X-Git-Url: http://git.shadowcat.co.uk/gitweb/gitweb.cgi?a=blobdiff_plain;f=lib%2FSQL%2FAbstract.pm;h=bf0022875f502fe139acae9016c6e0619d17f311;hb=4ee5e99cdf609a1ab37e8a9de8fcf222f7fd5010;hp=58f99ccac71ccc30e44c662fc05288a1e869ac4e;hpb=171a709f3628dad4f66300a49a1340ac41498730;p=dbsrgits%2FSQL-Abstract.git diff --git a/lib/SQL/Abstract.pm b/lib/SQL/Abstract.pm index 58f99cc..bf00228 100644 --- a/lib/SQL/Abstract.pm +++ b/lib/SQL/Abstract.pm @@ -8,17 +8,17 @@ package SQL::Abstract; # see doc at end of file use Carp; use strict; use warnings; -use List::Util qw/first/; -use Scalar::Util qw/blessed/; +use List::Util (); +use Scalar::Util (); #====================================================================== # GLOBALS #====================================================================== -our $VERSION = '1.61'; +our $VERSION = '1.68'; # This would confuse some packagers -#$VERSION = eval $VERSION; # numify for warning-free dev releases +$VERSION = eval $VERSION if $VERSION =~ /_/; # numify for warning-free dev releases our $AUTOLOAD; @@ -84,18 +84,18 @@ sub new { # try to recognize which are the 'equality' and 'unequality' ops # (temporary quickfix, should go through a more seasoned API) - $opt{equality_op} = qr/^(\Q$opt{cmp}\E|is|(is\s+)?like)$/i; - $opt{inequality_op} = qr/^(!=|<>|(is\s+)?not(\s+like)?)$/i; + $opt{equality_op} = qr/^(\Q$opt{cmp}\E|is|(is\s+)?like)$/i; + $opt{inequality_op} = qr/^(!=|<>|(is\s+)?not(\s+like)?)$/i; # SQL booleans $opt{sqltrue} ||= '1=1'; $opt{sqlfalse} ||= '0=1'; - # special operators + # special operators $opt{special_ops} ||= []; push @{$opt{special_ops}}, @BUILTIN_SPECIAL_OPS; - # unary operators + # unary operators $opt{unary_ops} ||= []; push @{$opt{unary_ops}}, @BUILTIN_UNARY_OPS; @@ -118,18 +118,24 @@ sub insert { my ($sql, @bind) = $self->$method($data); $sql = join " ", $self->_sqlcase('insert into'), $table, $sql; - if (my $fields = $options->{returning}) { - my $f = $self->_SWITCH_refkind($fields, { - ARRAYREF => sub {join ', ', map { $self->_quote($_) } @$fields;}, - SCALAR => sub {$self->_quote($fields)}, - SCALARREF => sub {$$fields}, - }); - $sql .= join " ", $self->_sqlcase(' returning'), $f; + if (my $ret = $options->{returning}) { + $sql .= $self->_insert_returning ($ret); } return wantarray ? ($sql, @bind) : $sql; } +sub _insert_returning { + my ($self, $fields) = @_; + + my $f = $self->_SWITCH_refkind($fields, { + ARRAYREF => sub {join ', ', map { $self->_quote($_) } @$fields;}, + SCALAR => sub {$self->_quote($fields)}, + SCALARREF => sub {$$fields}, + }); + return join (' ', $self->_sqlcase(' returning'), $f); +} + sub _insert_HASHREF { # explicit list of fields and then values my ($self, $data) = @_; @@ -185,7 +191,7 @@ sub _insert_values { $self->_SWITCH_refkind($v, { - ARRAYREF => sub { + ARRAYREF => sub { if ($self->{array_datatypes}) { # if array datatype are activated push @values, '?'; push @all_bind, $self->_bindtype($column, $v); @@ -205,7 +211,7 @@ sub _insert_values { push @all_bind, @bind; }, - # THINK : anything useful to do with a HASHREF ? + # THINK : anything useful to do with a HASHREF ? HASHREF => sub { # (nothing, but old SQLA passed it through) #TODO in SQLA >= 2.0 it will die instead belch "HASH ref as bind value in