X-Git-Url: http://git.shadowcat.co.uk/gitweb/gitweb.cgi?a=blobdiff_plain;f=lib%2FSQL%2FAbstract%2FTree.pm;h=0cceed81c3e0ba3e9b082e9344b4c264b425c68e;hb=fb98df489bcd8f1d0edeb5258768f3d2ef6c5654;hp=b1ec93dcda9a81880a9304f56ca8dc88d261313a;hpb=84c65032102185049278b5ebc5651458b92152dd;p=dbsrgits%2FSQL-Abstract.git diff --git a/lib/SQL/Abstract/Tree.pm b/lib/SQL/Abstract/Tree.pm index b1ec93d..0cceed8 100644 --- a/lib/SQL/Abstract/Tree.pm +++ b/lib/SQL/Abstract/Tree.pm @@ -2,28 +2,52 @@ package SQL::Abstract::Tree; use strict; use warnings; +no warnings 'qw'; use Carp; -use List::Util; +use Hash::Merge qw//; use base 'Class::Accessor::Grouped'; __PACKAGE__->mk_group_accessors( simple => $_ ) for qw( newline indent_string indent_amount colormap indentmap fill_in_placeholders + placeholder_surround ); -# 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; +my $merger = Hash::Merge->new; + +$merger->specify_behavior({ + SCALAR => { + SCALAR => sub { $_[1] }, + ARRAY => sub { [ $_[0], @{$_[1]} ] }, + HASH => sub { $_[1] }, + }, + ARRAY => { + SCALAR => sub { $_[1] }, + ARRAY => sub { $_[1] }, + HASH => sub { $_[1] }, + }, + HASH => { + SCALAR => sub { $_[1] }, + ARRAY => sub { [ values %{$_[0]}, @{$_[1]} ] }, + HASH => sub { Hash::Merge::_merge_hashes( $_[0], $_[1] ) }, + }, +}, 'SQLA::Tree Behavior' ); + +my $op_look_ahead = '(?: (?= [\s\)\(\;] ) | \z)'; +my $op_look_behind = '(?: (?<= [\,\s\)\(] ) | \A )'; + +my $quote_left = qr/[\`\'\"\[]/; +my $quote_right = qr/[\`\'\"\]]/; + +my $placeholder_re = qr/(?: \? | \$\d+ )/x; # 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. +# 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 = ( +# anchored to op boundaries (excluding quotes) to match the whole token. +my @expression_start_keywords = ( 'SELECT', 'UPDATE', 'INSERT \s+ INTO', @@ -32,57 +56,93 @@ my @expression_terminator_sql_keywords = ( 'SET', '(?: (?: - (?: \b (?: LEFT | RIGHT | FULL ) \s+ )? - (?: \b (?: CROSS | INNER | OUTER ) \s+ )? + (?: (?: LEFT | RIGHT | FULL ) \s+ )? + (?: (?: CROSS | INNER | OUTER ) \s+ )? )? JOIN )', 'ON', 'WHERE', - 'VALUES', + '(?: DEFAULT \s+ )? VALUES', 'EXISTS', 'GROUP \s+ BY', 'HAVING', 'ORDER \s+ BY', + 'SKIP', + 'FIRST', 'LIMIT', 'OFFSET', 'FOR', 'UNION', 'INTERSECT', 'EXCEPT', + 'BEGIN \s+ WORK', + 'COMMIT', + 'ROLLBACK \s+ TO \s+ SAVEPOINT', + 'ROLLBACK', + 'SAVEPOINT', + 'RELEASE \s+ SAVEPOINT', 'RETURNING', 'ROW_NUMBER \s* \( \s* \) \s+ OVER', ); +my $expr_start_re = join ("\n\t|\n", @expression_start_keywords ); +$expr_start_re = qr/ $op_look_behind (?i: $expr_start_re ) $op_look_ahead /x; + # These are binary operator keywords always a single LHS and RHS # * AND/OR are handled separately as they are N-ary # * so is NOT as being unary # * BETWEEN without paranthesis around the ANDed arguments (which # makes it a non-binary op) is detected and accomodated in # _recurse_parse() -my $stuff_around_mathops = qr/[\w\s\`\'\"\)]/; -my @binary_op_keywords = ( - ( map - { - ' ^ ' . quotemeta ($_) . "(?= \$ | $stuff_around_mathops ) ", - " (?<= $stuff_around_mathops)" . quotemeta ($_) . "(?