epub, the resultant mobi file and the job details available by clicking the jobs icon in the lower right area of calibre. This will be of immense help to everyone who is using calibre to convert to mobi.

[alexandroid]

The bug for Calibre was posted a long time ago - http://bugs.calibre-ebook.com/ticket/1366 and it is waiting for a fix in underlying pdf2html engine Calibre uses. The bug for PDF2HTML also old and sits there since 2008: https://sourceforge.net/tracker/?fun...39&atid=444239