Re: [TCLCORE] Making Tcl 9 go (way) faster [IO/pcre]

[Donal K. F. <don...@ma...>]

On 28/12/2012 00:09, Larry McVoy wrote:
> We use Jeff's PCRE patch in our fork of Tcl (that we need to merge forward
> to 8.6 and release).  We like it but the patch was incomplete, it breaks
> tests (that is my memory, it's possible we are based on a version of tcl
> that didn't pass tests).

Breaking tests is not *necessarily* a problem, if those tests are
depending on the exact way that Henry's code worked. Since PCRE uses a
different type of RE engine (stack-based instead of automata-theoretic)
I'd expect a number of edge cases to be different, yet for that
difference to be of minimal significance to most code.

> With much respect to Henry, we like PCRE better.
>

To my eyes, the best thing about PCRE is that it makes maintenance of it
someone else's problem. :-) If we had someone who was really willing to
put the effort in to maintain Henry's engine, I wouldn't be as in favour
of switching, but we are where we are. I've got too many other things to
do to have time to learn in depth how to handle Henry's code as well.

Donal.