Shell Programming and Scripting

@ andy391791: Did you run that scriptlet on the sample file in post#1? It will print exactly the desired fields. And, if extended by ORS=";", it will print the desired single line. The residual flaw is the trailing semicolon...

The RS= is an initial variable assignment before opening the file.
The main loop only has the condition NR%2==0 where the default action is {print}.

Not so easy in /bin/sh, but quite straightforward in awk (horses for courses):

Code:

#!/usr/bin/awk -f

BEGIN {
    FS=":"
}

{
    join()
    next
}

END {
    printf "\n"
}

function join() {
    gsub("\"","",$1)
    printf "%s%s", semi, $1
    semi = ";"
}

That will put it all on one line; if you want lines joined in batches of three, that is a little more work. The first step to solving a problem is specifying it precisely!

Assuming that there is only one quoted string before the colon (as in the sample input), it is also easy with any POSIX-conforming shell:

Code:

#!/bin/ksh
while IFS='"' read junk data junk
do	printf '%s%s' "$semi" "$data"
	semi=';'
done < "file.txt"
echo

This was written using a Korn shell and tested with ksh and bash with both of them producing the requested results:

Code:

Word1;Word2;This is Word3

Running it with zsh also produced the same results even though zsh obviously does not conform to the standard's quoting requirements for a shell (as mentioned in post #6 in this thread).