X-Git-Url: http://git.shadowcat.co.uk/gitweb/gitweb.cgi?a=blobdiff_plain;f=lib%2FSQL%2FAbstract%2FTest.pm;h=42491f9fa4181d398a59f21715a64c9dfa17c843;hb=b3b79607321d406a194b2aac205978d925b398c0;hp=d55f3a859f2df7d876706eab07a39aedf585d4f5;hpb=5aad8cf3d200d1538a5fad5b01895e1ccd3da80a;p=dbsrgits%2FSQL-Abstract.git diff --git a/lib/SQL/Abstract/Test.pm b/lib/SQL/Abstract/Test.pm index d55f3a8..42491f9 100644 --- a/lib/SQL/Abstract/Test.pm +++ b/lib/SQL/Abstract/Test.pm @@ -4,149 +4,308 @@ use strict; use warnings; use base qw/Test::Builder::Module Exporter/; use Data::Dumper; -use Carp; +use Test::Builder; +use SQL::Abstract::Tree; -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/; +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; +# 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', +); +my $unrollable_ops_re = join ' | ', @unrollable_ops; +$unrollable_ops_re = qr/$unrollable_ops_re/xi; + sub is_same_sql_bind { my ($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 - $tb->ok($same_sql && $same_bind, $msg); + # call Test::Builder::ok + 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, $msg) = @_; -sub eq_bind { - my ($bind_ref1, $bind_ref2) = @_; - return stringify_bind($bind_ref1) eq stringify_bind($bind_ref2); -} + # compare + my $same_sql = eq_sql($sql1, $sql2); -sub stringify_bind { - my $bind_ref = shift || []; + # call Test::Builder::ok + my $ret = $tb->ok($same_sql, $msg); - # some bind values can be arrayrefs (see L), - # so stringify them. - my @strings = map {ref $_ eq 'ARRAY' ? join('=>', @$_) : ($_ || '')} - @$bind_ref; + # add debugging info + if (!$same_sql) { + _sql_differ_diag($sql1, $sql2); + } - # join all values into a single string - return join "///", @strings; + # pass ok() result further + return $ret; } -sub eq_sql { - my ($left, $right) = @_; +sub is_same_bind { + my ($bind_ref1, $bind_ref2, $msg) = @_; - # ignore top-level parentheses - while ($left->[0] eq 'PAREN') {$left = $left->[1] } - while ($right->[0] eq 'PAREN') {$right = $right->[1]} + # compare + my $same_bind = eq_bind($bind_ref1, $bind_ref2); - # if operators are different - if ($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] " ) =~ 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 - } + # 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) = @_; -sub parse { - my $s = shift; + $tb->diag("BIND values differ\n" + ." got: " . Dumper($bind_ref1) + ."expected: " . Dumper($bind_ref2) + ); +} + +sub eq_sql_bind { + my ($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) = @_; - # tokenize string - my $tokens = [grep {!/^\s*$/} split /\s*(\(|\)|\bAND\b|\bOR\b)\s*/, $s]; + local $Data::Dumper::Useqq = 1; + local $Data::Dumper::Sortkeys = 1; - my $tree = _recurse_parse($tokens); - return $tree; + return Dumper($bind_ref1) eq Dumper($bind_ref2); } -sub _recurse_parse { - my $tokens = shift; +sub eq_sql { + my ($sql1, $sql2) = @_; - my $left; - while (1) { # left-associative parsing + # parse + my $tree1 = $sqlat->parse($sql1); + my $tree2 = $sqlat->parse($sql2); - my $lookahead = $tokens->[0]; - return $left if !defined($lookahead) || $lookahead eq ')'; + return 1 if _eq_sql($tree1, $tree2); +} - my $token = shift @$tokens; +sub _eq_sql { + my ($left, $right) = @_; - # nested expression in () - if ($token eq '(') { - my $right = _recurse_parse($tokens); - $token = shift @$tokens or croak "missing ')'"; - $token eq ')' or croak "unexpected token : $token"; - $left = $left ? [CONCAT => [$left, [PAREN => $right]]] - : [PAREN => $right]; + # one is defined the other not + if ( (defined $left) xor (defined $right) ) { + return 0; + } + # one is undefined, then so is the other + elsif (not defined $left) { + return 1; + } + # different amount of elements + elsif (@$left != @$right) { + $sql_differ = sprintf ("left: %s\nright: %s\n", map { $sqlat->unparse ($_) } ($left, $right) ); + return 0; + } + # one is empty - so is the other + elsif (@$left == 0) { + 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 { $sqlat->unparse ($_) } ($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]) ); } - # AND/OR - elsif ($token eq 'AND' || $token eq 'OR') { - my $right = _recurse_parse($tokens); - $left = [$token => [$left, $right]]; + return 1; + } + # both are an op-list combo + else { + + # unroll parenthesis if possible/allowed + _parenthesis_unroll ($_) 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 ( $ast->[0] =~ $unrollable_ops_re ) { + 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]; + $changes++; + } + + # a function binds tighter than a mathop - see if our ancestor is a + # mathop, and our content is: + # a single non-mathop child with a single PAREN grandchild which + # would indicate mathop ( nonmathop ( ... ) ) + # or a single non-mathop with a single LITERAL ( nonmathop ? ) + elsif ( + @{$child->[1]} == 1 + and + @{$child->[1][0][1]} == 1 + and + $ast->[0] =~ SQL::Abstract::Tree::_math_op_re() + and + $child->[1][0][0] !