regex - lazy and non-capturing - sql

Question

String to search:
VALUES ('9gfdg', to_date('1876/12/06','YYYY/MM/DD'), null)
Regex search so far:
VALUES\s*\(\s*'?\s*(.+?)\s*'?\s*,\s*'?\s*(.+?)\s*'?\s*,\s*'?\s*(.+?)\s*'?\s*\)
Regex replace to 3 groups: ie \1 \2 \3
I am aiming for a result of:
9gfdg to_date('1876/12/06' ,'YYYY/MM/DD') null
but instead get (because of that extra comma in to_Date and also lazy instead of greedy):
9gfdg to_date('1876/12/06 YYYY/MM/DD , null)
Note:
It is exactly 3 fields (the values within th 3 fields may be different but you get the idea of the format I am grappling with). ie each of the fields could have commas (usually character values, could be a keyword such as null, could be a number or could be a to_Date expression.
Regex engine is VBA/VBscript
Anyone have any pointers on fixing up this regex?

Answer 1

Here is a solution.
Notice the regex for $field: it is yet another application of the normal* (special normal*)* pattern, with normal being anything but a comma ([^,]) and special a comma as long as it is not followed by two single quotes (,(?!'')). The first normal, however, is made non empty using + instead of *.
Demonstration code in perl. The string concatenation operator in perl is a dot:
fge#erwin $ cat t.pl
#!/usr/bin/perl -W
use strict;
# Value separator: a comma optionally surrounded by spaces
my $value_separator = '\s*,\s*';
# Literal "null", and a number
my $null = 'null';
my $number = '\d+';
# Text field
my $normal = '[^,]'; # Anything but a comma
my $special = ",(?!'')"; # A comma, _not_ followed by two single quotes
my $field = "'$normal+(?:$special$normal*)*'"; # a text field
# A to_date() expression
my $to_date = 'to_date\(\s*' . $field . $value_separator . $field . '\s*\)';
# Any field
my $any_field = '(' . $null . '|' . $number . '|' . $field . '|' . $to_date . ')';
# The full regex
my $full_regex = '^\s*VALUES\s*\(\s*' . $any_field . $value_separator . $any_field
. $value_separator . $any_field . '\s*\)\s*$';
# This builds a compiled form of the regex
my $re = qr/$full_regex/;
# Read from stdin, try and match (m//), if match, print the three captured groups
while (<STDIN>) {
m/$re/ and print <<EOF;
Argument 1: -->$1<--
Argument 2: -->$2<--
Argument 3: -->$3<--
EOF
}
Demonstration output:
fge#erwin ~ $ perl t.pl
VALUES ('9gfdg', to_date('1876/12/06','YYYY/MM/DD'), null)
Argument 1: -->'9gfdg'<--
Argument 2: -->to_date('1876/12/06','YYYY/MM/DD')<--
Argument 3: -->null<--
VALUES('prout', 'ma', 'chere')
Argument 1: -->'prout'<--
Argument 2: -->'ma'<--
Argument 3: -->'chere'<--
VALUES(324, 'Aiie, a comma', to_date('whatever', 'is there, even commas'))
Argument 1: -->324<--
Argument 2: -->'Aiie, a comma'<--
Argument 3: -->to_date('whatever', 'is there, even commas')<--
One thing to note: you will notice that I don't ever use any lazy quantifiers, and not even the dot!
edit: special in a field is actually a comma not followed by two single quotes, not one

Answer 2

If only the second parameter can have commas in it, you could do something like:
^VALUES\s*\(\s*'?([^',]*)'?\s*,\s*(.*?)\s*,\s*'?([^',]*)'?\s*\)$
Otherwise I don't know what features that regex flavor supports, so hard to make something more fun. Altho you could always make a limited depth nested parentheses regex if (?R) is not supported.

Answer 3

For the more general case you could try something like:
^\s*
VALUES\s*
\(
\s*
(?: '([^']*)' | ( \w+ (?: \( [^()]* \) )? ) )
\s*,\s*
(?: '([^']*)' | ( \w+ (?: \( [^()]* \) )? ) )
\s*,\s*
(?: '([^']*)' | ( \w+ (?: \( [^()]* \) )? ) )
\s*
\)\s*
$
Spaces removed:
^\s*VALUES\s*\(\s*(?:'([^']*)'|(\w+(?:\([^()]*\))?))\s*,\s*(?:'([^']*)'|(\w+(?:\([^()]*\))?))\s*,\s*(?:'([^']*)'|(\w+(?:\([^()]*\))?))\s*\)\s*$
Replace with:
\1\2 \3\4 \5\6
Should work for one nested level of parentheses without any quoted parenthesis in them.
PS: Not tested. You can usually use the spaced regex if your flavor supports the /x flag.