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
+our @EXPORT_OK = qw/&is_same_sql_bind &is_same_sql &is_same_bind
+ &eq_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 $tb = __PACKAGE__->builder;
+# Parser states for _recurse_parse()
+use constant PARSE_TOP_LEVEL => 0;
+use constant PARSE_IN_EXPR => 1;
+use constant PARSE_IN_PARENS => 2;
+use constant PARSE_RHS => 3;
+
+# These SQL keywords always signal end of the current expression (except inside
+# of a parenthesized subexpression).
+# Format: A list of strings that will be compiled to extended syntax (ie.
+# /.../x) regexes, without capturing parentheses. They will be automatically
+# anchored to word boundaries to match the whole token).
+my @expression_terminator_sql_keywords = (
+ 'SELECT',
+ 'FROM',
+ '(?:
+ (?:
+ (?: \b (?: LEFT | RIGHT | FULL ) \s+ )?
+ (?: \b (?: CROSS | INNER | OUTER ) \s+ )?
+ )?
+ JOIN
+ )',
+ 'ON',
+ 'WHERE',
+ 'GROUP \s+ BY',
+ 'HAVING',
+ 'ORDER \s+ BY',
+ 'LIMIT',
+ 'OFFSET',
+ 'FOR',
+ 'UNION',
+ 'INTERSECT',
+ 'EXCEPT',
+);
+
+# These are binary operator keywords always a single LHS and RHS
+# * AND/OR are handled separately as they are N-ary
+# * BETWEEN without paranthesis around the ANDed arguments (which
+# makes it a non-binary op) is detected and accomodated in
+# _recurse_parse()
+my @binary_op_keywords = (
+ (map { "\Q$_\E" } (qw/< > != = <= >=/)),
+ '(?: NOT \s+)? LIKE',
+ '(?: NOT \s+)? BETWEEN',
+);
+
+my $tokenizer_re_str = join("\n\t|\n",
+ ( map { '\b' . $_ . '\b' } @expression_terminator_sql_keywords, 'AND', 'OR' ),
+ ( map { q! (?<= [\w\s\`\'\)] ) ! . $_ . q! (?= [\w\s\`\'\(] ) ! } @binary_op_keywords ),
+);
+
+my $tokenizer_re = qr/ \s* ( \( | \) | \? | $tokenizer_re_str ) \s* /xi;
+
+# All of these keywords allow their parameters to be specified with or without parenthesis without changing the semantics
+my @unrollable_ops = (
+ 'ON',
+ 'WHERE',
+ 'GROUP \s+ BY',
+ 'HAVING',
+ 'ORDER \s+ BY',
+);
+
sub is_same_sql_bind {
my ($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::Builder::ok
- $tb->ok($same_sql && $same_bind, $msg);
+ my $ret = $tb->ok($same_sql && $same_bind, $msg);
# add debugging info
if (!$same_sql) {
- $tb->diag("SQL expressions differ\n"
- ." got: $sql1\n"
- ."expected: $sql2\n"
- ."differing in :\n$sql_differ\n"
- );
+ _sql_differ_diag($sql1, $sql2);
}
if (!$same_bind) {
- $tb->diag("BIND values differ\n"
- ." got: " . Dumper($bind_ref1)
- ."expected: " . Dumper($bind_ref2)
- );
+ _bind_differ_diag($bind_ref1, $bind_ref2);
}
+
+ # pass ok() result further
+ return $ret;
}
+sub is_same_sql {
+ my ($sql1, $sql2, $msg) = @_;
-sub eq_bind {
+ # compare
+ my $same_sql = eq_sql($sql1, $sql2);
+
+ # call Test::Builder::ok
+ my $ret = $tb->ok($same_sql, $msg);
+
+ # add debugging info
+ if (!$same_sql) {
+ _sql_differ_diag($sql1, $sql2);
+ }
+
+ # pass ok() result further
+ return $ret;
+}
+
+sub is_same_bind {
+ my ($bind_ref1, $bind_ref2, $msg) = @_;
+
+ # compare
+ my $same_bind = eq_bind($bind_ref1, $bind_ref2);
+
+ # call Test::Builder::ok
+ 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 $ret;
+}
+
+sub _sql_differ_diag {
+ my ($sql1, $sql2) = @_;
+
+ $tb->diag("SQL expressions differ\n"
+ ." got: $sql1\n"
+ ."expected: $sql2\n"
+ ."differing in :\n$sql_differ\n"
+ );
+}
+
+sub _bind_differ_diag {
my ($bind_ref1, $bind_ref2) = @_;
- return stringify_bind($bind_ref1) eq stringify_bind($bind_ref2);
+
+ $tb->diag("BIND values differ\n"
+ ." got: " . Dumper($bind_ref1)
+ ."expected: " . Dumper($bind_ref2)
+ );
}
-sub stringify_bind {
- my $bind_ref = shift || [];
-
- # some bind values can be arrayrefs (see L<SQL::Abstract/bindtype>),
- # so stringify them.
