I have multiple files with different job names. The job name is specified as follows.

#SBATCH --job-name=01_job1 #Set the job name

I want to use sed/awk/grep to automatically get the name, that is to say, what follows '--job-name=' and precedes the comment '#Set the job name'. For the example above, I want to get 01_job1. The job name could be longer for several files, and there are multiple = signs in following lines in the file.

I have tried using grep -oP "job-name=\s \K\w " file and get an empty output. I suspect that this doesn't work because there is no space between 'name=' and '01_job1', so they must be understood as a single word.

I also unsuccessfully tried using awk '{for (I=1;I<NF;I ) if ($I == "name=") print $(I 1)}' file, attempting to find the characters after 'name='.

Lastly, I also unsuccessfully tried sed -e 's/name=\(.*\)#Set/\1/' file to find the characters between 'name=' and the beginning of the comment '#Set'. I receive the whole file as my output when I attempt this.

I appreciate any guidance. Thank you!!

CodePudding user response：

Use this, you was close, just correctness of your grep -oP attempt (the main issue if you are trying to match a space after = character):

$ grep -oP -- '--job-name=\K\S ' file
01_job1

Explanations:

From man bash

--

A -- signals the end of options and disables further option processing.
Any arguments after the -- are treated as filenames and arguments. An argument of - is equivalent to --

That's because if the string you'd grep start with a -, grep try to treat it as a switch

The regex matches as follows:

NODE                     EXPLANATION
----------------------------------------------------------------------
--job-name=                '--job-name='
----------------------------------------------------------------------
  \K                       restart the match look-around 
----------------------------------------------------------------------
  \S                       non-whitespace (all but \n, \r, \t, \f,
                           and " ") (1 or more times (matching the
                           most amount possible))
----------------------------------------------------------------------

CodePudding user response：

You need to match the whole string with sed and capture just what you need to get, and use -n option with the p flag:

sed -n 's/.*name=\([^[:space:]]*\).*/\1/p'

See the online demo:

#!/bin/bash
s='#SBATCH --job-name=01_job1           #Set the job name'
sed -n 's/.*name=\([^[:space:]]*\).*/\1/p' <<< "$s"
# => 01_job1

Details:

• -n - suppresses default line output
• .* - any text
• name= - a literal name= string
• \([^[:space:]]*\) - Group 1 (\1): any zero or more chars other than whitespace
• .* - any text
• p - print the result of the successful substitution.

CodePudding user response：

Simlar to the answer of Gilles Quenot

grep -oP -- '--job-name=\K.*(?= *# *Set the job name)'

This adds a look-ahead to ensure that the string is followed by #Set the job name

CodePudding user response：

1st solution: In GNU awk with your shown samples please try following awk code.

awk -v RS=' --job-name=\\S ' 'RT && split(RT,arr,"="){print arr[2]}' Input_file

OR a non-one liner form of above GNU awk code would be:

awk -v RS=' --job-name=\\S ' '
RT && split(RT,arr,"="){
   print arr[2]
}
' Input_file

---

2nd solution: Using any awk please try following code.

awk -F'[[:space:]] |--job-name=' '{print $3}' Input_file

---

3rd solution: Using GNU grep please try following code with your shown samples and using non-greedy .*? approach here in regex.

grep -oP '^.*?--job-name=\K\S ' Input_file

CodePudding user response：

You can use a lookbehind and lookahead with GNU grep to get exactly what you describe:

grep -oP '(?<=--job-name=)\S (?=\s #Set the job name)' file

Or with awk:

awk '/^#SBATCH[[:space:]] --job-name=/ && 
     /#Set the job name$/ {
        sub(/^[^=]*=/,"")   
        sub(/#[^#]*$/,"")   
        print
     }' file 

Or perl:

perl -lnE 'say $1 if /(?<=--job-name=)(\S )(?=\s #Set the job name)/'   file    

Any prints:

01_job1