[ruby-talk:9713] Re: Regexp engine (was Re: 101 Misconceptions About Dynamic Languages)

<[e-mail]> wrote:
>
> >>>>> "B" == Brian F Feldman <[e-mail]> writes:
>
>B> Henry Spencer's totally rewritten regex package which is now embedded in 
>Tcl
>B> 8.3 is _THE_ one to choose.
>
>  Do it have some incompatibilities with perl regexp ?
>
It used to.  TCL tried to be compatible with the POSIX
regular expression standard which is designed for DFA
engines.  There is essentially no compromise between
the POSIX standard and Perl's ability to have a
fine-grained switching between greedy and non-greedy.

This is a significant difference and I can find no
reference to it having changed.

Additionally Tcl used to not support most of the
standard RE features.  This may have changed.
>
>B>                            It should be the fastest one out there now, 
>and
>B> a nice license to boot.
>
>  Do you have some benchmark for regexp ?
>
A Google search for "TCL 8.3" and "regexp" turned up:

http://lists.bell-labs.com/pipermail/nsbd/2000q2/000001.html

which is not very promising.

Friedl is in the process of producing a new edition of
Mastering Regular Expressions.  He undoubtably has a
good benchmark suite.

Benchmarks are going to be misleading here.  Judging
from what was said in MRE edition 1, the TCL engine is
a DFA at heart.  However there is a quadratic support
for capturing subexpressions.  That means that in the
very common case of capturing subexpressions, on
average the TCL engine will be drastically slower than
an NFA.  However there is *NO* RE which is going to be
disasterously slow.

But backreferences won't be supported or supportable.

Cheers,
Ben
_________________________________________________________________
Get your FREE download of MSN Explorer at http://explorer.msn.com