insert is not supported"; @@ -254,7 +260,7 @@ sub update { my $label = $self->_quote($k); $self->_SWITCH_refkind($v, { - ARRAYREF => sub { + ARRAYREF => sub { if ($self->{array_datatypes}) { # array datatype push @set, "$label = ?"; push @all_bind, $self->_bindtype($k, $v); @@ -314,11 +320,11 @@ sub select { my $f = (ref $fields eq 'ARRAY') ? join ', ', map { $self->_quote($_) } @$fields : $fields; - my $sql = join(' ', $self->_sqlcase('select'), $f, + my $sql = join(' ', $self->_sqlcase('select'), $f, $self->_sqlcase('from'), $table) . $where_sql; - return wantarray ? ($sql, @bind) : $sql; + return wantarray ? ($sql, @bind) : $sql; } #====================================================================== @@ -335,7 +341,7 @@ sub delete { my($where_sql, @bind) = $self->where($where); my $sql = $self->_sqlcase('delete from') . " $table" . $where_sql; - return wantarray ? ($sql, @bind) : $sql; + return wantarray ? ($sql, @bind) : $sql; } @@ -358,7 +364,7 @@ sub where { $sql .= $self->_order_by($order); } - return wantarray ? ($sql, @bind) : $sql; + return wantarray ? ($sql, @bind) : $sql; } @@ -368,12 +374,11 @@ sub _recurse_where { # dispatch on appropriate method according to refkind of $where my $method = $self->_METHOD_FOR_refkind("_where", $where); + my ($sql, @bind) = $self->$method($where, $logic); - my ($sql, @bind) = $self->$method($where, $logic); - - # DBIx::Class directly calls _recurse_where in scalar context, so + # DBIx::Class directly calls _recurse_where in scalar context, so # we must implement it, even if not in the official API - return wantarray ? ($sql, @bind) : $sql; + return wantarray ? ($sql, @bind) : $sql; } @@ -393,7 +398,7 @@ sub _where_ARRAYREF { my (@sql_clauses, @all_bind); # need to use while() so can shift() for pairs - while (my $el = shift @clauses) { + while (my $el = shift @clauses) { # switch according to kind of $el and get corresponding ($sql, @bind) my ($sql, @bind) = $self->_SWITCH_refkind($el, { @@ -401,13 +406,17 @@ sub _where_ARRAYREF { # skip empty elements, otherwise get invalid trailing AND stuff ARRAYREF => sub {$self->_recurse_where($el) if @$el}, - ARRAYREFREF => sub { @{${$el}} if @{${$el}}}, + ARRAYREFREF => sub { + my ($s, @b) = @$$el; + $self->_assert_bindval_matches_bindtype(@b); + ($s, @b); + }, HASHREF => sub {$self->_recurse_where($el, 'and') if %$el}, # LDNOTE : previous SQLA code for hashrefs was creating a dirty # side-effect: the first hashref within an array would change # the global logic to 'AND'. So [ {cond1, cond2}, [cond3, cond4] ] - # was interpreted as "(cond1 AND cond2) OR (cond3 AND cond4)", + # was interpreted as "(cond1 AND cond2) OR (cond3 AND cond4)", # whereas it should be "(cond1 AND cond2) OR (cond3 OR cond4)". SCALARREF => sub { ($$el); }, @@ -433,8 +442,8 @@ sub _where_ARRAYREF { sub _where_ARRAYREFREF { my ($self, $where) = @_; - my ($sql, @bind) = @{${$where}}; - + my ($sql, @bind) = @$$where; + $self->_assert_bindval_matches_bindtype(@bind); return ($sql, @bind); } @@ -446,15 +455,48 @@ sub _where_HASHREF { my ($self, $where) = @_; my (@sql_clauses, @all_bind); - for my $k (sort keys %$where) { + for my $k (sort keys %$where) { my $v = $where->{$k}; - # ($k => $v) is either a special op or a regular hashpair - my ($sql, @bind) = ($k =~ /^(-.