- # furthermore, if L<SQL::Abstract/array_datatypes> is set to true, elements
- # of those arrayrefs can be arrayrefs, too.
- my @strings = map {
- ref $_ eq 'ARRAY'
- ? join('=>', map {
- ref $_ eq 'ARRAY'
- ? ('[' . join('=>', @$_) . ']')
- : (defined $_ ? $_ : '')
- } @$_)
- : (defined $_ ? $_ : '')
- } @$bind_ref;
-
- # join all values into a single string
- return join "///", @strings;
+sub eq_sql_bind {
+ my ($sql1, $bind_ref1, $sql2, $bind_ref2) = @_;
+
+ return eq_sql($sql1, $sql2) && eq_bind($bind_ref1, $bind_ref2);
+}
+
+
+sub eq_bind {
+ my ($bind_ref1, $bind_ref2) = @_;
+
+ return eq_deeply($bind_ref1, $bind_ref2);
}
sub eq_sql {
- my ($left, $right) = @_;
+ my ($sql1, $sql2) = @_;
- # ignore top-level parentheses
- while ($left->[0] eq 'PAREN') {$left = $left->[1] }
- while ($right->[0] eq 'PAREN') {$right = $right->[1]}
+ # parse
+ my $tree1 = parse($sql1);
+ my $tree2 = parse($sql2);
- # if operators are different
- if ($left->[0] ne $right->[0]) {
- $sql_differ = sprintf "OP [$left->[0]] != [$right->[0]] in\nleft: %s\nright: %s\n",
- unparse($left),
- unparse($right);
+ return 1 if _eq_sql($tree1, $tree2);
+}
+
+sub _eq_sql {
+ my ($left, $right) = @_;
+
+ # one is defined the other not
+ if ( (defined $left) xor (defined $right) ) {
return 0;
}
- # elsif operators are identical, compare operands
- else {
- if ($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;
+ # one is undefined, then so is the other
+ elsif (not defined $left) {
+ return 1;
+ }
+ # one is a list, the other is an op with a list
+ elsif (ref $left->[0] xor ref $right->[0]) {
+ $sql_differ = sprintf ("left: %s\nright: %s\n", map { unparse ($_) } ($left, $right) );
+ return 0;
+ }
+ # 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]) );
+ }
+ return 1;
+ }
+ # both are an op-list combo
+ else {
+
+ for my $ast ($left, $right) {
+
+ next unless (ref $ast->[1]);
+
+ # unroll parenthesis in an elaborate loop
+ my $changes;
+ do {
+
+ my @children;
+ $changes = 0;
+
+ for my $child (@{$ast->[1]}) {
+ if (not ref $child or not $child->[0] eq 'PAREN') {
+ push @children, $child;
+ next;
+ }
+
+ # unroll nested parenthesis
+ while ($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];
+ $changes++;
+ }
+
+ # only one element in the parenthesis which is a binary op with two EXPR sub-children
+ elsif (
+ @{$child->[1]} == 1
+ and
+ grep { $child->[1][0][0] =~ /^ $_ $/xi } (@binary_op_keywords)
+ and
+ $child->[1][0][1][0][0] eq 'EXPR'
+ and
+ $child->[1][0][1][1][0] eq 'EXPR'
+ ) {
+ push @children, $child->[1][0];
+ $changes++;
+ }
+
+ # otherwise no more mucking for this pass
+ else {
+ push @children, $child;
+ }
+ }
+
+ $ast->[1] = \@children;
+ } while ($changes);
}
- else { # binary operator
- return eq_sql($left->[1][0], $right->[1][0]) # left operand
- && eq_sql($left->[1][1], $right->[1][1]); # right operand
+
+ # if operators are different
+ if ($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][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]);
+ $sql_differ ||= sprintf ("left: %s\nright: %s\n", map { unparse ($_) } ($left, $right) ) if not $eq;
+ return $eq;
