Regarding naming, I deliberately switched from "irregex" (no "p") to "regexp", though I can't recall why. Perhaps just to differentiate. Shouldn't the plural be "regexen"? :)
Regarding match semantics, the intention was to deliberately leave this unspecified, though it's such a crucial issue that at least that decision should be stated clearly. Specifically, I didn't want to force implementors to use or reinvent the reference implementation, but to have the option of just compiling SREs to their native regexp implementations, like SCSH does. I think there should be a PFN clarifying this and also making it clear that leftmost-longest is preferred if at all possible. It's also worth noting that the non-greedy patterns, if supported, should be leftmost-shortest. Separately it may be worth noting that the sample implementation now supports all features except back references (which it likely never will). -- Alex On Sat, Oct 19, 2024 at 2:29 PM Arthur A. Gleckler <s...@speechcode.com> wrote: > On Fri, Oct 18, 2024 at 7:18 AM Daphne Preston-Kendal <d...@nonceword.org> > wrote: > > >> Also, it’s been raised by John before (on my instigation, heh), but I >> really do feel a PFN could clarify that leftmost-longest semantics are >> intended for submatches, if that is indeed the intention. (chibi regexp) >> uses leftmost-longest; I haven’t tested the ‘other’ sample implementation, >> irregex (no ‘p’!), but based on what Alex said when the issue was raised >> last I assume it does as well. >> <https://srfi-email.schemers.org/srfi-115/msg/3693241/> >> > > Alex, would you please chime in? > > Thanks. >
