Re^2: Analyzing regular expression performance

[Sprad]

Here's the one that inspired me to ask:

$sql = "
    create table #foo (variable1    char(20),
               variable2    char(30)    null)

    /*
    comments make it slow
    */
";


my $comment = qr/(?:\/\*.*?\*\/)/;                           # Simple 
+match of C-style comment
my $sp      = qr/(?:\s*$comment?\s*)*/;                      # This ju
+st looks for /* comments */, with whitespace on either side. 
                                                                # Sinc
+e comments can be anywhere, use this anywhere whitespace is allowed. 
                                            # The parens don't capture
+, so it won't affect your $1, $2 counts.
my $header  = qr/create${sp}table${sp}\#?\w+${sp}\(${sp}/i;  # "create
+ table foo (", where table name may have a # mark
my $name    = qr/[\#\w]+${sp}/;                              # Variabl
+e name, can contain # mark
my $type    = qr/\w+${sp}/;                                  # Variabl
+e type (e.g. 'char', 'number') 
my $op_char = qr/(?:\([\d\s,]+\)${sp})?/;                    # Bounds 
+for array types, can be multidimensional (e.g. '(10, 20)', '(30)')
my $words   = qr/(?:[\w]+(?:\([^\)]*\))?[${sp})]*)*/;        # Text af
+ter variable type -- may be 'null', 'not null', a 'check' block, etc.
my $subst   = qr/${sp}\{[^\}]+\}${sp}/;                      # Substit
+ution placeholder -- some text in {curlies}
my $decl    = qr/($name $type $op_char $words)/x;            # A compl
+ete variable declaration
my $comma   = qr/,${sp}/;                                    # Literal
+ comma, with possible trailing space
my $rest    = qr/\)${sp}/;                                   # Closing
+ the block with a close-paren

if ($sql =~ /
        ($header)
        ((?: $decl $comma | $subst $comma | $subst)*
        (?: $decl | $subst))
        ($rest)
/x) {
    print "Match!\n";
    print "1: >>>$1<<<\n";
    print "2: >>>$2<<<\n";
}

It takes about 45 seconds on my box. The longer the comment, the longer the runtime.