+ }
}
}
}
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 length $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) || $lookahead eq ')';
+ if ( not defined($lookahead)
+ or
+ ($state == PARSE_IN_PARENS && $lookahead eq ')')
+ or
+ ($state == PARSE_IN_EXPR && grep { $lookahead =~ /^ $_ $/xi } ('\)', @expression_terminator_sql_keywords ) )
+ or
+ ($state == PARSE_RHS && grep { $lookahead =~ /^ $_ $/xi } ('\)', @expression_terminator_sql_keywords, @binary_op_keywords, 'AND', 'OR' ) )
+ ) {
+ return $left;
+ }
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]]]
- : [PAREN => $right];
+ $left = $left ? [@$left, [PAREN => [$right] ]]
+ : [PAREN => [$right] ];
}
# AND/OR
- elsif ($token eq 'AND' || $token eq 'OR') {
- my $right = _recurse_parse($tokens);
- $left = [$token => [$left, $right]];
+ elsif ($token =~ /^ (?: OR | AND ) $/xi ) {
+ my $op = uc $token;
+ my $right = _recurse_parse($tokens, PARSE_IN_EXPR);
+
+ # Merge chunks if logic matches
+ if (ref $right and $op eq $right->[0]) {
+ $left = [ (shift @$right ), [$left, map { @$_ } @$right] ];
+ }
+ else {
+ $left = [$op => [$left, $right]];
+ }
+ }
+ # binary operator keywords
+ elsif (grep { $token =~ /^ $_ $/xi } @binary_op_keywords ) {
+ my $op = uc $token;
+ my $right = _recurse_parse($tokens, PARSE_RHS);
+
+ # A between with a simple EXPR for a 1st RHS argument needs a
+ # rerun of the search to (hopefully) find the proper AND construct
+ if ($op eq 'BETWEEN' and $right->[0] eq 'EXPR') {
+ unshift @$tokens, $right->[1][0];
+ $right = _recurse_parse($tokens, PARSE_IN_EXPR);
+ }
+
+ $left = [$op => [$left, $right] ];
+ }
+ # expression terminator keywords (as they start a new expression)
+ elsif (grep { $token =~ /^ $_ $/xi } @expression_terminator_sql_keywords ) {
+ my $op = uc $token;
+ my $right = _recurse_parse($tokens, PARSE_IN_EXPR);
+ $left = $left ? [@$left, [$op => [$right] ]]
+ : [[ $op => [$right] ]];
}
# leaf expression
else {
- $left = $left ? [CONCAT => [$left, [EXPR => $token]]]
- : [EXPR => $token];
+ $left = $left ? [@$left, [EXPR => [$token] ] ]
+ : [ EXPR => [$token] ];
}
}
}
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]}->();
+
+ if (not $tree ) {
+ return '';
+ }
+ elsif (ref $tree->[0]) {
+ return join (" ", map { unparse ($_) } @$tree);
+ }
+ elsif ($tree->[0] eq 'EXPR') {
+ return $tree->[1][0];
+ }
+ elsif ($tree->[0] eq 'PAREN') {
+ return sprintf '(%s)', join (" ", map {unparse($_)} @{$tree->[1]});
+ }
+ elsif ($tree->[0] eq 'OR' or $tree->[0] eq 'AND' or (grep { $tree->[0] =~ /^ $_ $/xi } @binary_op_keywords ) ) {
+ return join (" $tree->[0] ", map {unparse($_)} @{$tree->[1]});
+ }
+ else {
+ return sprintf '%s %s', $tree->[0], unparse ($tree->[1]);
+ }
}
use SQL::Abstract;
use Test::More;
- use SQL::Abstract::Test import => ['is_same_sql_bind'];
+ use SQL::Abstract::Test import => [qw/
+ 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,
$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,
+ $expected_sql, \@expected_bind);
+
+ my $sql_same = eq_sql($given_sql, $expected_sql);
+ my $bind_same = eq_bind(\@given_bind, \@expected_bind);
+
