CODEWITHSUNDEEP
perlbashawksed
John
John

Reputation: 41

Linux :search for specific columns with particular string and print on screen

Its kind of daily record, sample below, though value format will not change but position of field/column of record keep changing which is creating a problem while using awk,sed,grep.

Filename.txt - with millions of record

abcd D20140624 Useragent username userid 
abcd D20140625 Useragent username1 userid1
D20140626 Useragent username2 userid2

result should be: 

D20140624 username userid
D20140625 username1 userid1
D20140626 username2 userid2

IF i use cat Filename.txt|awk -f ' ' '{print $2,$4,$5}' - invalid result

Similarly sed gives invalid result.

Can anyone help me with this.

Upvotes: 1

Views: 114

Answers (7)

Miller
Miller

Reputation: 35208

Using a perl one-liner, using indexes from the end of the fields:

perl -lane 'print "@F[-4,-2,-1]"' file

Or with more explicit logic:

perl -lane 'print @F == 5 ? "@F[1,3,4]" : "@F[0,2,3]"' file

Explanation:

Switches:

  • -l: Enable line ending processing, specifies line terminator
  • -a: Splits the line on space and loads them in an array @F
  • -n: Creates a while(<>){..} loop for each “line” in your input file.
  • -e: Tells perl to execute the code on command line.

Upvotes: 0

Ed Morton
Ed Morton

Reputation: 204458

-f is the argument to tell awk to read its script from a file so when you say "invalid result" I assume you're getting an error message like can't open source file ' '.

I THINK You probably were trying to use -F but then ' ' is the default FS value so there's no need to set it explicitly.

Once you've got past that issue, to get the output you want from that input file is just:

$ awk '{print $(NF-3), $(NF-1), $NF}' file
D20140624 username userid
D20140625 username1 userid1
D20140626 username2 userid2

Upvotes: 0

Kalanidhi
Kalanidhi

Reputation: 5092

You can also use sed command 

sed -r 's/.*(D[0-9]+) \w+ (.*)/\1 \2/g' file_name

Upvotes: 0

Avinash Raj
Avinash Raj

Reputation: 174826

Through GNU sed,

$ sed -r 's/^.*(D\S*).*(usern\S*).*(useri\S*).*/\1 \2 \3/g' file
D20140624 username userid
D20140625 username1 userid1
D20140626 username2 userid2

Upvotes: 0

Kent
Kent

Reputation: 195229

awk '{for(i=1;i<=NF;i++)if($i~/^D[0-9]{8}$/){n=i;break}}
     {print $n,$(NF-1),$NF}' file

gives:

D20140624 username userid
D20140625 username1 userid1
D20140626 username2 userid2

it searches for the the first column that matching D.... no matter where it is, print it and the last two column. you didn't specify the rule in detail, so I came with this.

Upvotes: 0

Farvardin
Farvardin

Reputation: 5424

use this:

awk '{ if(NF==5) print $2,$4,$5; else print $1,$3,$4; }'

Upvotes: 1

Jotne
Jotne

Reputation: 41460

You can do like this with awk

awk '!/^D20[0-9][0-9]/ {$1="";sub(/^ /,"")}1'
D20140624 Useragent username userid
D20140625 Useragent username1 userid1
D20140626 Useragent username2 userid2

If first field dos not start with a year, remove it and remove extra space.

Upvotes: 1

Related Questions