CODEWITHSUNDEEP
linuxshellunixawksed
user2809888
user2809888

Reputation:

transpose the rows to column when pattern found

Here is the sample file which I was to process. I thought of using keyword "description" as some sort of RS but don't know how to do that ,also its not consistant.

Background: I am processing a log file which contains, date/time stamp (APR12) in first line and in second line there is a description about the log. This description is availabele for few logs and missig for few. 

001 APR12 aaa bbb
Description: This is a test file.
002 APR12 aaa bbb
Description: This is another test file.
003 APR12 aaa XXX
004 APR12 aaa bbb
Description: This is another,after skipping one.

Desired output:

001 APR12 aaa bbb Description: This is a test file.
002 APR12 aaa bbb Description: This is another test file.
003 APR12 aaa XXX
004 APR12 aaa bbb Description: This is another,after skipping one.

Upvotes: 1

Views: 771

Answers (6)

NeronLeVelu
NeronLeVelu

Reputation: 10039

sed ':a
N;$!ba
s/\n\([^0-9]\)/ \1/g' YourFile
  • Until you don't have huge file (loaded in memory).
  • join line that are not starting with number

If you have a GNU sed version after 4.2.2 that allow -z (-Z option). Thanks to @JJoao for this optimized code.

sed -z 's/\n\(^[0-9]\)/ \1/g' YourFile

Upvotes: 1

JJoao
JJoao

Reputation: 5337

perl -p0e 's!\n(?=Des)! !g'

(untested) -- It loads all the file into memory...

Upvotes: 0

potong
potong

Reputation: 58351

This might work for you (GNU sed):

sed 'N;s/\n\(Description\)/ \1/;P;D' file

Read pairs of lines throughout the file and replace the newline with a space if the second line of the pair begins Description.

Upvotes: 1

hek2mgl
hek2mgl

Reputation: 157947

Probably too complicated, but here comes a solution with sed:

# Does the line contain description?

# Yes ...
/Description/{
    # Exchange hold and pattern space
    x
    # Append hold space to pattern space
    # separated by newline 
    G
    # Remove that newline by a space
    s/\n\+/ /gp 
}

# No ...
/Description/! {
    # Exchange hold and pattern buffer
    x
    # The hold buffer contains a prefix line
    /Description/! {
        # Print it
        p
    }
    # Exchange hold and pattern buffer again
    x
    # Store current line in the hold buffer   
    h   
}

Upvotes: 0

Ed Morton
Ed Morton

Reputation: 203169

$ awk '{printf "%s%s", (/^[0-9]/?rs:FS), $0; rs=RS} END{print ""}' file
001 APR12 aaa bbb Description: This is a test file.
002 APR12 aaa bbb Description: This is another test file.
003 APR12 aaa XXX
004 APR12 aaa bbb Description: This is another,after skipping one.

Upvotes: 3

Tom Fenech
Tom Fenech

Reputation: 74596

You could add a newline every time the current line does not begin with "Description":

awk 'NR>1 && !/^Description/{print ""}{printf "%s ", $0}' file

The NR>1 prevents a newline from being added at the start of the output.

You may also want to add an END block to add a newline to the end of the output, if any lines were processed: END{if(NR)print ""}.

Upvotes: 1

Related Questions