QuickFont

NICE! Sorry to hear about the girlfriend but its all in the name of science right? :D

Thanks Tksuoran, your feedback is extremely useful. I realise that I have approached writing this library slightly naively.

>>You really, really should have kerning pairs.

Absolutely. I knew something looked a bit wrong.

I will need to manually compute the kerning information for texture fonts (as glyph shape may have changed during texturing), and I'm not too worried about this because although it will be computationally very intensive, it can be done once per font and saved to the .qfont file so that it does not later affect application load times.

My concern is that manually calculating kerning every time a font is loaded (even those loaded directly from a .ttf file) will be slow. I take it that I can extract kerning information from the ttf files?

>>Don't forget to be gamma correct when you scale or filter.

Fair point. I will attempt to modify my scaling algorithm to compensate for this at some point. Even if it is very slow to compute, it would be a very nice feature to have when generating the texture fonts in the first place.

>>All GL stuff is using the old fixed function. These days we would like to have shader support.

Would this be a lot of work to support, and how would I go about doing it?

Excellent work ...

An ideal complement to work with OpenTK.

Miguel

>>for example after letter f there seems to be too much space in the last picture.

Which f do you mean? I've highlighted all fs in the last picture.

(I've deleted the pre-kerning pictures to avoid confusion)

As far as I can tell it's working perfectly. The only exception I've had is the _ character which tends to be below the glyphs and therefore gets kerned to the max. What I'll do is to allow the user to define kerning exceptions, and by default I will have such an exception set for underscore.

I've compared it visually with a handful of fonts and it seems pretty flawless. The few cases where there are differences, I've not even been convinced that the manual kerning values are better than those selected automatically by my code. Though to be fair I've not tested it exhaustively, and as mentioned '_' gets auto-kerned wrong.

An easy way to see just how well the auto kerning is working is to set the character spacing (independent of kerning) to zero. I've done that in the second screenshot above. Basically it does exactly what I coded it to do: puts the glyphs as close as they'll go to each other without overlapping. :P