Convert all file extensions to lower-case

Question

I'm trying to lower-case all my extensions regardless of what it is. So far, from what I've seen, you have to specify what file extensions you want to convert to lower-case. However, I just want to lower-case everything after the first last dot . in the name.

How can I do that in bash?

Answer 1

Solution

You can solve the task in one line:

find . -name '*.*' -exec sh -c '
  a=$(echo "$0" | sed -r "s/([^.]*)\$/\L\1/");
  [ "$a" != "$0" ] && mv "$0" "$a" ' {} \;

Note: this will break for filenames that contain newlines. But bear with me for now.

Example of usage

$ mkdir C; touch 1.TXT a.TXT B.TXT C/D.TXT
$ find .
.
./C
./C/D.TXT
./1.TXT
./a.TXT
./B.TXT

$ find . -name '*.*' -exec sh -c 'a=$(echo "$0" | sed -r "s/([^.]*)\$/\L\1/"); [ "$a" != "$0" ] && mv "$0" "$a" ' {} \;

$ find .
.
./C
./C/D.txt
./a.txt
./B.txt
./1.txt

Explanation

You find all files in current directory (.) that have period . in its name (-name '*.*') and run the command for each file:

a=$(echo "$0" | sed -r "s/([^.]*)\$/\L\1/");
[ "$a" != "$0" ] && mv "{}" "$a"

That command means: try to convert file extension to lowercase (that makes sed):

$ echo 1.txt | sed -r "s/([^.]*)\$/\L\1/"
1.txt
$ echo 2.TXT | sed -r "s/([^.]*)\$/\L\1/"
2.txt

and save the result to the a variable.

If something was changed [ "$a" != "$0" ], rename the file mv "$0" "$a".

The name of the file being processed ({}) passed to sh -c as its additional argument and it is seen inside the command line as $0. It makes the script safe, because in this case the shell take {} as a data, not as a code-part, as when it is specified directly in the command line. (I thank @gniourf_gniourf for pointing me at this really important issue).

As you can see, if you use {} directly in the script, it's possible to have some shell-injections in the filenames, something like:

; rm -rf * ;

In this case the injection will be considered by the shell as a part of the code and they will be executed.

While-version

Clearer, but a little bit longer, version of the script:

find . -name '*.*' | while IFS= read -r f
do
  a=$(echo "$f" | sed -r "s/([^.]*)\$/\L\1/");
  [ "$a" != "$f" ] && mv "$f" "$a"
done

This still breaks for filenames containing newlines. To fix this issue, you need to have a find that supports -print0 (like GNU find) and Bash (so that read supports the -d delimiter switch):

find . -name '*.*' -print0 | while IFS= read -r -d '' f
do
  a=$(echo "$f" | sed -r "s/([^.]*)\$/\L\1/");
  [ "$a" != "$f" ] && mv "$f" "$a"
done

This still breaks for files that contain trailing newlines (as they will be absorbed by the a=$(...) subshell. If you really want a foolproof method (and you should!), with a recent version of Bash (Bash≥4.0) that supports the ,, parameter expansion here's the ultimate solution:

find . -name '*.*' -print0 | while IFS= read -r -d '' f
do
  base=${f%.*}
  ext=${f##*.}
  a=$base.${ext,,}
  [ "$a" != "$f" ] && mv -- "$f" "$a"
done

Back to the original solution

Or in one find go (back to the original solution with some fixes that makes it really foolproof):

find . -name '*.*' -type f -exec bash -c 'base=${0%.*} ext=${0##*.} a=$base.${ext,,}; [ "$a" != "$0" ] && mv -- "$0" "$a"' {} \;

I added -type f so that only regular files are renamed. Without this, you could still have problems if directory names are renamed before file names. If you also want to rename directories (and links, pipes, etc.) you should use -depth:

find . -depth -name '*.*' -type f -exec bash -c 'base=${0%.*} ext=${0##*.} a=$base.${ext,,}; [ "$a" != "$0" ] && mv -- "$0" "$a"' {} \;

so that find performs a depth-first search.

You may argue that it's not efficient to spawn a bash process for each file found. That's correct, and the previous loop version would then be better.