karl williamson wrote: >Here's my reasoning for excluding from consideration most of these. I've >scanned the tokenizer code, and there appear to be heuristics to decide >if something is an interpolated variable, the end of the pattern, and the >boundaries of character classes. So it seems a lot less dangerous to >exclude $ % @ ; [ ] $ % @ I agree with. Why do you include semicolon? I don't see a problem with it. Brackets I think should be OK here if they're used in a paired way. So not suitable for our current job, but potentially available for something else. > Also, quotes as you said, " ' ` >And, since / is the common regex delimiter, excluding it seems like a >good idea for human confusability issues, I'd exclude / here, because we've got to leave some character that's always available as a delimiter, and since / is the standard delimiter I'd pick that one. But " ' ` are *not* commonly quote characters for regexps, so I don't think it's necessary to exclude them. Indeed, ' is already used as a metacharacter in (?'NAME'pattern). > as well as anything that is a >paired delimiter, so } is out. Yes, save that for paired usage. {} pairs are already used in several parts of regexp syntax, and I wouldn't want to use either of them in any other way. Both for mnemonic reasons and to avoid screwing up {} quoting of regexps. > Similar concerns get rid of \ Yes. > I'd >rather keep _ in reserve as it is a word character, I was thinking _ would be a fairly good choice for our current use. The spacer uses you suggest, or at least some of them, would probably be better served by space characters and /x. But I don't object to keeping our hands off _ for now. >I'd rather not use ^, again for human usability issues, as that often is >the first thing in a pattern. I don't see a problem with (?^...) meaning something that has nothing to do with the ^ anchor. > A comma doesn't seem to me to convey the >right meaning, Agreed. Would be fine for some other use. > I didn't want to use tilde >because of the visual confusability with -, As previously noted, I disagree with this being a problem. > and not * because of a >number of things that start like (*PRUNE), I don't think that should stop us. (?*...) is quite distinct from (*...). > so that left the period. Period is fine. For our present purpose, I'd be happy with any of _ ^ ~ * . -zeframThread Previous | Thread Next