= \$ | $stuff_around_mathops ) ", - } - (qw/< > != <> = <= >=/) - ), - ( map - { '\b (?: NOT \s+)?' . $_ . '\b' } - (qw/IN BETWEEN LIKE/) - ), + +# this will be included in the $binary_op_re, the distinction is interesting during +# testing as one is tighter than the other, plus mathops have different look +# ahead/behind (e.g. "x"="y" ) +my @math_op_keywords = (qw/ < > != <> = <= >= /); +my $math_re = join ("\n\t|\n", map + { "(?: (?<= [\\w\\s] | $quote_right ) | \\A )" . quotemeta ($_) . "(?: (?= [\\w\\s] | $quote_left ) | \\z )" } + @math_op_keywords ); +$math_re = qr/$math_re/x; -my $tokenizer_re_str = join("\n\t|\n", - ( map { '\b' . $_ . '\b' } @expression_terminator_sql_keywords, 'AND', 'OR', 'NOT'), - @binary_op_keywords, +sub _math_op_re { $math_re } + + +my $binary_op_re = '(?: NOT \s+)? (?:' . join ('|', qw/IN BETWEEN R?LIKE/) . ')'; +$binary_op_re = join "\n\t|\n", + "$op_look_behind (?i: $binary_op_re ) $op_look_ahead", + $math_re, + $op_look_behind . 'IS (?:\s+ NOT)?' . "(?= \\s+ NULL \\b | $op_look_ahead )", +; +$binary_op_re = qr/$binary_op_re/x; + +sub _binary_op_re { $binary_op_re } + +my $all_known_re = join("\n\t|\n", + $expr_start_re, + $binary_op_re, + "$op_look_behind (?i: AND|OR|NOT ) $op_look_ahead", + (map { quotemeta $_ } qw/, ( ) */), + $placeholder_re, ); -my $tokenizer_re = qr/ \s* ( $tokenizer_re_str | \( | \) | \? ) \s* /xi; +$all_known_re = qr/$all_known_re/x; -sub _binary_op_keywords { @binary_op_keywords } +#this one *is* capturing for the split below +# splits on whitespace if all else fails +my $tokenizer_re = qr/ \s* ( $all_known_re ) \s* | \s+ /x; + +# 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_IN_FUNC => 3; +use constant PARSE_RHS => 4; + +my $expr_term_re = qr/ ^ (?: $expr_start_re | \) ) $/x; +my $rhs_term_re = qr/ ^ (?: $expr_term_re | $binary_op_re | (?i: AND | OR | NOT | \, ) ) $/x; +my $func_start_re = qr/^ (?: \* | $placeholder_re | \( ) $/x; my %indents = ( select => 0, @@ -94,32 +154,88 @@ my %indents = ( join => 1, 'left join' => 1, on => 2, + having => 0, 'group by' => 0, 'order by' => 0, set => 1, into => 1, values => 1, + limit => 1, + offset => 1, + skip => 1, + first => 1, ); my %profiles = ( console => { fill_in_placeholders => 1, + placeholder_surround => ['?/', ''], indent_string => ' ', indent_amount => 2, newline => "\n", colormap => {}, - indentmap => { %indents }, + indentmap => \%indents, + + eval { require Term::ANSIColor } + ? do { + my $c = \&Term::ANSIColor::color; + + my $red = [$c->('red') , $c->('reset')]; + my $cyan = [$c->('cyan') , $c->('reset')]; + my $green = [$c->('green') , $c->('reset')]; + my $yellow = [$c->('yellow') , $c->('reset')]; + my $blue = [$c->('blue') , $c->('reset')]; + my $magenta = [$c->('magenta'), $c->('reset')]; + my $b_o_w = [$c->('black on_white'), $c->('reset')]; + ( + placeholder_surround => [$c->('black on_magenta'), $c->('reset')], + colormap => { + 'begin work' => $b_o_w, + commit => $b_o_w, + rollback => $b_o_w, + savepoint => $b_o_w, + 'rollback to savepoint' => $b_o_w, + 'release savepoint' => $b_o_w, + + select => $red, + 'insert into' => $red, + update => $red, + 'delete from' => $red, + + set => $cyan, + from => $cyan, + + where => $green, + values => $yellow, + + join => $magenta, + 'left join' => $magenta, + on => $blue, + + 'group by' => $yellow, + having => $yellow, + 'order by' => $yellow, + + skip => $green, + first => $green, + limit => $green, + offset => $green, + } + ); + } : (), }, console_monochrome => { fill_in_placeholders => 1, + placeholder_surround => ['?/', ''], indent_string => ' ', indent_amount => 2, newline => "\n", colormap => {}, - indentmap => { %indents }, + indentmap => \%indents, }, html => { fill_in_placeholders => 1, + placeholder_surround => ['', ''], indent_string => ' ', indent_amount => 2, newline => "
