use warnings;
use base qw/Test::Builder::Module Exporter/;
use Data::Dumper;
-use Carp;
use Test::Builder;
-use Test::Deep qw(eq_deeply);
+use Test::Deep ();
+use SQL::Abstract::Tree;
our @EXPORT_OK = qw/&is_same_sql_bind &is_same_sql &is_same_bind
- &eq_sql_bind &eq_sql &eq_bind
+ &eq_sql_bind &eq_sql &eq_bind
$case_sensitive $sql_differ/;
+my $sqlat = SQL::Abstract::Tree->new;
+
our $case_sensitive = 0;
+our $parenthesis_significant = 0;
+our $order_by_asc_significant = 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;
-use constant PARSE_IN_EXPR => 1;
-use constant 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, $msg) = @_;
}
-sub eq_bind {
- my ($bind_ref1, $bind_ref2) = @_;
-
- return eq_deeply($bind_ref1, $bind_ref2);
-}
+sub eq_bind { goto &Test::Deep::eq_deeply };
sub eq_sql {
my ($sql1, $sql2) = @_;
# parse
- my $tree1 = parse($sql1);
- my $tree2 = parse($sql2);
+ my $tree1 = $sqlat->parse($sql1);
+ my $tree2 = $sqlat->parse($sql2);
- return _eq_sql($tree1, $tree2);
+ undef $sql_differ;
+ return 1 if _eq_sql($tree1, $tree2);
}
sub _eq_sql {
my ($left, $right) = @_;
- # ignore top-level parentheses
- while ($left and $left->[0] and $left->[0] eq 'PAREN') {$left = $left->[1]}
- while ($right and $right->[0] and $right->[0] eq 'PAREN') {$right = $right->[1]}
-
# one is defined the other not
if ( (defined $left) xor (defined $right) ) {
+ $sql_differ = sprintf ("[%s] != [%s]\n", map { defined $_ ? $sqlat->unparse ($_) : 'N/A' } ($left, $right) );
return 0;
}
+
# one is undefined, then so is the other
elsif (not defined $left) {
return 1;
}
- # if operators are different
- elsif ($left->[0] ne $right->[0]) {
- $sql_differ = sprintf "OP [$left->[0]] != [$right->[0]] in\nleft: %s\nright: %s\n",
- unparse($left),
- unparse($right);
- return 0;
- }
- # elsif operators are identical, compare operands
- else {
- if ($left->[0] eq 'EXPR' ) { # unary operator
- (my $l = " $left->[1] " ) =~ s/\s+/ /g;
- (my $r = " $right->[1] ") =~ s/\s+/ /g;
- my $eq = $case_sensitive ? $l eq $r : uc($l) eq uc($r);
- $sql_differ = "[$left->[1]] != [$right->[1]]\n" if not $eq;
- 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
- }
- }
-}
-
-sub parse {
- my $s = shift;
-
- # 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 $token !~ /^$/;
+ # both are empty
+ elsif (@$left == 0 and @$right == 0) {
+ return 1;
}
- my $tree = _recurse_parse($tokens, PARSE_TOP_LEVEL);
- return $tree;
-}
+ # one is empty
+ if (@$left == 0 or @$right == 0) {
+ $sql_differ = sprintf ("left: %s\nright: %s\n", map { @$_ ? $sqlat->unparse ($_) : 'N/A'} ($left, $right) );
+ return 0;
+ }
-sub _recurse_parse {
- my ($tokens, $state) = @_;
+ # one is a list, the other is an op with a list
+ elsif (ref $left->[0] xor ref $right->[0]) {
+ $sql_differ = sprintf ("[%s] != [%s]\nleft: %s\nright: %s\n", map
+ { ref $_ ? $sqlat->unparse ($_) : $_ }
+ ($left->[0], $right->[0], $left, $right)
+ );
+ return 0;
+ }
- my $left;
- while (1) { # left-associative parsing
+ # both are lists
+ elsif (ref $left->[0]) {
+ for (my $i = 0; $i <= $#$left or $i <= $#$right; $i++ ) {
+ if (not _eq_sql ($left->[$i], $right->[$i]) ) {
+ if (! $sql_differ or $sql_differ !~ /left\:\s .+ right:\s/xs) {
+ $sql_differ ||= '';
+ $sql_differ .= "\n" unless $sql_differ =~ /\n\z/;
+ $sql_differ .= sprintf ("left: %s\nright: %s\n", map { $sqlat->unparse ($_) } ($left, $right) );
+ }
+ return 0;
+ }
+ }
+ return 1;
+ }
- my $lookahead = $tokens->[0];
- return $left if !defined($lookahead)
- || ($state == PARSE_IN_PARENS && $lookahead eq ')')
- || ($state == PARSE_IN_EXPR && grep { $lookahead =~ /^$_$/xi }
- '\)', @expression_terminator_sql_keywords
- );
+ # both are ops
+ else {
- my $token = shift @$tokens;
+ # unroll parenthesis if possible/allowed
+ unless ( $parenthesis_significant ) {
+ $sqlat->_parenthesis_unroll($_) for $left, $right;
+ }
- # nested expression in ()
- if ($token eq '(') {
- 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]]]
- : [PAREN => $right];
