use strict;
use warnings;
-use Test::More;
-use base 'Exporter';
+use base qw/Test::Builder::Module Exporter/;
use Data::Dumper;
use Carp;
+use Test::Builder;
+use Test::Deep qw(eq_deeply);
our @EXPORT_OK = qw/&is_same_sql_bind &eq_sql &eq_bind
$case_sensitive $sql_differ/;
our $case_sensitive = 0;
our $sql_differ; # keeps track of differing portion between SQLs
+our $tb = __PACKAGE__->builder;
+
+# Parser states for _recurse_parse()
+use constant {
+ PARSE_TOP_LEVEL => 0,
+ PARSE_IN_EXPR => 1,
+ PARSE_IN_PARENS => 2,
+};
+
+# These SQL keywords always signal end of the current expression (except inside
+# of a parenthesized subexpression).
+# Format: A list of strings that will be compiled to extended syntax (ie.
+# /.../x) regexes, without capturing parentheses. They will be automatically
+# anchored to word boundaries to match the whole token).
+my @expression_terminator_sql_keywords = (
+ 'FROM',
+ '(?:
+ (?:
+ (?: \b (?: LEFT | RIGHT | FULL ) \s+ )?
+ (?: \b (?: CROSS | INNER | OUTER ) \s+ )?
+ )?
+ JOIN
+ )',
+ 'ON',
+ 'WHERE',
+ 'GROUP \s+ BY',
+ 'HAVING',
+ 'ORDER \s+ BY',
+ 'LIMIT',
+ 'OFFSET',
+ 'FOR',
+ 'UNION',
+ 'INTERSECT',
+ 'EXCEPT',
+);
+
+my $tokenizer_re_str = join('|',
+ map { '\b' . $_ . '\b' }
+ @expression_terminator_sql_keywords, 'AND', 'OR'
+);
+
+my $tokenizer_re = qr/
+ \s*
+ (
+ \(
+ |
+ \)
+ |
+ $tokenizer_re_str
+ )
+ \s*
+/xi;
+
sub is_same_sql_bind {
my ($sql1, $bind_ref1, $sql2, $bind_ref2, $msg) = @_;
# compare
- my $tree1 = parse($sql1);
- my $tree2 = parse($sql2);
- my $same_sql = eq_sql($tree1, $tree2);
+ my $same_sql = eq_sql($sql1, $sql2);
my $same_bind = eq_bind($bind_ref1, $bind_ref2);
- # call Test::More::ok
- ok($same_sql && $same_bind, $msg);
+ # call Test::Builder::ok
+ $tb->ok($same_sql && $same_bind, $msg);
# add debugging info
if (!$same_sql) {
- diag "SQL expressions differ\n"
+ $tb->diag("SQL expressions differ\n"
." got: $sql1\n"
."expected: $sql2\n"
- ."differing in :\n$sql_differ\n";
- ;
+ ."differing in :\n$sql_differ\n"
+ );
}
if (!$same_bind) {
- diag "BIND values differ\n"
+ $tb->diag("BIND values differ\n"
." got: " . Dumper($bind_ref1)
."expected: " . Dumper($bind_ref2)
- ;
+ );
}
}
-
sub eq_bind {
my ($bind_ref1, $bind_ref2) = @_;
- return stringify_bind($bind_ref1) eq stringify_bind($bind_ref2);
+
+ return eq_deeply($bind_ref1, $bind_ref2);
}
-sub stringify_bind {
- my $bind_ref = shift || [];
+sub eq_sql {
+ my ($sql1, $sql2) = @_;
- # some bind values can be arrayrefs (see L<SQL::Abstract/bindtype>),
- # so stringify them.
