I have a csv data file :

A,B,C,D,"A,B",E,"GG,H"
E,F,G,H,I,J,"S,P"

A|B|C|D|"A,B"|E|"GG,H"
E|F|G|H|I|J|"S,P"

Borrowing from Scrutinizer's idea...

awk 'NR%2{gsub(/,/,"|")}1;END{printf("\b")}' RS='"' ORS='"' infile

The post closes there with no result on awk.

Only the perl approach works, I am looking out for awk approach

perl -MText::ParseWords -nle'
  print join "|", parse_line(",",0, $_);
  ' infile

awk -F'"' '{gsub(/,/,"|",$1);gsub(/,/,"|",$3);} 1' yourFile

Have you seen my post???

yes thanks, that works fine but w/ nawk, btw can you please explain what NR%2 is doing here and what does '1' at the end mean in gsub

$ cat dd
A,"B,B",C,D,"A,B",E,"GG,H"
"E,F",F,G,H,I,J,"S,P"

$ nawk 'NR%2{gsub(/,/,"|")}1' RS='"' ORS='"' dd
A|"B,B"|C|D|"A,B"|E|"GG,H"
"E,F"|F|G|H|I|J|"S,P"

As you might have realised, the input (and also the output) record separator is a " . So, all odd-numbered records ( NR%2 ) are outside the double-quotes "realm" and all the , within these records need to be replaced with | . The other records do not need to be bothered about. The 1 at the end of the script prints all the records irrespective of the substitution.