grep output of a text file

Shell Programming and Scripting
1

Hi. I have a unix script 'if' condition that greps through a file and searches for the strings ERROR and WARNING.

if egrep -q 'ERROR|WARNING' myfile.txt; then

When this 'if' condition is true then I want to be able to capture 10 lines before the string and 10 lines after the string. So a total of only 20 lines.

Can anyone help me with a way to grep through a files and then output only 20 lines of the file (10 before the string - 10 after) that surround the string when it's found?

Any help would be greatly appreciated!

2 
grep -A 10 -B 10 -E 'ERROR|WARNING' myfile.txt

?

3

Hmm, just returning:

grep -A 10 -B 10 -E 'ERROR' myfile.txt
grep: Not a recognized flag: A
4

Getting lines after it is doable since grep has a special option to help with that, -m. That will cause it to stop reading after the number of matches you tell it.

#!/bin/sh

# open the file
exec 5<filename

# Read up to the first match.
# because of -m, grep will leave the open file on the line after it.
# Then print 10 more lines.
( grep -m 1 "MATCH" && head -n 10 ) <&5

# close the file
exec 5<&-

Getting the lines before it is much harder, since a shell can't seek backwards.

---------- Post updated at 11:18 AM ---------- Previous update was at 11:17 AM ----------

You never said what your system was. That's a GNU option and I guess you don't have GNU.

5

Hmm...my system is AIX 5.3. Is this what you mean? Does this help?

6

Yes, it means you don't have GNU grep, that's usually found in Linux, though occasionally installed elsewhere as a third-party app for its convenient options. Here's a brute-force non-GNU solution.

#!/bin/ksh

TMP="/tmp/$$"

while read LINE
do
        if [[ "${LINE}" == *ERROR* ]] || [ "${LINE}" == *WARNING* ]]
        then
                tail -n 10 "$TMP"
                echo "$LINE"
                head -n 10
                : > "$TMP"
        else
                echo "$LINE" >> "$TMP"
        fi
done < inputfile

rm -f "$TMP"

It's not perfect. If more than one error happens within the 10 lines it might not capture all the context.

---------- Post updated at 11:47 AM ---------- Previous update was at 11:32 AM ----------

#!/bin/ksh

TMP="/tmp/$$"

CONTEXT=0

while read LINE
do
        if [[ "${LINE}" == *ERROR* ]] || [ "${LINE}" == *WARNING* ]]
        then
		[[ "${CONTEXT}" -le 0 ]] && tail -n 10 "$TMP"
		CONTEXT=10
        fi

	if [[ "${CONTEXT}" -gt 0 ]]
	then
		echo "${LINE}"
	fi

        echo "${LINE}" >> "$TMP"

	((CONTEXT--))

done < inputfile

rm -f "$TMP"

This one should make better output, but is slower.

1 Like
7

or try this one :