Re^3: Splitting on escapable delimiter

[ikegami]

Nope, that returns an extra (empty) field most of the time, and there's no way to know when. For example, 'a@b' incorrectly returns 3 fields, although 'a@' correctly returns 2.

[Roy Johnson]

That's weird. I'm trying to figure out why mine takes another pass after encountering end of string (even if I specify that the separator can be end-of-string). Meanwhile, I'll point out that your method finds only one field in '@' (instead of two empty ones — if there's no newline after the @).

Update: this seems to work:

  my @pieces = /(?:^|\G@)((?:#.|[^@])*)/sg;

Still don't know why I couldn't get it to realize it had hit end of string. My guess is, if end-of-string is not required to match, it takes another shot to make sure nothing matches.

[ikegami]

Still don't know why I couldn't get it to realize it had hit end of string

It's seen as somewhat of a bug. It happens when the search expression can match a zero length string. In the followig, the parens indicate what the three passes of your earlier s/// matches:

(a@)(b)()

Or in terms of @- and @+:

pass	@-	@+
1	0	2
2	2	3
3	3	3

The /g loop ends when the fourth pass matches the same thing as the third pass and pos+1 is beyond the end of the string.