Shell Programming and Scripting

I have a script like this:

while read abbrev; do

sed 's/'$(echo "$abbrev" | tr "[a-z]" "[A-Z]")'/'"$abbrev"'/g' 'output_rgt.'$$ >'tmp.'$$
mv 'tmp.'$$ 'output_rgt.'$$

done<'dict.shortcuts.'$$

And I don't know how to leave "mv" out. ('dict.shortcuts.'$$ may be potentially long)

Thank you for any suggestions!

EDIT1: I know I can use two variables and save what is 'source' file and what is 'destination' file but it's not elegant too.

EDIT2:

I've rewritten it like this:

while read abbrev; do

# lookedUpAbbrev= ..
# protectedAbbred=...

echo 's/ '"$lookedUpAbbrev"'\(.*\..*\)/ {'"$protectedAbbrev"'}\1 /g;' >>'dict.commands.'$$

done<'dict.shortcuts.'$$

cat 'output_rgt.'$$ | sed -f 'dict.commands.'$$ >'tmp.'$$
mv 'tmp.'$$ 'output_rgt.'$$

I dunno. There are ways to avoid using a temporary file but sometimes temporary files are just the most robust way to get things done. Just make sure you set up the proper traps to remove the temporary when the script ends, and also if it's interrupted.

Another way whitout temporary file for sed commands :

Code:

while read abbrev; do

   # lookedUpAbbrev= ..
   # protectedAbbred=...

   echo "s/$lookedUpAbbrev\(.*\..*\)/ {$protectedAbbrev}\1 /g;" 

done <'dict.shortcuts.'$$ |
sed -f -  output_rgt.$$ >'tmp.'$$
mv 'tmp.'$$ 'output_rgt.'$$

Jean-Pierre.

thank you!

I've adjusted the script as Jean-Pierre advised but it's still too slow:

I've a file with 480 abbreviations so that sed works with 480 substitutions and file output_rgt.$$ may have 120kB (subtitles of a movie) and it's terribly slow. Isn't there any faster way how to perform these substitions?

Moving it all into a single Perl (or perhaps awk) script would reduce the overhead of spawning subprocesses to perform what seem to be fairly mundane transformations of the input. That's just a speculation, though. Can you split it up in two parts and time them? Does generating the sed script take a lot of time, and running it in sed go quickly, or is it the other way around? If the generation part dominates then I'd definitely try tackling that in Perl (and then doing the rest in Perl as well makes sense).

Another thing: the bottleneck might be the regexes. Does it help if you replace .*\..* with (say) [^.]*\.[^.]* (assuming this is an acceptable trade for your scenario, of course)?