UNIX for Dummies Questions & Answers

Is this where you get hung up?

for u in *.ABC; do for t in $u.raw.sedfile*; do sed -f $t $u > temp; mv temp $u; done; done &&

No matter, try this:


Regarding your large sed file that you call with "sed -f", can you instead do something like this instead.

for name in `cat large.sed.file`
do
echo changing $name
sed $name
done > output.file

You would need to have the sed command there to read in each line one at a time from the "for" line.

Hope this helps.

That would work - however with sed requiring that the output be directed to a new file - each line of the sedfile that is applied would require the result to be > output.file then renamed back to the original file to continue.

I think that would become very inefficient - as I'd be doing a mv for each line in the sed file (1,000's of them.) With sed -f I can do 199 at a time before I need to do an mv.

So a file with 10,000 commands would, with your suggestion, make 10,000 sed commands and 10,000 mv commands. The way I have it it makes 50 sed -f commands (arguably still is doing 10,000) - but only 50 mv commands. That's the difference I think.

I can see the benefit - but think that the cost in processing speed outweighs the change in favour of the current script.

The script does work. There's no problem with it. What I really wanted was the ability to have sed -f process more than 199 commands at a time. I figure from the responses to date that this is not possible (built into sed I assume). So might just leave it as it is.

after banging away all i found is that it is probably version specific, i used a file with 300+ sed commands and it ran fine, i'm not sure how far you want to go but you could try a different version of sed, unfortunately i'm not sure what version i'm using, sed -V is supposed to give it but not on my aix 4.3

later

Thanks everyone for looking at it.

Its not really worth my effort to locate another version of sed - considering the script works - and to be of any benefit - I'd really need a sed that could read unlimited lines.

No point moving to 500 lines or even 1,000 as I'd still have the use the script I have.

One more thing.

You could use an append for the last line of my suggestion.

for name in `cat large.sed.file`
do
echo changing $name
sed $name
done >> output.file

This would allow you to append unlimited lines to this output file before you decide to do a mv back to the original file.

Give it a look. Even if you don't change your script.

I might be mistaken - but won't that append the whole file each time?

The output of the sed command is the whole file with it's changes.

If I then append the output to a file - and follow up with with the next loop - won't it just add the entire file with a few lines changed again - onto the end ot the file....hence a file that is 'n' times the length of the sed file?

I see where you're coming from - I have considered a number of options - and so far as I can see what I have is the most efficient method.

The only way I could improve it is to have my sed file with multiple substitutions on one line..... -e s/old/new/g -e s/old/new/g ...

Then I may be able to call upto 199 lines (each line however would execute maybe 20 commands.....then I could do it with one file........

That however would entail me creating a file with 20 columns rather than one - which is a whole lot of manipulation - to acheive something that already works..... might have a play to see if it works anyway!

Theoretically, I believe that the for loop allows sed to process one line at a time so it would loop thru your large file one line at a time and append one line at a time. Also, I believe that sed does process the whole file at once but I am not sure what your script would do being run thru a for loop.

Let me know what happens.