+)/) ? $self->_where_op_in_hash($1, $v) - : do { - my $method = $self->_METHOD_FOR_refkind("_where_hashpair", $v); - $self->$method($k, $v); - }; + # ($k => $v) is either a special unary op or a regular hashpair + my ($sql, @bind) = do { + if ($k =~ /^-./) { + # put the operator in canonical form + my $op = $k; + $op =~ s/^-//; # remove initial dash + $op =~ s/[_\t ]+/ /g; # underscores and whitespace become single spaces + $op =~ s/^\s+|\s+$//g;# remove leading/trailing space + + $self->_debug("Unary OP(-$op) within hashref, recursing..."); + + my $op_entry = List::Util::first {$op =~ $_->{regex}} @{$self->{unary_ops}}; + if (my $handler = $op_entry->{handler}) { + if (not ref $handler) { + if ($op =~ s/\s?\d+$//) { + belch 'Use of [and|or|nest]_N modifiers is deprecated and will be removed in SQLA v2.0. ' + . "You probably wanted ...-and => [ -$op => COND1, -$op => COND2 ... ]"; + } + $self->$handler ($op, $v); + } + elsif (ref $handler eq 'CODE') { + $handler->($self, $op, $v); + } + else { + puke "Illegal handler for operator $k - expecting a method name or a coderef"; + } + } + else { + $self->debug("Generic unary OP: $k - recursing as function"); + my ($sql, @bind) = $self->_where_func_generic ($op, $v); + $sql = "($sql)" unless (defined($self->{_nested_func_lhs}) && ($self->{_nested_func_lhs} eq $k)); # top level vs nested + ($sql, @bind); + } + } + else { + my $method = $self->_METHOD_FOR_refkind("_where_hashpair", $v); + $self->$method($k, $v); + } + }; push @sql_clauses, $sql; push @all_bind, @bind; @@ -463,40 +505,34 @@ sub _where_HASHREF { return $self->_join_sql_clauses('and', \@sql_clauses, \@all_bind); } +sub _where_func_generic { + my ($self, $op, $rhs) = @_; -sub _where_op_in_hash { - my ($self, $orig_op, $v) = @_; + my ($sql, @bind) = $self->_SWITCH_refkind ($rhs, { + SCALAR => sub { + puke "Illegal use of top-level '$op'" + unless $self->{_nested_func_lhs}; - # put the operator in canonical form - my $op = $orig_op; - $op =~ s/^-//; # remove initial dash - $op =~ s/[_\t ]+/ /g; # underscores and whitespace become single spaces - $op =~ s/^\s+|\s+$//g;# remove leading/trailing space + return ( + $self->_convert('?'), + $self->_bindtype($self->{_nested_func_lhs}, $rhs) + ); + }, + FALLBACK => sub { + $self->_recurse_where ($rhs) + }, + }); - $self->_debug("OP(-$op) within hashref, recursing..."); + $sql = sprintf ('%s %s', + $self->_sqlcase($op), + $sql, + ); - my $op_entry = first {$op =~ $_->{regex}} @{$self->{unary_ops}}; - my $handler = $op_entry->{handler}; - if (! $handler) { - puke "unknown operator: $orig_op"; - } - elsif (not ref $handler) { - if ($op =~ s/\s?\d+$//) { - belch 'Use of [and|or|nest]_N modifiers is deprecated and will be removed in SQLA v2.0. ' - . "You probably wanted ...