Le 28/01/2013 11:45, Jürgen Spitzmüller a écrit :
I understand that a compose char is a compose char. But how many compose chars are there over the whole unicode range? My point is: If Qt already knows this, and if we can catch this information, why should we hardcode a list of code points ourselves? (Note that I do not know whether Qt _really_ provides this information).
This is exactly what I had in mind, actually.
Also, can you remember me why we paint char by char at all? If I set \force_paint_single_char to false, the ligatures automatically get painted correctly.
We compute metrics char-by-char but by default we draw text at word level. This is stupid and causes problems in particular on OS X where there seem to be more ligatures, even in plain ascii text.
One of the things I would like to do in Milano (if it proves possible) is to implement word-level metrics computation, at least for LtR text.
It would be great to get rid of our Bidi stuff too, but I will not try anything there, I can't read any of these languages.
JMarc