=head1 DESCRIPTION
This module is only intended for authors of tests on
$expected_sql, \@expected_bind, $test_msg);
Compares given and expected pairs of C<($sql, \@bind)>, and calls
-L<Test::Builder/ok> on the result, with C<$test_msg> as message. If the
-test fails, a detailed diagnostic is printed. For clients which use
-L<Test::Build>, this is the only function that needs to be
-imported.
+L<Test::Builder/ok> on the result, with C<$test_msg> as message. If the test
+fails, a detailed diagnostic is printed. For clients which use L<Test::More>,
+this is the one of the three functions (L</is_same_sql_bind>, L</is_same_sql>,
+L</is_same_bind>) that needs to be imported.
+
+=head2 is_same_sql
+
+ is_same_sql($given_sql, $expected_sql, $test_msg);
+
+Compares given and expected SQL statements, and calls L<Test::Builder/ok> on
+the result, with C<$test_msg> as message. If the test fails, a detailed
+diagnostic is printed. For clients which use L<Test::More>, this is the one of
+the three functions (L</is_same_sql_bind>, L</is_same_sql>, L</is_same_bind>)
+that needs to be imported.
+
+=head2 is_same_bind
+
+ is_same_bind(\@given_bind, \@expected_bind, $test_msg);
+
+Compares given and expected bind values, and calls L<Test::Builder/ok> on the
+result, with C<$test_msg> as message. If the test fails, a detailed diagnostic
+is printed. For clients which use L<Test::More>, this is the one of the three
+functions (L</is_same_sql_bind>, L</is_same_sql>, L</is_same_bind>) that needs
+to be imported.
+
+=head2 eq_sql_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
+L</is_same_sql_bind>, but it just returns a boolean value and does not print
+diagnostics or talk to L<Test::Builder>.
=head2 eq_sql
my $is_same = eq_sql($given_sql, $expected_sql);
-Compares the abstract syntax of two SQL statements. If the result is
-false, global variable L</sql_differ> will contain the SQL portion
-where a difference was encountered; this is useful for printing diagnostics.
+Compares the abstract syntax of two SQL statements. Similar to L</is_same_sql>,
+but it just returns a boolean value and does not print diagnostics or talk to
+L<Test::Builder>. If the result is false, the global variable L</$sql_differ>
+will contain the SQL portion where a difference was encountered; this is useful
+for printing diagnostics.
=head2 eq_bind
my $is_same = eq_sql(\@given_bind, \@expected_bind);
-Compares two lists of bind values, taking into account
-the fact that some of the values may be
-arrayrefs (see L<SQL::Abstract/bindtype>).
+Compares two lists of bind values, taking into account the fact that some of
+the values may be arrayrefs (see L<SQL::Abstract/bindtype>). Similar to
+L</is_same_bind>, but it just returns a boolean value and does not print
+diagnostics or talk to L<Test::Builder>.
=head1 GLOBAL VARIABLES
-=head2 case_sensitive
+=head2 $case_sensitive
If true, SQL comparisons will be case-sensitive. Default is false;
-=head2 sql_differ
+=head2 $sql_differ
When L</eq_sql> returns false, the global variable
C<$sql_differ> contains the SQL portion
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.