~ SQL::Abstract::Tree::_math_op_re + and + ( + $child->[1][0][1][0][0] eq 'PAREN' + or + $child->[1][0][1][0][0] eq 'LITERAL' + ) + ) { + push @children, $child->[1][0]; + $changes++; + } + + + # otherwise no more mucking for this pass + else { + push @children, $child; + } + } + + $ast->[1] = \@children; + } while ($changes); -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 { $sqlat->parse(@_) } 1; @@ -160,12 +319,25 @@ SQL::Abstract::Test - Helper function for testing SQL::Abstract use SQL::Abstract; use Test::More; - use SQL::Abstract::Test import => ['is_same_sql_bind']; - + use SQL::Abstract::Test import => [qw/ + is_same_sql_bind is_same_sql is_same_bind + eq_sql_bind eq_sql eq_bind + /]; + my ($sql, @bind) = SQL::Abstract->new->select(%args); - is_same_sql_bind($given_sql, \@given_bind, + + is_same_sql_bind($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 $bind_same = eq_bind(\@given_bind, \@expected_bind); + =head1 DESCRIPTION This module is only intended for authors of tests on @@ -178,47 +350,88 @@ ignoring differences in spaces or in levels of parentheses. Therefore the tests will pass as long as the semantics is preserved, even if the surface syntax has changed. -B : this is only a half-cooked semantic equivalence; -parsing is simple-minded, and comparison of SQL abstract syntax trees -ignores commutativity or associativity of AND/OR operators, Morgan -laws, etc. +B : the semantic equivalence handling is pretty limited. +A lot of effort goes into distinguishing significant from +non-significant parenthesis, including AND/OR operator associativity. +Currently this module does not support commutativity and more +intelligent transformations like Morgan laws, etc. + +For a good overview of what this test framework is capable of refer +to C =head1 FUNCTIONS =head2 is_same_sql_bind - is_same_sql_bind($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 -L on the result, with C<$test_msg> as message. If the -test fails, a detailed diagnostic is printed. For clients which use -L, this is the only function that needs to be -imported. +L on the result, with C<$test_msg> as message. If the test +fails, a detailed diagnostic is printed. For clients which use L, +this is the one of the three functions (L, L, +L) that needs to be imported. + +=head2 is_same_sql + + is_same_sql($given_sql, $expected_sql, $test_msg); + +Compares given and expected SQL statements, and calls L on +the result, with C<$test_msg> as message. If the test fails, a detailed +diagnostic is printed. For clients which use L, this is the one of +the three functions (L, L, L) +that needs to be imported. + +=head2 is_same_bind + + is_same_bind(\@given_bind, \@expected_bind, $test_msg); + +Compares given and expected bind values, and calls L on the +result, with C<$test_msg> as message. If the test fails, a detailed diagnostic +is printed. For clients which use L, this is the one of the three +functions (L, L, L) 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, but it just returns a boolean value and does not print +diagnostics or talk to L. =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 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, +but it just returns a boolean value and does not print diagnostics or talk to +L. If the result is false, the global variable L +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). +Compares two lists of bind values, taking into account the fact that some of +the values may be arrayrefs (see L). Similar to +L, but it just returns a boolean value and does not print +diagnostics or talk to L. =head1 GLOBAL VARIABLES -=head2 case_sensitive +=head2 $case_sensitive If true, SQL comparisons will be case-sensitive. Default is false; -=head2 sql_differ +=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 $sql_differ When L returns false, the global variable C<$sql_differ> contains the SQL portion @@ -227,15 +440,19 @@ where a difference was encountered. =head1 SEE ALSO -L, L. +L, L, L. -=head1 AUTHOR +=head1 AUTHORS Laurent Dami, Elaurent.dami AT etat geneve chE +Norbert Buchmuller + +Peter Rabbitson + =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.