[reportlab-users] ReportLab vs LaTex

François Pinard pinard at iro.umontreal.ca
Sat Dec 30 10:10:29 EST 2006

[Andy Robinson]

>But there are difficulties; keeping it fast would need a lot of C code,

>and some of the information you need is not available in most font


I'm far from being a font specialist, yet I think I read that kerning is
part of at least some font formats. There is some elementary
information about ligatures that could be easily related to Unicode
tables, as there are not so many of them, but the problem might be that
they are language dependent: some ligatures are not welcome in some
languages. Difficult to please everybody. Consider the "ij" ligature
for example: it is not so welcome in French, and the suggestion that
y-diaeresis (255 in Latin-1) could be reused to that goal was not fully
acceptable either (the Finnish -- if I remember well -- objected).

>We had an early contribution which did a Knuth-like line-breaking

>algorithm and it took forever, despite only handling one font face and


Indeed, even if dynamic programming (the technique) drastically reduces
the time complexity of problems, it still requires much doing for going
below O(n^2), and for long paragraphs, this might mean a lot of time.
Going from Python to C here makes nothing more than a linear speed-up,
so it only pushes the limit a bit further, and would not buy so much.
We probably need a better algorithm, than a change of language.

(Thanks for your reply, Andy.) Keep happy, all.

François Pinard http://pinard.progiciels-bpi.ca

More information about the reportlab-users mailing list