X-Git-Url: http://git.shadowcat.co.uk/gitweb/gitweb.cgi?a=blobdiff_plain;f=lib%2FSQL%2FAbstract%2FTest.pm;h=202d1307a39a6fce9d95cd62c16d1e39e5540a87;hb=0c2de280869928d9ff1ee95f36a9a45318766990;hp=82eed8f4affd2d31b11858add8ff3ec2299ebb9f;hpb=09abf3a06717253ac4eb7859bf9145c1f9b1b5af;p=dbsrgits%2FSQL-Abstract.git diff --git a/lib/SQL/Abstract/Test.pm b/lib/SQL/Abstract/Test.pm index 82eed8f..202d130 100644 --- a/lib/SQL/Abstract/Test.pm +++ b/lib/SQL/Abstract/Test.pm @@ -4,79 +4,23 @@ use strict; 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 $tb = __PACKAGE__->builder; -# Parser states for _recurse_parse() -use constant PARSE_TOP_LEVEL => 0; -use constant PARSE_IN_EXPR => 1; -use constant PARSE_IN_PARENS => 2; -use constant PARSE_RHS => 3; - -# These SQL keywords always signal end of the current expression (except inside -# of a parenthesized subexpression). -# Format: A list of strings that will be compiled to extended syntax (ie. -# /.../x) regexes, without capturing parentheses. They will be automatically -# anchored to word boundaries to match the whole token). -my @expression_terminator_sql_keywords = ( - 'SELECT', - 'FROM', - '(?: - (?: - (?: \b (?: LEFT | RIGHT | FULL ) \s+ )? - (?: \b (?: CROSS | INNER | OUTER ) \s+ )? - )? - JOIN - )', - 'ON', - 'WHERE', - 'GROUP \s+ BY', - 'HAVING', - 'ORDER \s+ BY', - 'LIMIT', - 'OFFSET', - 'FOR', - 'UNION', - 'INTERSECT', - 'EXCEPT', -); - -# These are binary operator keywords always a single LHS and RHS -# * AND/OR are handled separately as they are N-ary -# * BETWEEN without paranthesis around the ANDed arguments (which -# makes it a non-binary op) is detected and accomodated in -# _recurse_parse() -my @binary_op_keywords = ( - (map { "\Q$_\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, $msg) = @_; @@ -161,19 +105,16 @@ sub eq_sql_bind { } -sub eq_bind { - my ($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); + undef $sql_differ; return 1 if _eq_sql($tree1, $tree2); } @@ -182,228 +123,90 @@ sub _eq_sql { # 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 $left) { return 1; } + + # both are empty + elsif (@$left == 0 and @$right == 0) { + return 1; + } + + # one is empty + if (@$left == 0 or @$right == 0) { + $sql_differ = sprintf ("left: %s\nright: %s\n", map { @$_ ? $sqlat->unparse ($_) : 'N/A'} ($left, $right) ); + return 0; + } + # 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) ); + $sql_differ = sprintf ("[%s] != [%s]\nleft: %s\nright: %s\n", map + { ref $_ ? $sqlat->unparse ($_) : $_ } + ($left->[0], $right->[0], $left, $right) + ); return 0; } - # one is a list, so is the other + + # both are lists elsif (ref $left->[0]) { for (my $i = 0; $i <= $#$left or $i <= $#$right; $i++ ) { - return 0 if (not _eq_sql ($left->[$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; } - # both are an op-list combo + + # both are ops 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; - } - } + # unroll parenthesis if possible/allowed + unless ( $parenthesis_significant ) { + $sqlat->_parenthesis_unroll($_) for $left, $right; + } - $ast->[1] = \@children; - } while ($changes); + # unroll ASC order by's + unless ($order_by_asc_significant) { + $sqlat->_strip_asc_from_order_by($_) for $left, $right; } - # if operators are different - if ($left->[0] ne $right->[0]) { + if ( $left->[0] ne $right->[0] ) { $sql_differ = sprintf "OP [$left->[0]] != [$right->[0]] in\nleft: %s\nright: %s\n", - unparse($left), - unparse($right); + $sqlat->unparse($left), + $sqlat->unparse($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, 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, PARSE_TOP_LEVEL); - return $tree; -} - -sub _recurse_parse { - my ($tokens, $state) = @_; - - my $left; - while (1) { # left-associative parsing - - my $lookahead = $tokens->[0]; - if ( not defined($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, PARSE_IN_PARENS); - $token = shift @$tokens or croak "missing closing ')' around block " . unparse ($right); - $token eq ')' or croak "unexpected token '$token' terminating block " . unparse ($right); - $left = $left ? [@$left, [PAREN => [$right] ]] - : [PAREN => [$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 $eq; } - # AND/OR - elsif ($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 + # if operators are identical, compare operands else { - $left = $left ? [@$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; - - 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]); - } -} - - +sub parse { $sqlat->parse(@_) } 1; @@ -421,16 +224,16 @@ SQL::Abstract::Test - Helper function for testing SQL::Abstract is_same_sql_bind is_same_sql is_same_bind eq_sql_bind eq_sql eq_bind /]; - + my ($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, $expected_sql); @@ -448,16 +251,20 @@ ignoring differences in spaces or in levels of parentheses. Therefore the tests will pass as long as the semantics is preserved, even if the surface syntax has changed. -B : this is only a half-cooked semantic equivalence; -parsing is simple-minded, and comparison of SQL abstract syntax trees -ignores commutativity or associativity of AND/OR operators, Morgan -laws, etc. +B : the semantic equivalence handling is pretty limited. +A lot of effort goes into distinguishing significant from +non-significant parenthesis, including AND/OR operator associativity. +Currently this module does not support commutativity and more +intelligent transformations like Morgan laws, etc. + +For a good overview of what this test framework is capable of refer +to C =head1 FUNCTIONS =head2 is_same_sql_bind - is_same_sql_bind($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 @@ -488,7 +295,7 @@ to be imported. =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 @@ -520,6 +327,17 @@ diagnostics or talk to L. 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 vs C. +Defaults to false; + +=head2 $order_by_asc_significant + +If true SQL comparison will consider C and +C to be different. Default is false; + =head2 $sql_differ When L returns false, the global variable @@ -537,9 +355,11 @@ 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.