🚀 go-pugleaf

bkph@kauai.ai.mit.edu (Berthold K.P. Horn) writes:


> Using Subrs for common subshapes is a great idea, with one major problem:
> hinting.  The rmoveto's, rlineto's and rcurveto's are all relative, and so
> can be easily used to generate the same shape in different positions.
> But hints are all absolute (with respect to the left side bearing point).
> So the shared code better not contain any hints.
...> But please
> don't destroy a good commercial quality font that uses hint replacement, 
> for example, by applying this method for extracting common subshapes...
> Ah, I just noticed you said `PD fonts' in your posting (:-).
I would expect commercial fonts not to need any "optimization" of this 
kind anyway (since I expect "commercial" should also mean "professional" 
and this should imply knowledge of how to handle formats etc - I know 
this implication is not always true).
Anyway: what whould be the problem of extracting identical sequences 
of commands into Subrs? Their meaning relative to the Charstring they 
get called from is not changed by this even if by chance it should be 
a different one in some cases. I think the problem you seem to describe 
does not exist.
Thomas Wolff@inf.fu-berlin.de