- my @strings = map {ref $_ eq 'ARRAY' ? join('=>', @$_) : ($_ || '')}
- @$bind_ref;
+ # parse
+ my $tree1 = parse($sql1);
+ my $tree2 = parse($sql2);
- # join all values into a single string
- return join "///", @strings;
+ return _eq_sql($tree1, $tree2);
}
-sub eq_sql {
+sub _eq_sql {
my ($left, $right) = @_;
# ignore top-level parentheses
return $eq;
}
else { # binary operator
- return eq_sql($left->[1][0], $right->[1][0]) # left operand
- && eq_sql($left->[1][1], $right->[1][1]); # right operand
+ return _eq_sql($left->[1][0], $right->[1][0]) # left operand
+ && _eq_sql($left->[1][1], $right->[1][1]); # right operand
}
}
}
sub parse {
my $s = shift;
- # tokenize string
- my $tokens = [grep {!/^\s*$/} split /\s*(\(|\)|\bAND\b|\bOR\b)\s*/, $s];
+ # tokenize string, and remove all optional whitespace
+ my $tokens = [];
+ foreach my $token (split $tokenizer_re, $s) {
+ $token =~ s/\s+/ /g;
+ $token =~ s/\s+([^\w\s])/$1/g;
+ $token =~ s/([^\w\s])\s+/$1/g;
+ push @$tokens, $token if $token !~ /^$/;
+ }
- my $tree = _recurse_parse($tokens);
+ my $tree = _recurse_parse($tokens, PARSE_TOP_LEVEL);
return $tree;
}
sub _recurse_parse {
- my $tokens = shift;
+ my ($tokens, $state) = @_;
my $left;
while (1) { # left-associative parsing
my $lookahead = $tokens->[0];
- return $left if !defined($lookahead) || $lookahead eq ')';
+ return $left if !defined($lookahead)
+ || ($state == PARSE_IN_PARENS && $lookahead eq ')')
+ || ($state == PARSE_IN_EXPR && grep { $lookahead =~ /^$_$/xi }
+ '\)', @expression_terminator_sql_keywords
+ );
my $token = shift @$tokens;
# nested expression in ()
if ($token eq '(') {
- my $right = _recurse_parse($tokens);
+ my $right = _recurse_parse($tokens, PARSE_IN_PARENS);
$token = shift @$tokens or croak "missing ')'";
$token eq ')' or croak "unexpected token : $token";
$left = $left ? [CONCAT => [$left, [PAREN => $right]]]
}
# AND/OR
elsif ($token eq 'AND' || $token eq 'OR') {
- my $right = _recurse_parse($tokens);
+ my $right = _recurse_parse($tokens, PARSE_IN_EXPR);
$left = [$token => [$left, $right]];
}
+ # expression terminator keywords (as they start a new expression)
+ elsif (grep { $token =~ /^$_$/xi } @expression_terminator_sql_keywords) {
+ my $right = _recurse_parse($tokens, PARSE_IN_EXPR);
+ $left = $left ? [CONCAT => [$left, [CONCAT => [[EXPR => $token], [PAREN => $right]]]]]
+ : [CONCAT => [[EXPR => $token], [PAREN => $right]]];
+ }
# leaf expression
else {
$left = $left ? [CONCAT => [$left, [EXPR => $token]]]
use SQL::Abstract;
use Test::More;
- use SQL::Abstract::Test qw/is_same_sql_bind/;
+ use SQL::Abstract::Test import => ['is_same_sql_bind'];
my ($sql, @bind) = SQL::Abstract->new->select(%args);
is_same_sql_bind($given_sql, \@given_bind,
$expected_sql, \@expected_bind, $test_msg);
Compares given and expected pairs of C<($sql, \@bind)>, and calls
-L<Test::More/ok> on the result, with C<$test_msg> as message. If the
+L<Test::Builder/ok> on the result, with C<$test_msg> as message. If the
test fails, a detailed diagnostic is printed. For clients which use
-L<Test::More|Test::More>, this is the only function that needs to be
+L<Test::Build>, this is the only function that needs to be
imported.
=head2 eq_sql
=head1 SEE ALSO
-L<SQL::Abstract>, L<Test::More>.
+L<SQL::Abstract>, L<Test::More>, L<Test::Builder>.
-=head1 AUTHOR
+=head1 AUTHORS
Laurent Dami, E<lt>laurent.dami AT etat geneve chE<gt>
+Norbert Buchmuller <norbi@nix.hu>
+
=head1 COPYRIGHT AND LICENSE
Copyright 2008 by Laurent Dami.