CODEWITHSUNDEEP
bashawksed
ziggurat
ziggurat

Reputation: 3

How do I extract words between matching characters and combine them into new lines in awk/sed?

any idea how I can remove all the formatting from these responses and preserve only the 'values' portion separated by comma? I've achieved something similar but doing multiple runs and separate scripts.

response:
{
    type => 'query'
    timestamp => '1444304880'
    serial => '0000012970'
    address => '192.168.1.1'
    profile => 'common'
    query-id => '001'
    flags => '(NET, CORP)'
    version => '1.0.0.3'
}
response:
{
    type => 'query'
    timestamp => '1444305643'
    serial => '0000012971'
    address => '192.168.1.2'
    profile => 'common'
    query-id => '002'
    flags => '(CORP)'
    version => '1.0.0.3'
}

Ideal output:

query, 1444304880, 0000012970, 192.168.1.1, common, 001, (NET, CORP), 1.0.0.3
query, 1444305643, 0000012971, 192.168.1.2, common, 002, (CORP), 1.0.0.3

I noticed I can go two ways, first one is simply print out the $3 and $4 columns: awk '{print $3, $4}' dump.txt That gives me:

'query'
'1444304880'
'0000012970'
'192.168.1.1'

But it also includes the spaces created by the '{ }' which I can eliminate. My other option is to go this way and take out the response {} section.

sed "s/\'//g" dump.txt | awk '/\{/{flag=1;next}/\}/{flag=0}flag'

But then I have to bring each line up using: 

sed -e '/type/{N;s/\n//;}'

Any help in doing this in a nicer way is appreciated.

Upvotes: 0

Views: 53

Answers (4)

Jose Ricardo Bustos M.
Jose Ricardo Bustos M.

Reputation: 8174

Another solution, using gnu-awk and FPAT

awk -vFPAT="['][^\n]+[']" -vRS="{" -vOFS="," '
    NR>1{$1=$1; gsub(/\x27/,""); print}' file

you get,

query,1444304880,0000012970,192.168.1.1,common,001,(NET, CORP),1.0.0.3
query,1444305643,0000012971,192.168.1.2,common,002,(CORP),1.0.0.3

Upvotes: 0

glenn jackman
glenn jackman

Reputation: 247022

awk: uses "=>" as the field separator

awk -F "=>" '
    # a line with 2 fields, remove single quotes and print with a comma
    NF == 2 {gsub(/\x27/, "", $2); printf "%s,", $2}
    # end of record, overwrite the trailing command and add a newline
    $0 == "}" {printf "\b \n"}   
' file

Upvotes: 0

Cyrus
Cyrus

Reputation: 88766

With GNU grep and paste:

grep -Po "=> '\K.*(?=')" file | paste -d , - - - - - - - -

Output:

query,1444304880,0000012970,192.168.1.1,common,001,(NET, CORP),1.0.0.3
query,1444305643,0000012971,192.168.1.2,common,002,(CORP),1.0.0.3

Upvotes: 2

pcantalupo
pcantalupo

Reputation: 2226

Assuming that all your input is structured as in your example, this should work:

cut -d '>' -f 2 foo.txt | grep "^ " | paste -d, - - - - - - - - | tr -d "'" | sed 's/^ //'

Upvotes: 0

Related Questions