Here's where I came out on all of that.  Regex is useful only because 
it is: (i) concise; (ii) powerful; and (iii) very fast.  It is a lot 
like BitBlt in that regard, certainly like string searches, all of 
which are Squeak primtiives.  Regex won't be useful in Squeak, 
certainly not in the tools, unless it is powerfully fast, like a string 
search.

rePlugin offers significantly more power (full Perl5) at excellent 
speed than any other regex I have seen.  It is strong, well-tested and 
robust.

On Dec 1, 2003, at 11:02 PM, smallsqueak wrote:

>
> Avi Bryant wrote :
>
>>
>> On Dec 1, 2003, at 3:56 PM, smallsqueak.net wrote:
>>
>>>
>>> Hi Avi,
>>>
>>>> FWIW, I vote that if we're going to have the core depend
>> on a regex
>>>> package, it should be one of the Smalltalk ones - either
>> Vassili's or
>>>> the one from SmaCC.
>>>
>>>     Which of these will work in a headless environment,
>>>     something like SqueakScript ?
>>
>> Er, any of them, I would think.  Why wouldn't they?
>>
>
>   That's great.
>   So, regex does not need Morphic to run correctly ;-)
>
>   But, how can we know for sure ?
>   Can MC help (via dependency analysis) ?
>
>   For the rePlugin, I know for sure
>   and it will run even with Squat
>   (which may end up with just ProtoObject,
>    Metaclass and supporting cast ;-)
>
>   BTW, why 'it should be one of the Smalltalk ones' ?
>   There are at least 3 of them at the last count ;-)
>
>   Considering that Smalltalk is very productive,
>   we may have a dozen of these in no time ;-)
>
>   Considering the current rate of evolution of Squeak,
>   what's your guesstimate for the sweat needed to test
>   and maintain a native Squeak version of regex
>   comparable to the current (old) rePlugin ?
>   (and don't forget that KCP is sweeping left and
>    right, pulling the rug right under your feet ;-)
>
>   Cheers,
>
>   PhiHo.
>
>