mjhjmtu

I'm renaming a bunch of music files, stripping of the track numbers using sed. However I'm puzzled by sed's behavior towards the '+' character for regex expressions.

What I wanted to do:

for f in *-[0-9]**; do echo "$f" | sed 's/-[0-9]+/-/g'; done

But that didn't work as intended. This workaround did the job:

for f in *-[0-9]**; do echo "$f" | sed 's/-[0-9][0-9]*/-/g'; done

However I would like to know what the issue is with using '+' in sed... Any ideas?

+ is an extended regular expression modifier. sed does basic regular expressions by default.

To make sed parse your expressions as extended regular expressions, use sed -E. GNU sed also understands + in basic regular expressions if you use it as +.

Also, your pattern *-[0-9]** looks strange. The - does not need escaping (not in the shell pattern and not in the regular expression), and unless you use globstar in bash, the ** pattern will expand to itself. With globstar enabled in bash *-[0-9]** would expand to all pathnames that contain -N in the first filename component (where N is a digit).

What I think you'd like to do is something like this (probably not quite, but I don't know what your filenames look like):

for name in *-[0-9][0-9]*; do
 newname=$name%-*
 printf 'Would rename "%s" into "%s"n' "$name" "$newname"
 # mv "$name" "$newname"
done

This matches all names in the current directory that contains -NN (N is a digit). It strips off everything from the last - in the filename and renames the file (the actual renaming is commented out).