Re: Gnumeric parsing efficiency

[Morten Welinder <mortenw gnome org>]

All this is obviously compiler-dependent, but a quick spot check
says that the compiler that I use (gcc) will create jump tables
for the switch statements we use.  These tend to be dense
with about 10-20 cases.

However, the larger point here should be that we know of no
cases that (a) are even moderately slow, and (b) slowness is
even remotely attributable to switching.

So at present we do not have this problem.  If we did, I imagine
that the solution would be to compile the abstract syntax trees
into some pseudo code and execute that.

Morten