[COFF] Requesting thoughts on extended regular expressions in grep.

Hi Dan, It could be the quickest way to express the intent. I think those two can overlap. :-) Why, if the grammar is type three in Chomsky's hierarchy, i.e. a regular grammar? I think sticking with code aimed at regular grammars, or more likely regexps, will do better than, say, a parser generator for a type-two context-free grammar. As well as the lex(1) family, there's Ragel as another example. http://www.colm.net/open-source/ragel/ Well, we're back to the RE v. regexp again. /^[0-9]+\.jpeg$/ is matched by some engines by first checking the last five bytes are ‘.jpeg’. $ debugperl -Dr -e \ ... Matching REx "^[0-9]+\.jpeg$" against "123546789012354678901235467890123546789012.jpg" Intuit: trying to determine minimum start position... doing 'check' fbm scan, [1..46] gave -1 Did not find floating substr ".jpeg"$... Match rejected by optimizer Freeing REx: "^[0-9]+\.jpeg$" $ Boyer-Moore string searching can be used. Common-subregexp-elimination can spot repetitive fragment of regexp and factor them into a single set of states along with pairing the route into them with the appropriate route out. The more regexp engines are optimised, the more benefit to the programmer from sticking to a regexp rather than, say, ad hoc parsing. The theory of REs is interesting and important, but regexps deviate from it ever more. -- Cheers, Ralph.