CODEWITHSUNDEEP
linuxshellsedpipe
gx.
gx.

Reputation: 129

Can't get this simple sed command

I'm having a problem with sed and i can't figure it out, and i'm also amateur.

The objective of my code is to keep monitoring a file(OutputFile.dat) until possible strings(NaN or STOP) are found and then kill the program running on the background. It worked when I'd tried only one string. However when I'd tried to implement more possible matches the code didn't work out.

    ../program inputfile &> OutputFile.dat &
    tail -f OutputFile.dat | sed -n '/NaN/q;/STOP/q'
    killall program

I've tried a lot of different things but i can't figure this out. Alternative suggestions to achieve the same thing are also welcome.

Thanks in advance

Upvotes: 3

Views: 101

Answers (1)

that other guy
that other guy

Reputation: 123470

This is tricky due to buffering and race conditions. The current problems with your code are:

  1. OutputFile.dat is created asynchronously in the background and may not be there when tail tries to look for it.

    You can fix this by opening the file synchronously

  2. program may buffer output into larger chunks for efficiency reasons, because it doesn't think you care about timing.

    You can try to fix this by either asking program not to buffer, or by using stdbuf on GNU/Linux.

  3. After sed recognizes a match, tail will need to detect and write another buffer-full of data before it realizes that, exits, and lets the script continue.

    You can work around this by structuring your script to wait for sed but not for tail.

Since you say your script works for one keyword, the showstopper for you is likely #3, if your second keyword appears towards the end or a pause in the output.

All together, this would be:

#!/bin/bash
( stdbuf -o 0 -e 0 ../program inputfile & ) &> OutputFile.dat
sed -n '/NaN/q;/STOP/q' <(tail -f OutputFile.dat)
killall program

Upvotes: 4

Related Questions