Here's an approach based on the observation of certain similarities (common prefix characters) in the data fields of interest in the three different types of data files. No discrimination between the three data file types is needed in the code.
Some notes of caution:
-
The code shown assumes the data being fed to it is valid. It is intended only as an example of a regex-based approach.
-
The code is critically dependent on the definition of the $rx_oct regex. (I gave it this name because it superficially suggests an IP octet.) The OP shows only limited examples of this sub-field in the range (10 .. 13). You (brad_nov) will have to change this regex to reflect the real data – or else maybe reveal an actual spec!
>perl -wMstrict -le
"my @records = (
'1|1212|34353|56575|||||4|~some~~pi=[10.10.10.10.10],uid=[11]}~',
'1|1212|34353|56575|||||4|~som~~390=10.10.10.10.11,391=222,394~',
'1|1212|34353|56575|||||4|~somedata~10.10.10.10.12~3333~~a~~~~',
);
;;
my $rx_oct = qr{ \d{1,3} }xms;
my $rx_quint = qr{ $rx_oct (?: \. $rx_oct){4} }xms;
;;
my $rx_dotted = qr{ (?<! \d) $rx_quint (?! \d) }xms;
my $rx_int = qr{ \d+ }xms;
;;
for my $record (@records) {
print qq{'$record'};
my ($const, $var) = $record =~ m{ ( \A .+) \| ( .* \z) }xms;
my (undef, $dotted, $int) =
$var =~ m{ (\D) ($rx_dotted) .*? \1 ($rx_int) }xms;
my $new_record = join '|', $const, $dotted, $int;
print qq{'$new_record' \n};
}
"
'1|1212|34353|56575|||||4|~some~~pi=[10.10.10.10.10],uid=[11]}~'
'1|1212|34353|56575|||||4|10.10.10.10.10|11'
'1|1212|34353|56575|||||4|~som~~390=10.10.10.10.11,391=222,394~'
'1|1212|34353|56575|||||4|10.10.10.10.11|222'
'1|1212|34353|56575|||||4|~somedata~10.10.10.10.12~3333~~a~~~~'
'1|1212|34353|56575|||||4|10.10.10.10.12|3333'
Update: After playing around with this a bit and doing a little, um, testing, I think I would change the definition of $rx_dotted as follows (change to final look-ahead):
my $rx_dotted = qr{ (?<! \d) $rx_quint (?! [.\d]) }xms;
This change does not affect behavior for valid records.