[SOLVED] Bash: how to populate a list of arbitrarily named files?

catkin · 03-14-2010, 03:29 AM

Hello

How to build a list of files under a directory that may have any permissible characters in the name, that is anything except NUL?

The only possible (?) bash data structure to contain a list of such names is an array because NUL cannot be used as a list item separator so no X-separated list can safely be used; there is no "X" that might not be part of a file name.

OK -- but how to populate such an array? Here's what I've tried.

Code:

#!/bin/bash

# Set up test files
dir=$(mktemp -d "/tmp/${0##*/}.XXXXXX")
touch $dir/foo $dir/bar

# This does populate the array but is useless because assignments set
# in a pipeline are lost after execution
i=0
/usr/bin/find $dir -type f -print0 | while IFS= read -r -d '' filename
do
    echo "DEBUG: pipeline: filename is '$filename'"
    files[$i]="$filename"
    let i=i+1
done

# And this fails.  I don't know why; maybe the NUL characters are lost
# in expansion.
while IFS= read -r -d '' filename
do
    echo "DEBUG: here string: filename is '$filename'"
    files[$i]="$filename"
    let i=i+1
done <<< "$( /usr/bin/find $dir -type f -print0 )"

# Remind to remove test files
echo "You probably want to run /bin/rm -fr $dir now"

blacky_5251 · 03-14-2010, 03:39 AM

What about:-

Code:

#!/bin/bash

# Set up test files
dir=$(mktemp -d "/tmp/${0##*/}.XXXXXX")
touch $dir/foo $dir/bar

i=0
for filename in $(ls | xargs)
do
    echo "DEBUG: pipeline: filename is '$filename'"
    files[$i]="$filename"
    let i=i+1
done

catkin · 03-14-2010, 03:58 AM

Quote:

Originally Posted by blacky_5251

What about:-

Thanks Ian

but that will not work for file names including difficult characters such as linefeed and backspace as explained in detail here (and it will not descend into subdirectories).

blacky_5251 · 03-14-2010, 05:40 AM

How about this then:-

Code:

for filename in $(find $dir -type f -print0 | xargs -0)
do
  files[$i]="$filename"
  let i=i+1
  echo DEBUG "$filename"
done

catkin · 03-14-2010, 06:04 AM

Thanks Ian

That works perfectly.

tuxdev · 03-14-2010, 11:44 AM

Actually, that fails for spaces in filenames as well. for thing in $() construct is almost certainly wrong and the exact behavior is very dependent on the value of IFS.

Use this instead:

Code:

while IFS="" read -r -d "" file ; do
   files+=("$file")
done < <(find ... -print0)

And as always, Use More Quotes!

catkin · 03-14-2010, 12:05 PM

Quote:

Originally Posted by tuxdev

Actually, that fails for spaces in filenames as well. for thing in $() construct is almost certainly wrong and the exact behavior is very dependent on the value of IFS.

Use this instead:

Code:

while IFS="" read -r -d "" file ; do
   files+=("$file")
done < <(find ... -print0)

And as always, Use More Quotes!

Thanks tuxdev

I had just discovered that and came back to this thread to mark it [UNSOLVED] so it was great to find your solution.

How does the done < <( <command> ) construct work? I don't understand "< <".

In case anyone is interested here's the test script (some stylistic changes from tuxdev's suggestion but essentially the same thing).

Code:

#!/bin/bash

# Set up test files
dir=$(mktemp -d "/tmp/${0##*/}.XXXXXX")
touch $dir/foo
touch $dir/'path^Ha^Mlogical'
touch $dir/'name with spaces and    tab'

rm -f /tmp/trash

while IFS= read -r -d '' file
do
   files+=("$file")
done < <(find $dir -type f -print0)
for ((i=0;i<${#files[*]};i++))
do
    echo "${files[$i]}" >> /tmp/trash
done

vi /tmp/trash

PTrenholme · 03-14-2010, 12:52 PM

Just curious: What does your "arbitrary character file name except null" code do about slash (/) characters in the file names? Oh, wait: You aren't generating the names, just trying to read them, so a slash, in addition to null, would be impossible.

All the << does is use the result of the find command as stdin until it's exhausted. See section 3.6 (Redirection) of info bash for details.

catkin · 03-14-2010, 01:04 PM

Quote:

Originally Posted by PTrenholme

All the << does is use the result of the find command as stdin until it's exhausted. See section 3.6 (Redirection) of info bash for details.

That's what I thought at first -- that < < is the same as << but I tried it and << generates a syntax error whereas < < does what is required.

I would also like to know what the parentheses in files+=("$file") do. Without them, when the /tmp/trash generated by the test script is viewed in vi all the file names are concatenated. With the parentheses, the file names are listed one per line. ???

EDIT: Have understood the parentheses. It is explained in The GNU Bash Reference, 3.6 Shell Parameters. <array name>+=<value> adds <value> to the existing contents of the last element of the array whereas <array name>+=(<value>) adds a new element to the array containing <value>. Nice!

blacky_5251 · 03-14-2010, 02:46 PM

The <(command) is known as process substitution. The results of "command" is temporarily stored in a file (/dev/fd) and then the file is used as input as if it had been classically redirected.

catkin · 03-14-2010, 02:58 PM

Quote:

Originally Posted by blacky_5251

The <(command) is known as process substitution. The results of "command" is temporarily stored in a file (/dev/fd) and then the file is used as input as if it had been classically redirected.

Thanks Ian

Given that info, found the reference at http://www.gnu.org/software/bash/man...s-Substitution. Always something to learn about bash!