use warnings;
use base qw/Test::Builder::Module Exporter/;
use Data::Dumper;
-use Carp;
use Test::Builder;
-use Test::Deep qw(eq_deeply);
+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 $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
- )',
+# All of these keywords allow their parameters to be specified with or without parenthesis without changing the semantics
+my @unrollable_ops = (
'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) = @_;
my $same_bind = eq_bind($bind_ref1, $bind_ref2);
# call Test::Builder::ok
- $tb->ok($same_sql && $same_bind, $msg);
+ my $ret = $tb->ok($same_sql && $same_bind, $msg);
# add debugging info
if (!$same_sql) {
if (!$same_bind) {
_bind_differ_diag($bind_ref1, $bind_ref2);
}
+
+ # pass ok() result further
+ return $ret;
}
sub is_same_sql {
my $same_sql = eq_sql($sql1, $sql2);
# call Test::Builder::ok
- $tb->ok($same_sql, $msg);
+ my $ret = $tb->ok($same_sql, $msg);
# add debugging info
if (!$same_sql) {
_sql_differ_diag($sql1, $sql2);
}
+
+ # pass ok() result further
+ return $ret;
}
sub is_same_bind {
my $same_bind = eq_bind($bind_ref1, $bind_ref2);
# call Test::Builder::ok
- $tb->ok($same_bind, $msg);
+ my $ret = $tb->ok($same_bind, $msg);
# add debugging info
if (!$same_bind) {
_bind_differ_diag($bind_ref1, $bind_ref2);
}
+
+ # pass ok() result further
+ return $ret;
}
sub _sql_differ_diag {
sub eq_bind {
my ($bind_ref1, $bind_ref2) = @_;
- return eq_deeply($bind_ref1, $bind_ref2);
+ local $Data::Dumper::Useqq = 1;
+ local $Data::Dumper::Sortkeys = 1;
+
+ return Dumper($bind_ref1) eq Dumper($bind_ref2);
}
sub eq_sql {
my ($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, $tree2);
+ return 1 if _eq_sql($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) ) {
return 0;
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($right);
+ # different amount of elements
+ elsif (@$left != @$right) {
+ $sql_differ = sprintf ("left: %s\nright: %s\n", map { $sqlat->unparse ($_) } ($left, $right) );
return 0;
}
- # elsif operators are identical, compare operands
- else {
- if ($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], $right->[1][1]); # right operand
- }
+ # one is empty - so is the other
+ elsif (@$left == 0) {
+ return 1;
}
-}
-
-
-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 $token !~ /^$/;
+ # one is a list, the other is an op with a list
+ elsif (ref $left->[0] xor ref $right->[0]) {
+ $sql_differ = sprintf ("left: %s\nright: %s\n", map { $sqlat->unparse ($_) } ($left, $right) );
+ return 0;
}
-
- my $tree = _recurse_parse($tokens, PARSE_TOP_LEVEL);
- return $tree;
-}
-
-sub _recurse_parse {
- my ($tokens, $state) = @_;
-
- my $left;
- while (1) { # left-associative parsing
-
- 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
- );
-
- my $token = shift @$tokens;
-
- # 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];
+ # one is a list, so is the other
+ elsif (ref $left->[0]) {
+ for (my $i = 0; $i <= $#$left or $i <= $#$right; $i++ ) {
+ return 0 if (not _eq_sql ($left->[$i], $right->[$i]) );
}
- # AND/OR
- elsif ($token eq 'AND' || $token eq 'OR') {
- 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 => $right]]];
+ return 1;
+ }
+ # both are an op-list combo
+ else {
+
+ # unroll parenthesis if possible/allowed
+ _parenthesis_unroll ($_) for ($left, $right);
+
+ # if operators are different
+ 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;
}
- # leaf expression
+ # elsif operators are identical, compare operands
else {
- $left = $left ? [CONCAT => [$left, [EXPR => $token]]]
- : [EXPR => $token];
+ if ($left->[0] eq 'LITERAL' ) { # unary
+ (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;
+ }
+ else {
+ 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 _parenthesis_unroll {
+ my $ast = shift;
+
+ return if $parenthesis_significant;
+ return unless (ref $ast and ref $ast->[1]);
+
+ my $changes;
+ do {
+ my @children;
+ $changes = 0;
+
+ for my $child (@{$ast->[1]}) {
+ # the current node in this loop is *always* a PAREN
+ if (not ref $child or not $child->[0] eq 'PAREN') {
+ push @children, $child;
+ next;
+ }
+
+ # unroll nested parenthesis
+ while ( @{$child->[1]} && $child->[1][0][0] eq 'PAREN') {
+ $child = $child->[1][0];
+ $changes++;
+ }
+
+ # if the parenthesis are wrapped around an AND/OR matching the parent AND/OR - open the parenthesis up and merge the list
+ if (
+ ( $ast->[0] eq 'AND' or $ast->[0] eq 'OR')
+ and
+ $child->[1][0][0] eq $ast->[0]
+ ) {
+ push @children, @{$child->[1][0][1]};
+ $changes++;
+ }
+
+ # if the parent operator explcitly allows it nuke the parenthesis
+ elsif ( grep { $ast->[0] =~ /^ $_ $/xi } @unrollable_ops ) {
+ push @children, $child->[1][0];
+ $changes++;
+ }
+
+ # only *ONE* LITERAL element
+ elsif (
+ @{$child->[1]} == 1 && $child->[1][0][0] eq 'LITERAL'
+ ) {
+ push @children, $child->[1][0];
+ $changes++;
+ }
+
+ # only one element in the parenthesis which is a binary op
+ # and has exactly two grandchildren
+ # the only time when we can *not* unroll this is when both
+ # the parent and the child are mathops (in which case we'll
+ # break precedence) or when the child is BETWEEN (special
+ # case)
+ elsif (
+ @{$child->[1]} == 1
+ and
+ $child->[1][0][0] =~ SQL::Abstract::Tree::_binary_op_re()
+ and
+ $child->[1][0][0] ne 'BETWEEN'
+ and
+ @{$child->[1][0][1]} == 2
+ and
+ ! (
+ $child->[1][0][0] =~ SQL::Abstract::Tree::_math_op_re()
+ and
+ $ast->[0] =~ SQL::Abstract::Tree::_math_op_re()
+ )
+ ) {
+ push @children, $child->[1][0];
+ $changes++;
+ }
+
+ # a function binds tighter than a mathop - see if our ancestor is a
+ # mathop, and our content is a single non-mathop child with a single
+ # PAREN grandchild which would indicate mathop ( nonmathop ( ... ) )
+ elsif (
+ @{$child->[1]} == 1
+ and
+ @{$child->[1][0][1]} == 1
+ and
+ $child->[1][0][1][0][0] eq 'PAREN'
+ and
+ $ast->[0] =~ SQL::Abstract::Tree::_math_op_re()
+ and
+ $child->[1][0][0] !~ SQL::Abstract::Tree::_math_op_re
+ ) {
+ push @children, $child->[1][0];
+ $changes++;
+ }
+
+
+ # otherwise no more mucking for this pass
+ else {
+ push @children, $child;
+ }
+ }
+ $ast->[1] = \@children;
-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]}->();
-}
+ } while ($changes);
+}
+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 for testing the C<-nest> modifier. Defaults to 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.