Bash – Grep by block of text\lines

bashscriptingshellshell-scripting

I have the text, that contains some lines. So, i need to make GREP of several lines. For example, i have repeating text and i should GREP get lines, that have this repeating key-words.

grep -o "test|test2" textfile

My text:

123|never for your|test
123421|never for your|test2
123412|never for your|test3
12341|never for your|test4
12311|never for your|test2
123312312|never for your|test
123321312|never for your|test2

I should have:

123|never for your|test
123421|never for your|test2
123312312|never for your|test
123321312|never for your|test2

It works, but it doesn't work how i want. It searchs in text, all words "test" & "test2". But i want to get textblocks, like some pattern, where only after "test" comes "test2".
Have you got any ideas ?

Best Answer

Brief shell script using sed. Makes a list of line numbers for the second case, and compares against line numbers for the first case. Prints matching pairs. Uses the first argument as the file name. Could easily be extended to take second and third arguments as patterns to match. Could save as findnext.sh, and run:

$ sh findnext.sh testfile

Should be quick as it only involves two passes over the file, and has the advantage of being completely portable.

#!/bin/sh 
# Line numbers matching test1
mt2=$(sed -ne '/test1/=' < $1 | tr '\n' '/')

for l in $(sed -ne '/test/=' < $1); do
    nextline=$(expr $l + 1)
    [ "${mt2#*$nextline/}" != "$mt2" ] && sed -ne $l,${nextline}p <$1
done

Related Solutions

Grep’ing through mailq

What about:

grep -C 1 said:

-C is context and returns one line before & after "said:". Also, for future reference, -A is for after, -B is for before.

grep -C 1 said: | grep ^[0-9] | cut -f1 -d " "

(for the list of IDs)

If you need something more complex, then you would have to use awk.

While loop read multiple lines from a grep

Zoredache pointed this out to me in chat, and poige mentioned it in his answer: this problem can be solved with a subshell.

When I had to change from a for loop that read a single variable from my grep at a time to a while read var1 var2 loop that allowed me to read in multiple variables, I was able to hang on to temporary variables I manipulated within the while loop by using parentheses to define an explicit subshell. Here's an example:

sum=0

grep volume configfile | head -n1 | (while read var1 var2; do
  let sum=var1+var2
done

echo "The sum is $sum.")

Without the parentheses, you will always echo a sum of 0. With them, you will echo the sum of the first two values in the first matching line of your grep.

Additionally, as Poige points out in another answer, you can use a subshell to populate an in-scope variable like this:

var=$( cat file | while read a b; do
    sum=a+b
    echo "$sum"
done)

echo "$var"

In this case, the value of var that you echo at the end would be the last sum you calculated in your loop, even though sum got destroyed at the end of the subshell.

Best Answer

Related Solutions

Grep’ing through mailq

While loop read multiple lines from a grep

Related Topic