\n", @@ -128,17 +244,34 @@ my %profiles = ( 'insert into' => ['' , ''], update => ['' , ''], 'delete from' => ['' , ''], - where => ['' , ''], + + set => ['', ''], from => ['' , ''], + + where => ['' , ''], + values => ['', ''], + join => ['' , ''], + 'left join' => ['',''], on => ['' , ''], + 'group by' => ['', ''], + having => ['', ''], 'order by' => ['', ''], - set => ['', ''], - into => ['', ''], - values => ['', ''], + + skip => ['', ''], + first => ['', ''], + limit => ['', ''], + offset => ['', ''], + + 'begin work' => ['', ''], + commit => ['', ''], + rollback => ['', ''], + savepoint => ['', ''], + 'rollback to savepoint' => ['', ''], + 'release savepoint' => ['', ''], }, - indentmap => { %indents }, + indentmap => \%indents, }, none => { colormap => {}, @@ -146,34 +279,15 @@ my %profiles = ( }, ); -eval { - require Term::ANSIColor; - $profiles{console}->{colormap} = { - select => [Term::ANSIColor::color('red'), Term::ANSIColor::color('reset')], - 'insert into' => [Term::ANSIColor::color('red'), Term::ANSIColor::color('reset')], - update => [Term::ANSIColor::color('red'), Term::ANSIColor::color('reset')], - 'delete from' => [Term::ANSIColor::color('red'), Term::ANSIColor::color('reset')], - - set => [Term::ANSIColor::color('cyan'), Term::ANSIColor::color('reset')], - from => [Term::ANSIColor::color('cyan'), Term::ANSIColor::color('reset')], - - where => [Term::ANSIColor::color('green'), Term::ANSIColor::color('reset')], - values => [Term::ANSIColor::color('yellow'), Term::ANSIColor::color('reset')], - - join => [Term::ANSIColor::color('magenta'), Term::ANSIColor::color('reset')], - 'left join' => [Term::ANSIColor::color('magenta'), Term::ANSIColor::color('reset')], - on => [Term::ANSIColor::color('blue'), Term::ANSIColor::color('reset')], - - 'group by' => [Term::ANSIColor::color('yellow'), Term::ANSIColor::color('reset')], - 'order by' => [Term::ANSIColor::color('yellow'), Term::ANSIColor::color('reset')], - }; -}; - sub new { - my ($class, $args) = @_; + my $class = shift; + my $args = shift || {}; my $profile = delete $args->{profile} || 'none'; - my $data = {%{$profiles{$profile}}, %{$args||{}}}; + + die "No such profile '$profile'!" unless exists $profiles{$profile}; + + my $data = $merger->merge( $profiles{$profile}, $args ); bless $data, $class } @@ -184,11 +298,15 @@ sub parse { # tokenize string, and remove all optional whitespace my $tokens = []; foreach my $token (split $tokenizer_re, $s) { - push @$tokens, $token if (length $token) && ($token =~ /\S/); + push @$tokens, $token if ( + defined $token + and + length $token + and + $token =~ /\S/ + ); } - - my $tree = $self->_recurse_parse($tokens, PARSE_TOP_LEVEL); - return $tree; + $self->_recurse_parse($tokens, PARSE_TOP_LEVEL); } sub _recurse_parse { @@ -202,11 +320,13 @@ sub _recurse_parse { or ($state == PARSE_IN_PARENS && $lookahead eq ')') or - ($state == PARSE_IN_EXPR && grep { $lookahead =~ /^ $_ $/xi } ('\)', @expression_terminator_sql_keywords ) ) + ($state == PARSE_IN_EXPR && $lookahead =~ $expr_term_re ) + or + ($state == PARSE_RHS && $lookahead =~ $rhs_term_re ) or - ($state == PARSE_RHS && grep { $lookahead =~ /^ $_ $/xi } ('\)', @expression_terminator_sql_keywords, @binary_op_keywords, 'AND', 'OR', 'NOT' ) ) + ($state == PARSE_IN_FUNC && $lookahead !