detailed bash script with cp and grep will not work with spaces in filenames

Question

I would never do it myself but people with Windows machines insist on putting spaces in filenames.

I've composed this detailed command which works perfectly except for the files with spaces in them. Have tried everything, single quotes, double quotes, ticks, escape with backslash.

The command is supposed to copy everything from a directory with certain file extensions, except a list of files which I don't want copied. Unfortunately some of those files contain spaces. This is the command:

cp $(ls *.txt *.docx | grep --invert-match --fixed-strings \
-e not_this_one.txt \
-e not_this_one_either.docx \
-e "no not me.txt" \
-e "please leave me out as well.docx")  ./destination_directory/

Any ideas how to make this work?

Answer 1

Use find and xargs instead of $(...) parameter expansion:

find *.txt *.docx \( \
      -name not_this_one.txt \
      -o -name not_this_one_either.docx \
      -o -name 'no not me.txt' \
      -o -name "please leave me out as well.docx" \
    \) -prune -o -print0 |
  xargs -0 cp -t /tmp/destination_directory

We use the -prune option to exclude things we don't want to copy.

We use -print0 on the find command to produce NUL-terminated filenames, which when piped to xargs -0 correctly handles filenames that contain whitespace.

Finally, we use the -t <target_directory> option on cp because this allows xargs to just append a list of filenames to the command (without the -t, the destination directory needs to come last, which complicates things a bit).

---

Or, use tar:

tar -cf- \
    --exclude=not_this_one.txt \
    --exclude='not_this_one_either.docx' \
    --exclude='no not me.txt' \
    --exclude='please leave me out as well.docx' *.txt *.docx |
  tar -C /tmp/destination_directory -xf-

(And of course, you could put your list of exclude patterns into a file and use --exclude-from instead.)

Answer 2

Once you send ls *.txt *.docx tru a pipe | the shell is generating an string where files are not separated. Well, they could be separated by spaces (or newlines), which leads to confusion of the next tool that try to use such string splitting filenames on spaces (or newlines).

Since you are tagging the question with bash, the expansion of the globs *.txt and *.docx is done by the shell (bash), and bash is able to reject some file patterns given in GLOBIGNORE, you can simply do:

( GLOBIGNORE='not_this_one.txt:not_this_one_either.docx:no not me.txt:please leave me out as well.docx';
  echo \
  cp -t /tmp/destination_directory *.txt *.docx
)

The (...) here is to ensure that the setting of GLOBIGNORE is erased once the subshell is closed. Of course, you can remove it and do a unset GLOBIGNORE after. The echo is there to show you the result of the command. If you are happy with what you get, remove the echo.