Understanding a sed line in a bash script.

[Erik1984]

I found a script to generate moodbars for Amarok http://wiki.ubuntuusers.de/Moodbar#Unter-Amarok.

The complete script:

#!/bin/bash
control_c()        # Den Vorgang abbrechen, wenn der Nutzer Strg+c drückt
{
 echo $1 > lastreadsong
 echo "Exiting..."
 exit
}
if [ -e lastreadsong ]; then
  read filetodelete < lastreadsong
  rm "$filetodelete" lastreadsong
fi
find . -type f -regextype posix-awk -iregex '.*\.(mp3|ogg|flac|wma)' | while read i
do
  trap 'control_c "$OUTF"' SIGINT
  TEMP="${i%.*}.mood"
  OUTF=`echo "$TEMP" | sed 's#\(.*\)/\([^,]*\)#\1/.\2#'`
  if [ ! -e "$OUTF" ]; then
    moodbar -o "$OUTF" "$i"
  fi
done

I understand what the script does but I don't see what the sed line does exactly. This is the part I'm interested in:

  TEMP="${i%.*}.mood"
  OUTF=`echo "$TEMP" | sed 's#\(.*\)/\([^,]*\)#\1/.\2#'`

How does it work? All that seems necessary is to add a "." before the filename, other than that TEMP seems to contain the right value.

[Vaphell]

TEMP line strips shortest possible rightside (%) '.'+anything (.*) - read: extension - and adds .mood
OUTF line pipes $TEMP value to sed which splits the whole path in 2 parts with the last /, effectively cutting it to path (\1) and filename (\2) parts. Then it glues them back with /. which effectively means the filename got prefixed with .
with -r parameter the picture is clearer
sed -r 's#(.*)/([^,]*)#\1/.\2#'

Why the filename part is matched by not-colon any number of times [^,]* - no idea.

$ i=/some/path/file.mp3
$ TEMP="${i%.*}.mood"
$ echo "$TEMP"
/some/path/file.mood
$ OUTF=$( echo "$TEMP" | sed -r 's#(.*)/([^,]*)#\1/.\2#' )
$ echo "$OUTF"
/some/path/.file.mood

the same behavior in pure bash using trimming shortest rightside (%) /* pattern and longest leftside (##) */ pattern. More efficient than creating separate sed process for each file

$ OUTF=${TEMP%/*}/.${TEMP##*/}
$ echo "$OUTF"
/some/path/.file.mood

[ofnuts]

Would not that be simpler (or at least more readable)?

OUTF=$(dirname $i)/.$(basename $i .mp3).mood

[Erik1984]

Thanks Vaphell and ofnuts! Well especially Vaphell here for the extensive explanation

Your solutions are a lot friendlier on the eyes let's say regular expressions are not my strongest point.

@ofnuts
Your solution looks elegant but the problem is that my collection contains music with different extensions. it could be .mp3 but also .ogg .wma (one pesky folder) or .flac

[Crusty Old Fart]

...Why the filename part is matched by not-colon any number of times [^,]* - no idea...

Well, not to be a nitpicker, but to correctly name the character for what it is, because I'll be referring to it later: it's a comma, not a colon.
The negated character class: [^,]* matches any non-comma character zero or more times. The author may know that none of his filename parts contain commas. So this allows him to match the entire filename part.
It might be argued that he could accomplish the same match by using: .*
However, the dot metacharacter does not actually match any single character. It matches any single character EXCEPT for newline characters. In essence, the dot metacharacter in Unix regex is short for: [^\n]
So, assuming the author knows that none of his filename parts contain commas, and he wants his match to include newline characters, then the negated character class: [^,]* works for him whereas: .* does not.
(Ref.: http://www.regular-expressions.info/dot.html)

Now everybody can sleep tonight

[Vaphell]

my bad, i always have trouble recalling english names of ,:;

[Erik1984]

Thanks again I'll mark this as solved as I have successfully executed the script with Vaphells code instead of the sed line.

musicdir="$HOME/Music/"
find $musicdir -type f -regextype posix-awk -iregex '.*\.(mp3|ogg|flac|wma|m4a)' | while read i
do
    TEMP="${i%.*}.mood"
    OUTF=${TEMP%/*}/.${TEMP##*/}    
    if [ ! -e "$OUTF" ];
    then
        moodbar -o "$OUTF" "$i"
    fi    
done

[Vaphell]

home/$USER is less rocksolid than simple $HOME (you don't have to have your homedir in /home, root cerainly doesn't) but $HOME will always point to the right place)

[Erik1984]

Thanks, changed it. I put $USER there to 'censor' my username which is in the script on my machine. $HOME seems a better choice to distribute the script for common use.

[ofnuts]

Thanks Vaphell and ofnuts! Well especially Vaphell here for the extensive explanation

Your solutions are a lot friendlier on the eyes let's say regular expressions are not my strongest point.

@ofnuts
Your solution looks elegant but the problem is that my collection contains music with different extensions. it could be .mp3 but also .ogg .wma (one pesky folder) or .flac

Hmmm Then technically nothing says you won't get "foo.mp3" and "foo.ogg" in the same folder, so the extension shouldn't be removed (.foo.mp3.mood).