i would strongly agree with others who have recommended use of a standard, well tested csv-parsing module.

however, if you really have to go commando (and do it on the command line to boot), the following (tested) code should work.

note that:

• NSRLFile.csv contains the example data from the op
• NSRLFile.psv contains the processed, pipe character-separated data
• the code makes no assumptions about the absence or presence whitespace around the value-separating commas or at the beginning or end of records
• double-quoted strings can contain anything, even backslash-escaped double-quotes (or any other backslash-escaped character)
• the only value supported other than a double-quoted string is a simple numeric with NO sign, decimal point or embedded commas
• the code makes no assumptions about the order of the fields in a record
• had to use \042 in place of a double-quote character because otherwise, the xp command-line interpreter gets confused and thinks the first pipe character it sees (in a regex) is an actual command-line pipe operator

perl -wlpe "use strict;   my $quo = qq(\042);  my $esc = qq(\\\\);
my $body = qr/[^$quo$esc]/;  my $escaped = qr/$esc ./x;  
my $sep = qr/ \s* , \s* | \s* \Z /x;
my $quoted = qr(  $quo $body* (?: $escaped $body* )* $quo )x;  
my $number = qr(\d+);
$_ = join '|', m/( $quoted | $number ) (?= $sep )/gx"  
NSRLFile.csv  > NSRLfile.psv
[download]