Re: [Devel] Problems with bbox code and cubic bezier curves

[Just van Rossum]

Tom Kacvinsky wrote:

> Well, it sounds like you are half a Van Gogh, then, Just! ;)

Heh, I knew that dutch expression wouldn't work too well in english ;-)

> Kidding aside, I took a look at the sample PDF file sent to me.  It is not 
> using a font at all, so it must be the case that FreeType is getting used for 
> things other than font rasterization.  So we need to be careful about bbox 
> stuff. 

But I don't understand *why* to *ever* worry about an exact bounding box? If the
outlines are decent, the cbox is fine, and if the outlines are not so decent,
then the cbox may be too large, but why should FT care?

To me, an exact bbox is only "nice to have", and therefore performance can't be
an issue either. If performance is an issue, you can't beat the cbox! I don't
believe that the cost of *always* calculating an exact bounding box weighs up
against wasting some space in fairly uncommon cases. Non-conforming fonts? Blame
the font. Rotated text? I really wouldn't care if it were slightly slower.

(Oh, and you can't crash a cbox function due to weird input...)

Just