• Craig Black (8/8) Feb 16 2006 With all this discussion of $ stuff for regular expressions, I was
• pragma (13/21) Feb 16 2006 That's a good question. I never did any timed comparisons of the templa...
• Walter Bright (3/4) Feb 16 2006 Sure. Compile with -profile .

"Craig Black" <cblack ara.com> writes:

With all this discussion of $ stuff for regular expressions, I was 
wondering, aren't the templated compile-time regular expressions eventually 
going to replace the current regular expression implementation?  I would 
assume they will be much much faster.  I think we should keep that in mind 
when adding syntax sugar to the language. We should ensure that whatever 
syntax sugar is added, it will be compatible with the newer faster regex 
stuff.

-Craig 

pragma <pragma_member pathlink.com> writes:

In article <dt2jvd$14s1$1 digitaldaemon.com>, Craig Black says...
With all this discussion of $ stuff for regular expressions, I was 
wondering, aren't the templated compile-time regular expressions eventually 
going to replace the current regular expression implementation?  I would 
assume they will be much much faster.  I think we should keep that in mind 
when adding syntax sugar to the language. We should ensure that whatever 
syntax sugar is added, it will be compatible with the newer faster regex 
stuff.

-Craig 

That's a good question.  I never did any timed comparisons of the template
solution versus Walter's implementation.  Although, making the templated version
opMatch friendly should be pretty easy to do.

Anyone have any nanosecond (Windows) timing code handy?  

Anyway, I really feel (as others do) that tying the evaulation of "<string>
<opMatch> <operand>" to std.regexp is not the ideal way forward (but it does
work).  If Walter were to move beyond this to a compile-time regex operation
instead, I'd insist that a non-library approach be used as it would invariably
yield a faster expression processor.  However, the unfortunate consequence of
that would be a harder to implement language standard as it would require a
whole 'nother codegen engine within the compiler itself.

- Eric Anderton at yahoo

"Walter Bright" <newshound digitalmars.com> writes:

"pragma" <pragma_member pathlink.com> wrote in message 
news:dt2lrp$16dt$1 digitaldaemon.com...
 Anyone have any nanosecond (Windows) timing code handy?

Sure. Compile with -profile <g>.