On Fri, 6 Aug 2010 00:30:51 Steve Wampler wrote:
> 
> I used to do (long ago and far away) a fair amount of programming in
>  S4 (nothing near to what the long-time S4 advocates have done, of
>  course!) My experience comparing what I did in S4 with what I do in
>  Icon/Unicon is that while S4 PM let me succinctly describe the
>  matching process, by the time I added in the code to actually do
>  things with the results, it wasn't that much more terse than
>  Icon/Unicon, and a lot harder for me to understand.
> 
That point has puzzled me too. I don't see the SNOBOL tutes actually 
building much in the way of data structures representing the input. I've 
had great success in Icon scanning stuff into local temps and then 
handing them off to a record constructor, or lately to an object 
constructor.

However one of the very first things I ever tried to do in Icon was a 
FORTRAN to C converter. My very first cut used far too much slicing and 
dicing and was quite slow. When I tried a second time making much better 
use of the internal string optimisations (not to mention better style) 
it was much faster. I can see how that type of job, munging lines to 
lines, could be quite sussinct in SNOBOL. Lines to structures, I'm not 
yet convinced.



------------------------------------------------------------------------------
This SF.net email is sponsored by 

Make an app they can't live without
Enter the BlackBerry Developer Challenge
http://p.sf.net/sfu/RIM-dev2dev 
_______________________________________________
Unicon-group mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/unicon-group

Reply via email to