When implementing
ocamlscript, I
realized that argument passing for a shebang script (a Unix script
starting with
#!) was rather undocumented. So here is a
tiny documentation of this rather contrived argument passing convention.
Suppose you have the OCaml program:
let _ =
for i = 0 to Array.length Sys.argv - 1 do
Format.printf "argv(%d): \"%s\"@\n" i Sys.argv.(i)
done
Compiled with:
ocamlopt -o argv-test argv-test.ml.
And suppose you have following
script-with-args shebang
script:
#!./argv-test intarg1 intarg2 intarg3
and following
script-without-args shebang script:
#!./argv-test
As a reference, if
argv-test is called directly, you have:
$ ./argv-test arg1 arg2 arg3
argv(0): "./argv-test"
argv(1): "arg1"
argv(2): "arg2"
argv(3): "arg3"
Now, let's compare the different ways of calling the shebang script:
$ ./script-without-args
argv(0): "./argv-test"
argv(1): "./script-without-args"
$ ./script-without-args arg1 arg2 arg3
argv(0): "./argv-test"
argv(1): "./script-without-args"
argv(2): "arg1"
argv(3): "arg2"
argv(4): "arg3"
$ ./script-with-args
argv(0): "./argv-test"
argv(1): "intarg1 intarg2 intarg3"
argv(2): "./script-with-args"
$ ./script-with-args arg1 arg2 arg3
argv(0): "./argv-test"
argv(1): "intarg1 intarg2 intarg3"
argv(2): "./script-with-args"
argv(3): "arg1"
argv(4): "arg2"
argv(5): "arg3"
You see the pattern? :) If the shebang script is called with arguments
given on the shebang line, those arguments are available in a single
string as argv(1). Otherwise, argv(1) is equal to the name of the
shebang script itself. If any argument in given on the command line to
the shebang script, they are available as normal argument, after binary
name, optional internal arguments and script name. To determine if the
shebang script has internal arguments: if the called binary has at least
three arguments, one should test if argv(1) is the name of an real file
(and thus the script has not argument) or not (and thus the script has
arguments).
As a side note, if one wants to use the
env utility:
$ cat ./env-script
#!/usr/bin/env ./argv-test
$ ./env-script
argv(0): "./argv-test"
argv(1): "./env-script"
$ ./env-script arg1 arg2 arg3
argv(0): "./argv-test"
argv(1): "./env-script"
argv(2): "arg1"
argv(3): "arg2"
argv(4): "arg3"
However
env does not support arguments on the shebang line:
$ cat ./env-script-with-args
#!/usr/bin/env ./argv-test intarg1 intarg2 intarg3
$ ./env-script-with-args
/usr/bin/env: ./argv-test intarg1 intarg2 intarg3: No such file or directory
2006-03-29T19:38:24Z []
permanent link
I wanted to know how files are distributed in my home directory, so I
wrote a quick
OCaml
program to do this analysis. This program gives following result on
my home directory (it counts links as files):
$ ./filesystem-analysis.ml ~
Total file size: 6120592065 bytes
Minimum non empty file size: 1.00 B
Maximum file size: 163.77 MB
Average file size: 28.81 KB
Total number of files: 207476
of which are empty files: 544
Total number of directories: 22932
File size distribution:
[ 1.00 B - 1.00 B ] 592 files - total average of 592.00 B
[ 2.00 B - 3.00 B ] 574 files - total average of 1.68 KB
[ 4.00 B - 7.00 B ] 772 files - total average of 4.52 KB
[ 8.00 B - 15.00 B ] 2205 files - total average of 25.84 KB
[ 16.00 B - 31.00 B ] 4077 files - total average of 95.55 KB
[ 32.00 B - 63.00 B ] 15371 files - total average of 720.52 KB
[ 64.00 B - 127.00 B ] 7598 files - total average of 712.31 KB
[ 128.00 B - 255.00 B ] 6258 files - total average of 1.15 MB
[ 256.00 B - 511.00 B ] 10922 files - total average of 4.00 MB
[ 512.00 B - 1023.00 B ] 21493 files - total average of 15.74 MB
[ 1.00 KB - 2.00 KB] 24063 files - total average of 35.25 MB
[ 2.00 KB - 4.00 KB] 35981 files - total average of 105.41 MB
[ 4.00 KB - 8.00 KB] 27669 files - total average of 162.12 MB
[ 8.00 KB - 16.00 KB] 16737 files - total average of 196.14 MB
[ 16.00 KB - 32.00 KB] 16654 files - total average of 390.33 MB
[ 32.00 KB - 64.00 KB] 7057 files - total average of 330.80 MB
[ 64.00 KB - 128.00 KB] 4190 files - total average of 392.81 MB
[ 128.00 KB - 256.00 KB] 2905 files - total average of 544.69 MB
[ 256.00 KB - 512.00 KB] 1036 files - total average of 388.50 MB
[ 512.00 KB - 1024.00 KB] 738 files - total average of 553.50 MB
[ 1.00 MB - 2.00 MB] 301 files - total average of 451.50 MB
[ 2.00 MB - 4.00 MB] 115 files - total average of 345.00 MB
[ 4.00 MB - 8.00 MB] 104 files - total average of 624.00 MB
[ 8.00 MB - 16.00 MB] 38 files - total average of 456.00 MB
[ 16.00 MB - 32.00 MB] 14 files - total average of 336.00 MB
[ 32.00 MB - 64.00 MB] 8 files - total average of 384.00 MB
[ 64.00 MB - 128.00 MB] 1 files - total average of 96.00 MB
[ 128.00 MB - 256.00 MB] 3 files - total average of 576.00 MB
Returned results are pretty expected. Over the 6GB of files in my home
directory, I have about 200,000 files of an average size of 30KB and
23,000 directories (about 10% of the number of files). The vast majority of
files are below 32KB but a small number of big files are taking the vast
majority of my disk space.
A more interesting point is that a non negligible number of files (10%)
is smaller than 64 bytes. In the case of the distributed backup system,
it would be time and space saving to save them within the master index.
2006-03-21T20:07:50Z []
permanent link