~ $func_start_re) # if there are multiple values - the parenthesis will switch the $state ) { - return $left; + return $left||(); } my $token = shift @$tokens; @@ -217,24 +337,25 @@ sub _recurse_parse { $token = shift @$tokens or croak "missing closing ')' around block " . $self->unparse($right); $token eq ')' or croak "unexpected token '$token' terminating block " . $self->unparse($right); - $left = $left ? [@$left, [PAREN => [$right] ]] - : [PAREN => [$right] ]; + $left = $left ? [$left, [PAREN => [$right||()] ]] + : [PAREN => [$right||()] ]; } - # AND/OR - elsif ($token =~ /^ (?: OR | AND ) $/xi ) { - my $op = uc $token; + # AND/OR and LIST (,) + elsif ($token =~ /^ (?: OR | AND | \, ) $/xi ) { + my $op = ($token eq ',') ? 'LIST' : uc $token; + my $right = $self->_recurse_parse($tokens, PARSE_IN_EXPR); # Merge chunks if logic matches if (ref $right and $op eq $right->[0]) { - $left = [ (shift @$right ), [$left, map { @$_ } @$right] ]; + $left = [ (shift @$right ), [$left||(), map { @$_ } @$right] ]; } else { - $left = [$op => [$left, $right]]; + $left = [$op => [ $left||(), $right||() ]]; } } # binary operator keywords - elsif (grep { $token =~ /^ $_ $/xi } @binary_op_keywords ) { + elsif ( $token =~ /^ $binary_op_re $ /x ) { my $op = uc $token; my $right = $self->_recurse_parse($tokens, PARSE_RHS); @@ -248,25 +369,39 @@ sub _recurse_parse { $left = [$op => [$left, $right] ]; } # expression terminator keywords (as they start a new expression) - elsif (grep { $token =~ /^ $_ $/xi } @expression_terminator_sql_keywords ) { + elsif ( $token =~ / ^ $expr_start_re $ /x ) { my $op = uc $token; my $right = $self->_recurse_parse($tokens, PARSE_IN_EXPR); - $left = $left ? [ $left, [$op => [$right] ]] - : [ $op => [$right] ]; + $left = $left ? [ $left, [$op => [$right||()] ]] + : [ $op => [$right||()] ]; } - # NOT (last as to allow all other NOT X pieces first) - elsif ( $token =~ /^ not $/ix ) { + # NOT + elsif ( $token =~ /^ NOT $/ix ) { my $op = uc $token; my $right = $self->_recurse_parse ($tokens, PARSE_RHS); $left = $left ? [ @$left, [$op => [$right] ]] : [ $op => [$right] ]; } - # literal (eat everything on the right until RHS termination) + elsif ( $token =~ $placeholder_re) { + $left = $left ? [ $left, [ PLACEHOLDER => [ $token ] ] ] + : [ PLACEHOLDER => [ $token ] ]; + } + # we're now in "unknown token" land - start eating tokens until + # we see something familiar else { - my $right = $self->_recurse_parse ($tokens, PARSE_RHS); - $left = $left ? [ $left, [LITERAL => [join ' ', $token, $self->unparse($right)||()] ] ] - : [ LITERAL => [join ' ', $token, $self->unparse($right)||()] ]; + my $right; + + # check if the current token is an unknown op-start + if (@$tokens and $tokens->[0] =~ $func_start_re) { + $right = [ $token => [ $self->_recurse_parse($tokens, PARSE_IN_FUNC) || () ] ]; + } + else { + $right = [ LITERAL => [ $token ] ]; + } + + $left = $left ? [ $left, $right ] + : $right; } } } @@ -281,7 +416,6 @@ sub format_keyword { return $keyword } - my %starters = ( select => 1, update => 1, @@ -289,7 +423,7 @@ my %starters = ( 'delete from' => 1, ); -sub whitespace { +sub pad_keyword { my ($self, $keyword, $depth) = @_; my $before = ''; @@ -297,7 +431,7 @@ sub whitespace { $before = $self->newline . $self->indent($depth + $self->indentmap->{lc $keyword}); } $before = '' if $depth == 0 and defined $starters{lc $keyword}; - return [$before, ' ']; + return [$before, '']; } sub indent { ($_[0]->indent_string||'') x ( ( $_[0]->indent_amount || 0 ) * $_[1] ) } @@ -309,51 +443,79 @@ sub _is_key { defined $tree && defined $self->indentmap->{lc $tree}; } -sub _fill_in_placeholder { +sub fill_in_placeholder { my ($self, $bindargs) = @_; if ($self->fill_in_placeholders) { - my $val = pop @{$bindargs}; + my $val = shift @{$bindargs} || ''; + $val =~ s/^(')(.*)(')$/$2/; + my ($lquo, $rquo) = ($1 || '', $3 || ''); + my ($left, $right) = @{$self->placeholder_surround}; $val =~ s/\\/\\\\/g; $val =~ s/'/\\'/g; - return qq('$val') + return qq($left$lquo$val$rquo$right) } return '?' } +# FIXME - terrible name for a user facing API sub unparse { - my ($self, $tree, $bindargs, $depth) = @_; + my ($self, $tree, $bindargs) = @_; + $self->_unparse($tree, [@{$bindargs||[]}], 0); +} - $depth ||= 0; +sub _unparse { + my ($self, $tree, $bindargs, $depth) = @_; - if (not $tree ) { + if (not $tree or not @$tree) { return ''; } - my $car = $tree->[0]; - my $cdr = $tree->[1]; + my ($car, $cdr) = @{$tree}[0,1]; + + if (! defined $car or (! ref $car and ! defined $cdr) ) { + require Data::Dumper; + Carp::confess( sprintf ( "Internal error - malformed branch at depth $depth:\n%s", + Data::Dumper::Dumper($tree) + ) ); + } if (ref $car) { - return join ('', map $self->unparse($_, $bindargs, $depth), @$tree); + return join (' ', map $self->_unparse($_, $bindargs, $depth), @$tree); } elsif ($car eq 'LITERAL') { - if ($cdr->[0] eq '?') { - return $self->_fill_in_placeholder($bindargs) - } return $cdr->[0]; } + elsif ($car eq 'PLACEHOLDER') { + return $self->fill_in_placeholder($bindargs); + } elsif ($car eq 'PAREN') { - return '(' . - join(' ', - map $self->unparse($_, $bindargs, $depth + 2), @{$cdr}) . - ($self->_is_key($cdr)?( $self->newline||'' ).$self->indent($depth + 1):'') . ') '; + return sprintf ('( %s )', + join (' ', map { $self->_unparse($_, $bindargs, $depth + 2) } @{$cdr} ) + . + ($self->_is_key($cdr) + ? ( $self->newline||'' ) . $self->indent($depth + 1) + : '' + ) + ); + } + elsif ($car eq 'AND' or $car eq 'OR' or $car =~ / ^ $binary_op_re $ /x ) { + return join (" $car ", map $self->_unparse($_, $bindargs, $depth), @{$cdr}); } - elsif ($car eq 'OR' or $car eq 'AND' or (grep { $car =~ /^ $_ $/xi } @binary_op_keywords ) ) { - return join (" $car ", map $self->unparse($_, $bindargs, $depth), @{$cdr}); + elsif ($car eq 'LIST' ) { + return join (', ', map $self->_unparse($_, $bindargs, $depth), @{$cdr}); } else { - my ($l, $r) = @{$self->whitespace($car, $depth)}; - return sprintf "$l%s %s$r", $self->format_keyword($car), $self->unparse($cdr, $bindargs, $depth); + my ($l, $r) = @{$self->pad_keyword($car, $depth)}; + + return sprintf "$l%s%s%s$r", + $self->format_keyword($car), + ( ref $cdr eq 'ARRAY' and ref $cdr->[0] eq 'ARRAY' and $cdr->[0][0] and $cdr->[0][0] eq 'PAREN' ) + ? '' # mysql-- + : ' ' + , + $self->_unparse($cdr, $bindargs, $depth), + ; } } @@ -363,6 +525,10 @@ sub format { my $self = shift; $self->unparse($self->parse($_[0]), $_[1]) } =pod +=head1 NAME + +SQL::Abstract::Tree - Represent SQL as an AST + =head1 SYNOPSIS my $sqla_tree = SQL::Abstract::Tree->new({ profile => 'console' }); @@ -379,8 +545,110 @@ sub format { my $self = shift; $self->unparse($self->parse($_[0]), $_[1]) } my $sqla_tree = SQL::Abstract::Tree->new({ profile => 'console' }); + $args = { + profile => 'console', # predefined profile to use (default: 'none') + fill_in_placeholders => 1, # true for placeholder population + placeholder_surround => # The strings that will be wrapped around + [GREEN, RESET], # populated placeholders if the above is set + indent_string => ' ', # the string used when indenting + indent_amount => 2, # how many of above string to use for a single + # indent level + newline => "\n", # string for newline + colormap => { + select => [RED, RESET], # a pair of strings defining what to surround + # the keyword with for colorization + # ... + }, + indentmap => { + select => 0, # A zero means that the keyword will start on + # a new line + from => 1, # Any other positive integer means that after + on => 2, # said newline it will get that many indents + # ... + }, + } + +Returns a new SQL::Abstract::Tree object. All arguments are optional. + +=head3 profiles + +There are four predefined profiles, C, C, C, +and C. Typically a user will probably just use C or +C, but if something about a profile bothers you, merely +use the profile and override the parts that you don't like. + =head2 format - $sqlat->format('SELECT * FROM bar') + $sqlat->format('SELECT * FROM bar WHERE x = ?', [1]) + +Takes C<$sql> and C<\@bindargs>. + +Returns a formatting string based on the string passed in + +=head2 parse + + $sqlat->parse('SELECT * FROM bar WHERE x = ?') + +Returns a "tree" representing passed in SQL. Please do not depend on the +structure of the returned tree. It may be stable at some point, but not yet. + +=head2 unparse + + $sqlat->parse($tree_structure, \@bindargs) + +Transform "tree" into SQL, applying various transforms on the way. + +=head2 format_keyword + + $sqlat->format_keyword('SELECT') + +Currently this just takes a keyword and puts the C stuff around it. +Later on it may do more and allow for coderef based transforms. + +=head2 pad_keyword + + my ($before, $after) = @{$sqlat->pad_keyword('SELECT')}; + +Returns whitespace to be inserted around a keyword. + +=head2 fill_in_placeholder + + my $value = $sqlat->fill_in_placeholder(\@bindargs) + +Removes last arg from passed arrayref and returns it, surrounded with +the values in placeholder_surround, and then surrounded with single quotes. + +=head2 indent + +Returns as many indent strings as indent amounts times the first argument. + +=head1 ACCESSORS + +=head2 colormap + +See L + +=head2 fill_in_placeholders + +See L + +=head2 indent_amount + +See L + +=head2 indent_string + +See L + +=head2 indentmap + +See L + +=head2 newline + +See L + +=head2 placeholder_surround + +See L -Returns a formatting string based on wthe string passed in