You can do a loop, and use $ to turn a field number into field contents:
for(N=1; N<=NF; N++) sub(/ /, "", $N);
...but I suspect you're overthinking this a touch -- why bother using fields at all? Do a search/replace on the line itself and you'll get it all done in one gsub:
awk '{ gsub(/[|] [|]/, "||") } 1' inputfile
---------- Post updated at 10:04 AM ---------- Previous update was at 10:01 AM ----------
Okay, that doesn't work because gsub refuses to overlap like that. But my first example still works when you set FS and OFS properly.
The regex might as well be /^ $/ for that case, to prevent it eating spaces in the middle of a string.
I just want to do on few fields alone. If othert fields have space alone, that wont be a problem...I have to do a search and replace on selective fields alone
The "3|4" is the list of fields you want to replace here. It's split() into the array F so that F[1]=3, F[2]=4, using the same separator as everything else... Then we loop over that array, for(N in F) which should do two loops with N=1 and N=2 respectively. We look up the field in F, convert it to field contents with $, and substitute on it.
I think just plain sub() should be fine here, since you don't need to do more than one substitution per token.
