[ How to split on NULs in shell ]

I am using zsh as a shell.

I would like to execute the unix find command and put the result into a shell array variable, something like:

FILES=($(find . -name '*.bak'))

so that I can iterate over the values with something like

for F in "$FILES[@]"; do echo "<<$F>>"; done

However, my filenames contain spaces at least, and perhaps other funky characters, so the above doesn't work. What does work is:

IFS=$(echo -n -e "\0"); FILES=($(find . -name '*.bak' -print0)); unset IFS

but that's fugly. This is already a bit beyond my comfort limit with zsh syntax, so I'm hoping someone can point me to some basic feature that I never knew about but should.

Answer 1

I tend to use read for that. A quick google search showed me zsh also seem to support that:

find . -name '*.bak' | while read file; do echo "<<$file>>"; done

That doesn't split with zero bytes, but it will make it work with file-names containing whitespace other than newlines. If the file-name appears at the very last of the command to be executed, you can use xargs, working also with newlines in filenames:

find . -name '*.bak' -print0 | xargs -0 cp -t /tmp/dst

copies all files found into the directory /tmp/dst. Downside of the xargs approach is that you don't have the filenames in a variable, of course. So this not always applicable.

Answer 2

Alternatively, since you're using zsh, you can use zsh's extended globbing syntax to find files rather than using the find command. As far as I know, all the functionality of find is present in the globbing syntax, and it handles filenames with whitespaces properly. See the zshexpn(1) manpage for more info. If you use zsh on a fulltime basis, it's well worth learning the syntax.

Answer 3

The only way I figured out is using eval:

(zyx:~/tmp) % F="$(find . -maxdepth 1 -name '* *' -print0)"
(zyx:~/tmp) % echo $F | hexdump -C
00000000  2e 2f 20 10 30 00 2e 2f  d0 96 d1 83 d1 80 d0 bd  |./ .0../........|
00000010  d0 b0 d0 bb 20 c2 ab d0  a1 d0 b0 d0 bc d0 b8 d0  |.... ...........|
00000020  b7 d0 b4 d0 b0 d1 82 c2  bb 2e d0 9c d0 b8 d1 82  |................|
00000030  d1 8e d0 b3 d0 b8 d0 bd  d0 b0 20 d0 9e d0 bb d1  |.......... .....|
00000040  8c d0 b3 d0 b0 2e 20 d0  93 d1 80 d0 b0 d0 bd d0  |...... .........|
00000050  b8 20 d0 be d1 82 d1 80  d0 b0 d0 b6 d0 b5 d0 bd  |. ..............|
00000060  d0 b8 d0 b9 2e 68 74 6d  6c 00 2e 2f 0a 20 0a 6e  |.....html../. .n|
00000070  00 2e 2f 20 7b 5b 5d 7d  20 28 29 21 26 7c 00 2e  |../ {[]} ()!&|..|
00000080  2f 74 65 73 74 32 20 2e  00 2e 2f 74 65 73 74 33  |/test2 .../test3|
00000090  20 2e 00 2e 2f 74 74 74  0a 74 74 0a 0a 74 20 00  | .../ttt.tt..t .|
000000a0  2e 2f 74 65 73 74 20 2e  00 2e 2f 74 74 0a 74 74  |./test .../tt.tt|
000000b0  0a 74 20 00 2e 2f 74 74  5c 20 74 0a 74 74 74 00  |.t ../tt\ t.ttt.|
000000c0  2e 2f 0a 20 0a 0a 00 2e  2f 7b 5c 5b 5d 7d 20 28  |./. ..../{\[]} (|
000000d0  29 21 26 7c 00 0a                                 |)!&|..|
000000d6
(zyx:~/tmp) % echo $F[1]
.
(zyx:~/tmp) % eval 'F=( ${(s.'$'\0''.)F} )'
(zyx:~/tmp) % echo $F[1]
./ 0

function SplitAt0()
{
    local -r VAR=$1
    shift
    local -a CMD
    set -A CMD $@
    local -r CMDRESULT="$($CMD)"
    eval "$VAR="'( ${(s.'$'\0''.)CMDRESULT} )'
}

Usage: SplitAt0 varname command [arguments]

I should have used ${(ps.\0.)F}, not ${(s.\0.)F}:

% F=${(ps.\0.)"$(find . -maxdepth 1 -name '* *' -print0)"}
% echo $F[1]
./ 0

Answer 4

I've tried

F=(*)

and it handles even the files with newlines in the filename.