CODEWITHSUNDEEP
bashawksed
Matt Elson
Matt Elson

Reputation: 4355

How to extract lines after founding specific string

My example text is,

AA BB  CC
DDD
process.get('name1')
process.get('name2')
process.get('name3')
process.get('name4')
process.get('name5')
process.get('name6')
EEE
FFF
...

I want to search the string "process.get('name1')" first, if found then extract the lines from "process.get('name1')" to "process.get('name6')".

How do I extract the lines using sed?

Upvotes: 0

Views: 185

Answers (2)

Ed Morton
Ed Morton

Reputation: 203684

sed is for simple substitutions on individual lines, for anything more interesting you should be using awk:

$ awk -v beg="process.get('name1')" -v end="process.get('name6')" \
    'index($0,beg){f=1} f; index($0,end){f=0}' file
process.get('name1')
process.get('name2')
process.get('name3')
process.get('name4')
process.get('name5')
process.get('name6')

Note that you could use a range in awk, just like you are forced to in sed:

awk -v beg="process.get('name1')" -v end="process.get('name6')" \
        'index($0,beg),index($0,end)' file

and you could use regexps after escaping metachars in awk, just like you are forced to in sed:

awk "/process\.get\('name1'\)/,/process\.get\('name6'\)/" file

but the first awk version above using strings instead of regexps and a flag variable is simpler (in as much as you don't have to figure out which chars are/aren't RE metacharacters), more robust and more easily extensible in future.

It's important to note that sed CANNOT operate on strings, just regexps, so when you say "I want to search for a string" you should stop trying to force sed to behave as if it can do that.

Imagine your search strings are passed in to a script as positional parameters $1 and $2. With awk you'd just init the awk variables from them in the expected way:

awk -v beg="$1" -v end="$2" 'index($0,beg){f=1} f; index($0,end){f=0}' file

whereas with sed you'd have to do something like:

beg=$(sed 's/[^^]/[&]/g; s/\^/\\^/g' <<< "$1")
end=$(sed 's/[^^]/[&]/g; s/\^/\\^/g' <<< "$2")
sed -n "/^${beg}$/,/^${end}$/p" file

to deactivate any metacharacters present. See Is it possible to escape regex metacharacters reliably with sed for details on escaping RE metachars for sed.

Finally - as mentioned above you COULD use a range expression with strings in awk:

awk -v beg="$1" -v end="$2" 'index($0,beg),index($0,end)' file

but I personally have never found that useful, there's always some slight requirements change comes along to make me wish I'd started out using a flag. See Is a /start/,/end/ range expression ever useful in awk? for details on that

Upvotes: 2

mauro
mauro

Reputation: 5950

This should work and... it uses sed as per OP request:

$ sed -n "/^process\.get('name1')$/,/^process\.get('name6')$/p" file

Upvotes: 3

Related Questions