829488 - Replace substitution with transliteration wherever possible to improve performance

Status: NEW

(Bugzilla :: Bugzilla-General, enhancement)

Description

[Koosha KM]

Attachment: Perl benchmark script

Quote from perldoc perlperf:

"Search and replace or tr
       If we have a string which needs to be modified, while a regex will
       almost always be much more flexible, "tr", an oft underused tool, can
       still be a useful."

it also has a benchmark script that produces interesting results. I changed the iteration to run ten times more; i.e. 1E7 instead of 1E6. On my machine the result is

Benchmark: timing 10000000 iterations of sr, tr...
        sr: 12 wallclock secs (12.21 usr +  0.01 sys = 12.22 CPU) @ 818330.61/s (n=10000000)
        tr:  4 wallclock secs ( 3.99 usr +  0.00 sys =  3.99 CPU) @ 2506265.66/s (n=10000000)

 
As you see, the difference is really noticeable. We should replace s/// with tr/// wherever possible.

I attach the benchmark script as well as the ack-grep result of files that have s///.

Comment 1

[Koosha KM]

Attachment: ack-grep

Comment 2

[Frédéric Buclin]

Are there places in Bugzilla were tr// would be suitable? tr// doesn't support regexps and doesn't play nice with different encodings.

Comment 3

[Koosha KM]

There are some places like:

Bugzilla/Token.pm
170:    $token =~ s/\+/-/g;
171:    $token =~ s/\//_/g;

Bugzilla/WebService/Server/JSONRPC.pm
93:        $name =~ s/-/_/g;

Bugzilla/WebService/Server/XMLRPC.pm
32:            $value =~ s/-//g;

buglist.cgi
126:        $bug_ids =~ s/[:-]/,/g;

reports.cgi
104:    $image_file =~ s/\+/-/g;
105:    $image_file =~ s/\//_/g;

and some more. But one needs to be watchful enough to capture all of them and still not make mistake where regexp must be used.