I wrote a tool that reads lines from stdin and prepends them with a timestamp. Out of reasons I did so in C, awk, perl, python and shell. After I was done I found out, that there is already such a tool, ts from moreutils. ts from moreutils is a Perl script. I did a little test comparing for how long each version needs to process some dummy input.
The basic function looks like this in shell:
while read -r line do d=$(date '+%Y%m%dT%H%M%S') printf "%s %s\n" "$d" "$line" done
Let's generate a file with many lines:
% dd if=/dev/zero bs=1024 count=1024 2>/dev/null | od -v > t_s % wc -l t_s 65537 t_sAnd with the help of a little script we get results:
% ./t t_s ts t_s real 0.190000 user 0.150000 sys 0.030000 C t_s real 0.020000 user 0.010000 sys 0.000000 awk t_s real 0.020000 user 0.020000 sys 0.000000 pl t_s real 0.080000 user 0.070000 sys 0.000000 py t_s real 0.180000 user 0.170000 sys 0.010000 sh t_s real 33.740000 user 27.780000 sys 9.250000 ./t t_s 28.23s user 9.31s system 109% cpu 34.233 totalThe first column identifies the version, where 'ts' is the version from moreutils. Awk is GNU awk here, because I needed the time functions which are not include in The One True AWK.
Let's see with a bigger file and without the shell version:
% dd if=/dev/zero bs=8192 count=8192 2>/dev/null | od -v > t_xl % wc -l t_xl 4194305 t_xl % ./t t_xl ts t_xl real 10.360000 user 8.000000 sys 2.340000 C t_xl real 0.920000 user 0.830000 sys 0.090000 awk t_xl real 1.350000 user 1.260000 sys 0.080000 pl t_xl real 4.280000 user 4.200000 sys 0.060000 py t_xl real 10.600000 user 10.520000 sys 0.070000 ./t t_xl 24.85s user 2.68s system 100% cpu 27.529 totalI'm really amazed how fast awk is. And it's very interesting to see the difference between my perl version and the original ts.
That's it for a sunday - back to real work!
Here is the code: the codez.
Last modified: 2023-10-29T12:32:36Z