Re: Documentation proposal

["Harvey J. Stein" <hjstein at bfr dot co dot il>]

Michael Vanier <mvanier@bbb.caltech.edu> writes:

 > I agree, but I also doubt that there's anything fundamental about guile
 > that precludes it being useful as a text-munging language; it's just the
 > speed of the current implementation of regexps etc.

With emphasis on etc.  When I was testing (on Redhat 5.1, I think -
might even have been 4.2), guile %read-line was slow compared to perl
& awk, but for my application it wasn't the killer.

It's true that guile regexp handling was much slower than perl & awk,
but this was because guile was using the system regexp library,
whereas perl & awk used their own regexp libraries.  I mailed in a
patch including awk's regexp libarary (the faster of the two) & making
the regexp lib switchable btw the system & the one included, but I
don't think it was ever incorporated.

But, it seemed like there was a lot of overhead in the internals of
the guile<->regexp lib interface (but don't hold me to this).

However, awk's regexp library is from glibc, and it seems that awk
just happened to have snarfed from a newer glibc than was on RH 5.1.

I haven't tested things on 6.2 yet...

-- 
Harvey Stein
Bloomberg LP
hjstein@bfr.co.il