[poppler] [PATCH] Selection improvements

[Albert Astals Cid]

A Divendres, 7 d'octubre de 2011, Andreas Becker vàreu escriure:
> > Not sure if you are sending and old patch again or you are a new person
> > doing
> > the same mistake, but we have discussed this a few times, starting a new
> > word
> > when the font changes is wrong.
> > 
> > As an example of why it is wrong your patch changes the output of
> > pdftotext in
> > http://wooster.hut.fi/kurssit/s88211/s88211MatlabProject.pdf
> > from
> > uses a channel model described in [4]
> > to
> > uses a channel model described in [4 ]
> 
> I am a new person.
> If only the space after "[4" is a problem, I can fix it using the variable
> spaceAfter (TextOutputDev.cc).
> 
> Another possibility is to store a GfxFont* array in TextWord, so that every
> letter has its own font.
> I tested this and it works fine, but the rest of poppler still acts as if
> one word had only one font. A word would have
> a "main font" and letters with individual fonts. It is not a very clean
> solution, but probably better
> than starting new words.

To be honest, I don't have much knowledge of that code, it is just that fixing 
one thing to break another doesn't seem a good solution :D So if you have a 
patch that does not give regressions in pdftotext and improves the thing you 
want to improve, I am happy.

Albert

P.S: Depending on how big the change is we might not commit it straight away 
since we are trying to merge everything from xpdf-3.03 release before doing 
big changes in the core code.