Which law?

[sawyl]

The name of the law I was completely and utterly unable to remember at lunch time was Benford's.

Comments

[gleet.livejournal.com]

Benford's Law is fun:

bruno paulstevenson $ find . -type f -exec ls -l {} \; | awk '{print $5}' | cut -c1-1 | sort | uniq -c
23674 1
14912 2
10707 3
8644 4
7967 5
7423 6
3456 7
3040 8
6472 9

(don't know what 's so special about 9)

[sawyl.livejournal.com]

Very cool. Here are my numbers:

  23810 eld004> find . -type f -printf "%s\n" | awk '{f[substr($1,0,1)]++;t++}
  > END{for(i=1;i<11;++i){printf("%d  %5.2f\n",i%10,100*f[i%10]/t)}}'
  1  45.67                                                                        
  2  27.94                                                                        
  3   9.99                                                                        
  4   4.06                                                                        
  5   3.07                                             
  6   2.37                                                                       
  7   2.17                                                    
  8   2.12                                                       
  9   2.19                                                       
  0   0.43                                                    
  23811 eld004>

What do you reckon? Could this be the birth of a new meme?

[gleet.livejournal.com]

ooh, you win with awk cleverness, but it could make a good meme in certain cirles. Unfortunately, I don't think I know anyone else who could do it.

[gleet.livejournal.com]

clearly I meant circles, but that printf in the find arg list is not known to my mac or sun. Is it a common extension?

[sawyl.livejournal.com]

I think the -printf option is a GNU special. The portable way to do it is probably:

find . -type f | xargs ls -l | awk '{f[substr($5,0,1)]++;t++}
END{for(i=1;i<11;++i){printf("%d %5.2f\n",i%10,100*f[i%10]/t)}}'

[sawyl.livejournal.com]

The awk thing is a holdover from the Cray days, which used to have problems (http://www.spikynorman.dsl.pipex.com/CrayWWWStuff/Cfaqp2.html#TOC7) launching new processes. We used to have macho competitions to see who could do the most with the fewest number of child processes, all of which culminated in my writing a vast, rambling, workload manager daemon in horrific combination of awk and Korn shell. Ouch.

[gleet.livejournal.com]

Or avoiding the empty files, and I guess stat is a little more efficient than ls, though I haven't timed both ways:

find . -type f ! -empty -exec stat -f %z {} \; | awk '{f[substr($1,0,1)]++;t++} END{for(i=1;i<11;++i){printf("%d %5.2f\n",i%10,100*f[i%10]/t)}}'