> A no-op?
Please note that in mathematics an empty set is always a sub set of all other sets.
And so does the empty pattern in Perl always match.
If no match has previously succeeded, this will (silently) act instead as a genuine empty pattern (which will always match).
Hence not a no-op!°
> that a code expression that doesn't do anything useful should get a special case so that it does something useful.
I disagree. The "repeat the last match" feature should be dependent on an explicit modifier like eg s///b ²
I'd welcome a feature to switch off the current default behavior in favor of an explicit modifier.
Reasoning: The ratio of useful vs confusing is too bad.
Update
°) Demo:
~ $ perl -de0
...
DB<1> $_='XXXX'
DB<2> s/(?:)/Y/ # genuine empty pattern
DB<3> p
YXXXX
DB<3> s/(?:)/Z/g
DB<4> p
ZYZXZXZXZXZ
DB<4>
²) in an earlier version I suggested s///l but this modifier is already taken. "b" (for before) seems to be free.
| [reply]
[d/l]
[select] |
100% agree. At $job - 2, we spent several days tracking a bug caused by s/$regex/$replace/ where $regex got accidentally empty and the substitution started replacing unrelated parts of the data.
Besides /b or a feature, another way would be to introduce a new construct, e.g. (*LAST), which you can use similarly to (*SKIP) etc.
map{substr$_->[0],$_->[1]||0,1}[\*||{},3],[[]],[ref qr-1,-,-1],[{}],[sub{}^*ARGV,3]
| [reply]
[d/l]
[select] |
| [reply] |
I expected it just do nothing, But it's more useful to take last matched pattern. | [reply] |
s/(?:)// will do nothing (or rather, replace one-zero length string with another zero-length string). So, if you have a possibly-empty pattern in a variable, you could use
s/(?:$pat)//
or
my $re = qr/$pat/;
s/$re//
The latter was buggy before 5.18 as it resulted in the same as s/// when $pat contained an empty string.
| [reply]
[d/l]
[select] |
glad to learn that there is a way to suppress the urge of abnormal behavior.
| [reply] |
