Here are the two conditions necessary for irregex-search to break its "leftmost, longest" contract:
1. Range iteration instead of star iteration.
2. An alternation where the first branch is a prefix of the second branch.
A minimal example:
$ mit-scheme --load irregex.scm
[...]
;Loading "irregex.scm"... done
1 ]=> (define (imsis re str) (irregex-match-substring (irregex-search re str)))
;Value: imsis
1 ]=> (imsis "(a|ab){0,3}" "abab")
;Value 35: "a"Irregex-match keeps working:
1 ]=> (define (imsim re str) (irregex-match-substring (irregex-match re str)))
;Value: imsim
1 ]=> (imsim "(a|ab){0,3}" "abab")
;Value 36: "abab"If only one condition is met, irregex-search works too:
1 ]=> (imsis "(ab|a){0,3}" "abab")
;Value 34: "abab"
1 ]=> (imsis "(a|ab)*" "abab")
;Value 33: "abab"I think this might be worth telling the shinnoid. I cannot see any obvious contact info on
http://synthcode.com/scheme/irregex/
but his git commits on
https://github.com/ashinn/irregex
are by "Alex Shinn <alexshinn@gmail.com>". Well, at least he's not one of those protonmail people. If someone could drop him a line, that would be great.
That's what I tried
Can't remember what the problem was
I could have miswritten it
Clearly, it was a typo of some sort. No worries.