On 7/2/21 12:49 AM, to...@tuxteam.de wrote:
Now to another pet peeve of mine: useless use of grep:
grep xxxx <myfile> | sed -e 's/bla/foo/' ...
Perhaps the underlying issue is useless use of shell pipelines ("The
Unix Way"):
2021-07-02 12:44:43 dpchrist@dipsy ~/sandbox/perl
$ cat useless-use-of-grep
This is the first line: bla bla bla.
This is line xxxx: bla bla bla.
This is the last line: bla bla bla.
2021-07-02 12:45:38 dpchrist@dipsy ~/sandbox/perl
$ grep xxxx useless-use-of-grep | sed -e 's/bla/foo/'
This is line xxxx: foo bla bla.
Versus "all-in-one":
2021-07-02 12:45:39 dpchrist@dipsy ~/sandbox/perl
$ perl -ne 's/bla/foo/ && print if /xxxx/' useless-use-of-grep
This is line xxxx: foo bla bla.
The former is better from a "golfing" standpoint (e.g. fewer characters
to type).
But, it is interesting that the benchmark results defy the common
perceptions of "userland tools are are fast" and "Perl is slow":
2021-07-02 12:59:55 dpchrist@dipsy ~/sandbox/perl
$ time for n in {1..10000}; do grep xxxx useless-use-of-grep | sed -e
's/bla/foo/' > /dev/null ; done
real 0m42.162s
user 0m32.925s
sys 0m45.787s
2021-07-02 13:00:45 dpchrist@dipsy ~/sandbox/perl
$ time for n in {1..10000}; do perl -ne 's/bla/foo/ && print if /xxxx/'
useless-use-of-grep > /dev/null; done
real 0m23.624s
user 0m14.308s
sys 0m10.826s
Perhaps this is because processing is trivial and a pipeline of two
commands has twice the process creation costs of a single command (?).
David