+ # unroll ASC order by's
+ unless ($order_by_asc_significant) {
+ $sqlat->_strip_asc_from_order_by($_) for $left, $right;
}
- # AND/OR
- elsif ($token eq 'AND' || $token eq 'OR') {
- my $right = _recurse_parse($tokens, PARSE_IN_EXPR);
- $left = [$token => [$left, $right]];
+
+ if ( $left->[0] ne $right->[0] ) {
+ $sql_differ = sprintf "OP [$left->[0]] != [$right->[0]] in\nleft: %s\nright: %s\n",
+ $sqlat->unparse($left),
+ $sqlat->unparse($right)
+ ;
+ return 0;
}
- # 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 => $right]]];
+
+ # literals have a different arg-sig
+ elsif ($left->[0] eq '-LITERAL') {
+ (my $l = " $left->[1][0] " ) =~ s/\s+/ /g;
+ (my $r = " $right->[1][0] ") =~ s/\s+/ /g;
+ my $eq = $case_sensitive ? $l eq $r : uc($l) eq uc($r);
+ $sql_differ = "[$l] != [$r]\n" if not $eq;
+ return $eq;
}
- # leaf expression
+
+ # if operators are identical, compare operands
else {
- $left = $left ? [CONCAT => [$left, [EXPR => $token]]]
- : [EXPR => $token];
+ my $eq = _eq_sql($left->[1], $right->[1]);
+ $sql_differ ||= sprintf ("left: %s\nright: %s\n", map { $sqlat->unparse ($_) } ($left, $right) ) if not $eq;
+ return $eq;
}
}
}
-
-
-sub unparse {
- my $tree = shift;
- my $dispatch = {
- EXPR => sub {$tree->[1] },
- PAREN => sub {"(" . unparse($tree->[1]) . ")" },
- CONCAT => sub {join " ", map {unparse($_)} @{$tree->[1]}},
- AND => sub {join " AND ", map {unparse($_)} @{$tree->[1]}},
- OR => sub {join " OR ", map {unparse($_)} @{$tree->[1]}},
- };
- $dispatch->{$tree->[0]}->();
-}
-
-
+sub parse { $sqlat->parse(@_) }
1;
is_same_sql_bind is_same_sql is_same_bind
eq_sql_bind eq_sql eq_bind
/];
-
+
my ($sql, @bind) = SQL::Abstract->new->select(%args);
- is_same_sql_bind($given_sql, \@given_bind,
+ is_same_sql_bind($given_sql, \@given_bind,
$expected_sql, \@expected_bind, $test_msg);
is_same_sql($given_sql, $expected_sql, $test_msg);
is_same_bind(\@given_bind, \@expected_bind, $test_msg);
- my $is_same = eq_sql_bind($given_sql, \@given_bind,
+ my $is_same = eq_sql_bind($given_sql, \@given_bind,
$expected_sql, \@expected_bind);
my $sql_same = eq_sql($given_sql, $expected_sql);
Therefore the tests will pass as long as the semantics
is preserved, even if the surface syntax has changed.
-B<Disclaimer> : this is only a half-cooked semantic equivalence;
-parsing is simple-minded, and comparison of SQL abstract syntax trees
-ignores commutativity or associativity of AND/OR operators, Morgan
-laws, etc.
+B<Disclaimer> : the semantic equivalence handling is pretty limited.
+A lot of effort goes into distinguishing significant from
+non-significant parenthesis, including AND/OR operator associativity.
+Currently this module does not support commutativity and more
+intelligent transformations like Morgan laws, etc.
+
+For a good overview of what this test framework is capable of refer
+to C<t/10test.t>
=head1 FUNCTIONS
=head2 is_same_sql_bind
- is_same_sql_bind($given_sql, \@given_bind,
+ 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
=head2 eq_sql_bind
- my $is_same = eq_sql_bind($given_sql, \@given_bind,
+ my $is_same = eq_sql_bind($given_sql, \@given_bind,
$expected_sql, \@expected_bind);
Compares given and expected pairs of C<($sql, \@bind)>. Similar to
If true, SQL comparisons will be case-sensitive. Default is false;
+=head2 $parenthesis_significant
+
+If true, SQL comparison will preserve and report difference in nested
+parenthesis. Useful while testing C<IN (( x ))> vs C<IN ( x )>.
+Defaults to false;
+
+=head2 $order_by_asc_significant
+
+If true SQL comparison will consider C<ORDER BY foo ASC> and
+C<ORDER BY foo> to be different. Default is false;
+
=head2 $sql_differ
When L</eq_sql> returns false, the global variable
Norbert Buchmuller <norbi@nix.hu>
+Peter Rabbitson <ribasushi@cpan.org>
+
=head1 COPYRIGHT AND LICENSE
Copyright 2008 by Laurent Dami.
This library is free software; you can redistribute it and/or modify
-it under the same terms as Perl itself.
+it under the same terms as Perl itself.