| Subject: | read_file very slow in Strawberry when $& is used |
As far as I can see, this only happens on win32 with strawberry perl (but
not cygwin). Compare
perl -MFile::Slurp -E"$t=read_file('file_of_20MB'); say length $t; $&"
to
perl -MFile::Slurp -E"$t=read_file('file_of_20MB'); say length $t"
On my system the first line takes about 20s to finish, the second line
takes less than 1s. No penalty for using $& seems to occur when
binmode=>':raw' is used.
This may be unavoidable and the general penalty for using $& is mentioned
in perlre, but I wouldn't expect regex matters to influence File::Slurp
so heavily. Minimum solution would be a warning in the pod.