syntax question. if I have a number of subdirectories within a target dir, and I want to output the names of the subs to a text file I can easily run:
ls > filelist.txt
on the target. But say all of my subs are named with a 7 character prefix like:
JR-5426_mydir
JR-5487_mydir2
JR-5517_mydir3
...
and I just want the prefixes. Is there an option to "ls" that will only output n characters per line?
Don't use ls in any programmatic context; it should be used strictly for presentation to humans -- ParsingLs gives details on why.
On bash 4.0 or later, the below will provide a deduplicated list of filename prefixes:
declare -A prefixes_seen=( ) # create an associative array -- aka "hash" or "map"
for file in *; do # iterate over all non-hidden directory entries
prefixes_seen[${file:0:2}]=1 # add the first two chars of each as a key in the map
done
printf '%s\n' "${!prefixes_seen[@]}" # print all keys in the map separated by newlines
That said, if instead of wanting a 2-character prefix you want everything before the first -, you can write something cleaner:
declare -A prefixes_seen=( )
for file in *-*; do
prefixes_seen[${file%%-*}]=1 # "${file%%-*}" cuts off "$file" at the first dash
done
printf '%s\n' "${!prefixes_seen[@]}"
...and if you don't care about deduplication:
for file in *-*; do
printf '%s\n' "${file%%-*}"
done
...or, sticking with the two-character rule:
for file in *; do
printf '%s\n' "${file:0:2}"
done
That said -- if you're trying to Do It Right, you shouldn't be using newlines to separate lists of filename characters either, because newlines are valid inside filenames on POSIX filesystems. Think about a file named f$'\n'oobar -- that is, with a literal newline in the second character; code written carelessly would see f as one prefix and oo as a second one, from this single name. Iterating over associative-array prefixes, as done for the deduplicating answers, is safer in this case, because it doesn't rely on any delimiter character.
To demonstrate the difference -- if instead of writing
printf '%s\n' "${!prefixes_seen[@]}"
you wrote
printf '%q\n' "${!prefixes_seen[@]}"
it would emit the prefix of the hypothetical file f$'\n'oobar as
$'f\n'
instead of
f
...with an extra newline below it.
If you want to pass lists of filenames (or, as here, filename prefixes) between programs, the safe way to do it is to NUL-delimit the elements -- as NULs are the single character which can't possibly exist in a valid UNIX path. (A filename also can't contain /, but a path obviously can).
A NUL-delimited list can be written like so:
printf '%s\0' "${!prefixes_seen[@]}"
...and read back into an identical data structure on the receiving end (should the receiving code be written in bash) like so:
declare -A prefixes_seen=( )
while IFS= read -r -d '' prefix; do
prefixes_seen[$prefix]=1
done
No, you use the cut command:
ls | cut -c1-7
