Double quotes vs asterisk filename expansion in Bash

Question

In the directories `~/temp/a/foo/` and `~/temp/b/foo foo/` I have some files named `bar1`, `bar2`, `bar bar1`, `bar bar2`, etc. I am trying to write a line of Bash that copies all these files in a directory containing "foo" as last part of the name to the folder above the respective "foo" folder. As long as there are no spaces in the file names, this is an easy task, but the two following commands fail when dealing with the `foo foo` directory: ``` for dir in `find . -type d -name '*foo'` ; do cp $dir/* "$(echo $dir|sed 's_foo__g')" ; done ``` (The `cp` command fails to see the last `foo` of `"foo foo"` as part of the same directory name.) ``` for dir in `find . -type d -name '*foo'` ; do cp "$dir/*" "$(echo $dir|sed 's_foo__g')" ; done ``` (`"$dir/*"` is not expanded.) Attempts like replacing `$dir/*` with `"$(echo $dir/*)"` have been even less successful. Is there an easy way to expand `$dir/*` so that `cp` understands?

Accepted Answer

Not only is a `for` loop wrong -- `sed` is also not the right tool for this job.

```
while IFS= read -r -d '' dir; do
  cp "$dir" "${dir/foo/}"
done < <(find . -type d -name '*foo' -print0)
```

Using `-print0` on the `find` (and `IFS= read -r -d ''`) ensures that filenames with newlines won't mess you up.

Using the `< <(...)` construct ensures that if the inside of your loop sets variables, changes directory state, or does similar things, those changes will survive (the right-hand side of a pipeline is in a subshell in bash, so piping into a while loop would mean that any changes to the shell's state made inside that while loop would be discarded on its exit otherwise).

Using `${dir/foo/}` is vastly more efficient than invoking `sed`, as it does the string replacement internal to bash.

Double quotes vs asterisk filename expansion in Bash

Solution

Problem