-and => [ -$op => COND1, -$op => COND2 ... ]"; - } - return $self->$handler ($op, $v); - } - elsif (ref $handler eq 'CODE') { - return $handler->($self, $op, $v); - } - else { - puke "Illegal handler for operator $orig_op - expecting a method name or a coderef"; - } + return ($sql, @bind); } sub _where_op_ANDOR { - my ($self, $op, $v) = @_; + my ($self, $op, $v) = @_; $self->_SWITCH_refkind($v, { ARRAYREF => sub { @@ -509,7 +545,7 @@ sub _where_op_ANDOR { : $self->_where_HASHREF($v); }, - SCALARREF => sub { + SCALARREF => sub { puke "-$op => \\\$scalar not supported, use -nest => ..."; }, @@ -528,73 +564,55 @@ sub _where_op_ANDOR { } sub _where_op_NEST { - my ($self, $op, $v) = @_; + my ($self, $op, $v) = @_; $self->_SWITCH_refkind($v, { - ARRAYREF => sub { - return $self->_where_ARRAYREF($v, ''); - }, - - HASHREF => sub { - return $self->_where_HASHREF($v); - }, - - SCALARREF => sub { # literal SQL - return ($$v); - }, - - ARRAYREFREF => sub { # literal SQL - return @{${$v}}; - }, - SCALAR => sub { # permissively interpreted as SQL belch "literal SQL should be -nest => \\'scalar' " . "instead of -nest => 'scalar' "; - return ($v); + return ($v); }, UNDEF => sub { puke "-$op => undef not supported"; }, + + FALLBACK => sub { + $self->_recurse_where ($v); + }, + }); } sub _where_op_BOOL { - my ($self, $op, $v) = @_; + my ($self, $op, $v) = @_; - my ( $prefix, $suffix ) = ( $op =~ /\bnot\b/i ) - ? ( '(NOT ', ')' ) + my ( $prefix, $suffix ) = ( $op =~ /\bnot\b/i ) + ? ( '(NOT ', ')' ) : ( '', '' ); - $self->_SWITCH_refkind($v, { - ARRAYREF => sub { - my ( $sql, @bind ) = $self->_where_ARRAYREF($v, ''); - return ( ($prefix . $sql . $suffix), @bind ); - }, - ARRAYREFREF => sub { - my ( $sql, @bind ) = @{ ${$v} }; - return ( ($prefix . $sql . $suffix), @bind ); - }, - - HASHREF => sub { - my ( $sql, @bind ) = $self->_where_HASHREF($v); - return ( ($prefix . $sql . $suffix), @bind ); - }, + my ($sql, @bind) = do { + $self->_SWITCH_refkind($v, { + SCALAR => sub { # interpreted as SQL column + $self->_convert($self->_quote($v)); + }, - SCALARREF => sub { # literal SQL - return ($prefix . $$v . $suffix); - }, + UNDEF => sub { + puke "-$op => undef not supported"; + }, - SCALAR => sub { # interpreted as SQL column - return ($prefix . $self->_convert($self->_quote($v)) . $suffix); - }, + FALLBACK => sub { + $self->_recurse_where ($v); + }, + }); + }; - UNDEF => sub { - puke "-$op => undef not supported"; - }, - }); + return ( + join ('', $prefix, $sql, $suffix), + @bind, + ); } @@ -621,7 +639,7 @@ sub _where_hashpair_ARRAYREF { my $logic = $op ? substr($op, 1) : ''; return $self->_recurse_where(\@distributed, $logic); - } + } else { # LDNOTE : not sure of this one. What does "distribute over nothing" mean? $self->_debug("empty ARRAY($k) means 0=1"); @@ -633,6 +651,8 @@ sub _where_hashpair_HASHREF { my ($self, $k, $v, $logic) = @_; $logic ||= 'and'; + local $self->{_nested_func_lhs} = $self->{_nested_func_lhs}; + my ($all_sql, @all_bind); for my $orig_op (sort keys %$v) { @@ -646,9 +666,12 @@ sub _where_hashpair_HASHREF { my ($sql, @bind); + # CASE: col-value logic modifiers + if ( $orig_op =~ /^ \- (and|or) $/xi ) { + ($sql, @bind) = $self->_where_hashpair_HASHREF($k, $val, $1); + } # CASE: special operators like -in or -between - my $special_op = first {$op =~ $_->{regex}} @{$self->{special_ops}}; - if ($special_op) { + elsif ( my $special_op = List::Util::first {$op =~ $_->{regex}} @{$self->{special_ops}} ) { my $handler = $special_op->{handler}; if (! $handler) { puke "No handler supplied for special operator $orig_op"; @@ -670,12 +693,6 @@ sub _where_hashpair_HASHREF { ($sql, @bind) = $self->_where_field_op_ARRAYREF($k, $op, $val); }, - SCALARREF => sub { # CASE: col => {op => \$scalar} (literal SQL without bind) - $sql = join ' ', $self->_convert($self->_quote($k)), - $self->_sqlcase($op), - $$val; - }, - ARRAYREFREF => sub { # CASE: col => {op => \[$sql, @bind]} (literal SQL with bind) my ($sub_sql, @sub_bind) = @$$val; $self->_assert_bindval_matches_bindtype(@sub_bind); @@ -685,10 +702,6 @@ sub _where_hashpair_HASHREF { @bind = @sub_bind; }, - HASHREF => sub { - ($sql, @bind) = $self->_where_hashpair_HASHREF($k, $val, $op); - }, - UNDEF => sub { # CASE: col => {op => undef} : sql "IS (NOT)? NULL" my $is = ($op =~ $self->{equality_op}) ? 'is' : ($op =~ $self->{inequality_op}) ? 'is not' : @@ -696,11 +709,17 @@ sub _where_hashpair_HASHREF { $sql = $self->_quote($k) . $self->_sqlcase(" $is null"); }, - FALLBACK => sub { # CASE: col => {op => $scalar} - $sql = join ' ', $self->_convert($self->_quote($k)), - $self->_sqlcase($op), - $self->_convert('?'); - @bind = $self->_bindtype($k, $val); + FALLBACK => sub { # CASE: col => {op/func => $stuff} + + # retain for proper column type bind + $self->{_nested_func_lhs} ||= $k; + + ($sql, @bind) = $self->_where_func_generic ($op, $val); + + $sql = join (' ', + $self->_convert($self->_quote($k)), + $self->{_nested_func_lhs} eq $k ? $sql : "($sql)", # top level vs nested + ); }, }); } @@ -734,18 +753,18 @@ sub _where_field_op_ARRAYREF { # distribute $op over each remaining member of @vals, append logic if exists return $self->_recurse_where([map { {$k => {$op, $_}} } @vals], $logic); - # LDNOTE : had planned to change the distribution logic when - # $op =~ $self->{inequality_op}, because of Morgan laws : + # LDNOTE : had planned to change the distribution logic when + # $op =~ $self->{inequality_op}, because of Morgan laws : # with {field => {'!=' => [22, 33]}}, it would be ridiculous to generate - # WHERE field != 22 OR field != 33 : the user probably means + # WHERE field != 22 OR field != 33 : the user probably means # WHERE field != 22 AND field != 33. # To do this, replace the above to roughly : # my $logic = ($op =~ $self->{inequality_op}) ? 'AND' : 'OR'; # return $self->_recurse_where([map { {$k => {$op, $_}} } @vals], $logic); - } + } else { - # try to DWIM on equality operators + # try to DWIM on equality operators # LDNOTE : not 100% sure this is the correct thing to do ... return ($self->{sqlfalse}) if $op =~ $self->{equality_op}; return ($self->{sqltrue}) if $op =~ $self->{inequality_op}; @@ -767,7 +786,7 @@ sub _where_hashpair_SCALARREF { sub _where_hashpair_ARRAYREFREF { my ($self, $k, $v) = @_; $self->_debug("REF($k) means literal SQL: @${$v}"); - my ($sql, @bind) = @${$v}; + my ($sql, @bind) = @$$v; $self->_assert_bindval_matches_bindtype(@bind); $sql = $self->_quote($k) . " " . $sql; return ($sql, @bind ); @@ -777,8 +796,8 @@ sub _where_hashpair_ARRAYREFREF { sub _where_hashpair_SCALAR { my ($self, $k, $v) = @_; $self->_debug("NOREF($k) means simple key=val: $k $self->{cmp} $v"); - my $sql = join ' ', $self->_convert($self->_quote($k)), - $self->_sqlcase($self->{cmp}), + my $sql = join ' ', $self->_convert($self->_quote($k)), + $self->_sqlcase($self->{cmp}), $self->_convert('?'); my @bind = $self->_bindtype($k, $v); return ( $sql, @bind); @@ -837,7 +856,9 @@ sub _where_field_BETWEEN { my ($clause, @bind) = $self->_SWITCH_refkind($vals, { ARRAYREFREF => sub { - return @$$vals; + my ($s, @b) = @$$vals; + $self->_assert_bindval_matches_bindtype(@b); + ($s, @b); }, SCALARREF => sub { return $$vals; @@ -857,6 +878,7 @@ sub _where_field_BETWEEN { }, ARRAYREFREF => sub { my ($sql, @bind) = @$$val; + $self->_assert_bindval_matches_bindtype(@bind); return ($self->_convert($sql), @bind); }, }); @@ -968,7 +990,11 @@ sub _order_by_chunks { map { $self->_order_by_chunks ($_ ) } @$arg; }, - ARRAYREFREF => sub { [ @$$arg ] }, + ARRAYREFREF => sub { + my ($s, @b) = @$$arg; + $self->_assert_bindval_matches_bindtype(@b); + [ $s, @b ]; + }, SCALAR => sub {$self->_quote($arg)}, @@ -978,11 +1004,11 @@ sub _order_by_chunks { HASHREF => sub { # get first pair in hash - my ($key, $val) = each %$arg; + my ($key, $val, @rest) = %$arg; return () unless $key; - if ( (keys %$arg) > 1 or not $key =~ /^-(desc|asc)/i ) { + if ( @rest or not $key =~ /^-(desc|asc)/i ) { puke "hash passed to _order_by must have exactly one key (-desc or -asc)"; } @@ -1032,80 +1058,78 @@ sub _table { # UTILITY FUNCTIONS #====================================================================== +# highly optimized, as it's called way too often sub _quote { - my $self = shift; - my $label = shift; - - $label or puke "can't quote an empty label"; - - # left and right quote characters - my ($ql, $qr, @other) = $self->_SWITCH_refkind($self->{quote_char}, { - SCALAR => sub {($self->{quote_char}, $self->{quote_char})}, - ARRAYREF => sub {@{$self->{quote_char}}}, - UNDEF => sub {()}, - }); - not @other - or puke "quote_char must be an arrayref of 2 values"; - - # no quoting if no quoting chars - $ql or return $label; + # my ($self, $label) = @_; - # no quoting for literal SQL - return $$label if ref($label) eq 'SCALAR'; + return '' unless defined $_[1]; + return ${$_[1]} if ref($_[1]) eq 'SCALAR'; - # separate table / column (if applicable) - my $sep = $self->{name_sep} || ''; - my @to_quote = $sep ? split /\Q$sep\E/, $label : ($label); + return $_[1] unless $_[0]->{quote_char}; - # do the quoting, except for "*" or for `table`.* - my @quoted = map { $_ eq '*' ? $_: $ql.$_.$qr} @to_quote; + my $qref = ref $_[0]->{quote_char}; + my ($l, $r); + if (!$qref) { + ($l, $r) = ( $_[0]->{quote_char}, $_[0]->{quote_char} ); + } + elsif ($qref eq 'ARRAY') { + ($l, $r) = @{$_[0]->{quote_char}}; + } + else { + puke "Unsupported quote_char format: $_[0]->{quote_char}"; + } - # reassemble and return. - return join $sep, @quoted; + # parts containing * are naturally unquoted + return join( $_[0]->{name_sep}||'', map + { $_ eq '*' ? $_ : $l . $_ . $r } + ( $_[0]->{name_sep} ? split (/\Q$_[0]->{name_sep}\E/, $_[1] ) : $_[1] ) + ); } # Conversion, if applicable sub _convert ($) { - my ($self, $arg) = @_; + #my ($self, $arg) = @_; # LDNOTE : modified the previous implementation below because # it was not consistent : the first "return" is always an array, # the second "return" is context-dependent. Anyway, _convert -# seems always used with just a single argument, so make it a +# seems always used with just a single argument, so make it a # scalar function. # return @_ unless $self->{convert}; # my $conv = $self->_sqlcase($self->{convert}); # my @ret = map { $conv.'('.$_.')' } @_; # return wantarray ? @ret : $ret[0]; - if ($self->{convert}) { - my $conv = $self->_sqlcase($self->{convert}); - $arg = $conv.'('.$arg.')'; + if ($_[0]->{convert}) { + return $_[0]->_sqlcase($_[0]->{convert}) .'(' . $_[1] . ')'; } - return $arg; + return $_[1]; } # And bindtype sub _bindtype (@) { - my $self = shift; - my($col, @vals) = @_; + #my ($self, $col, @vals) = @_; - #LDNOTE : changed original implementation below because it did not make + #LDNOTE : changed original implementation below because it did not make # sense when bindtype eq 'columns' and @vals > 1. # return $self->{bindtype} eq 'columns' ? [ $col, @vals ] : @vals; - return $self->{bindtype} eq 'columns' ? map {[$col, $_]} @vals : @vals; + # called often - tighten code + return $_[0]->{bindtype} eq 'columns' + ? map {[$_[1], $_]} @_[2 .. $#_] + : @_[2 .. $#_] + ; } # Dies if any element of @bind is not in [colname => value] format # if bindtype is 'columns'. sub _assert_bindval_matches_bindtype { - my ($self, @bind) = @_; - +# my ($self, @bind) = @_; + my $self = shift; if ($self->{bindtype} eq 'columns') { - foreach my $val (@bind) { - if (!defined $val || ref($val) ne 'ARRAY' || @$val != 2) { - die "bindtype 'columns' selected, you need to pass: [column_name => bind_value]" + for (@_) { + if (!defined $_ || ref($_) ne 'ARRAY' || @$_ != 2) { + puke "bindtype 'columns' selected, you need to pass: [column_name => bind_value]" } } } @@ -1130,11 +1154,9 @@ sub _join_sql_clauses { # Fix SQL case, if so requested sub _sqlcase { - my $self = shift; - # LDNOTE: if $self->{case} is true, then it contains 'lower', so we # don't touch the argument ... crooked logic, but let's not change it! - return $self->{case} ? $_[0] : uc($_[0]); + return $_[0]->{case} ? $_[1] : uc($_[1]); } @@ -1144,48 +1166,57 @@ sub _sqlcase { sub _refkind { my ($self, $data) = @_; - my $suffix = ''; - my $ref; - my $n_steps = 0; - - while (1) { - # blessed objects are treated like scalars - $ref = (blessed $data) ? '' : ref $data; - $n_steps += 1 if $ref; - last if $ref ne 'REF'; - $data = $$data; - } - my $base = $ref || (defined $data ? 'SCALAR' : 'UNDEF'); + return 'UNDEF' unless defined $data; - return $base . ('REF' x $n_steps); -} + # blessed objects are treated like scalars + my $ref = (Scalar::Util::blessed $data) ? '' : ref $data; + + return 'SCALAR' unless $ref; + my $n_steps = 1; + while ($ref eq 'REF') { + $data = $$data; + $ref = (Scalar::Util::blessed $data) ? '' : ref $data; + $n_steps++ if $ref; + } + return ($ref||'SCALAR') . ('REF' x $n_steps); +} sub _try_refkind { my ($self, $data) = @_; my @try = ($self->_refkind($data)); push @try, 'SCALAR_or_UNDEF' if $try[0] eq 'SCALAR' || $try[0] eq 'UNDEF'; push @try, 'FALLBACK'; - return @try; + return \@try; } sub _METHOD_FOR_refkind { my ($self, $meth_prefix, $data) = @_; - my $method = first {$_} map {$self->can($meth_prefix."_".$_)} - $self->_try_refkind($data) - or puke "cannot dispatch on '$meth_prefix' for ".$self->_refkind($data); - return $method; + + my $method; + for (@{$self->_try_refkind($data)}) { + $method = $self->can($meth_prefix."_".$_) + and last; + } + + return $method || puke "cannot dispatch on '$meth_prefix' for ".$self->_refkind($data); } sub _SWITCH_refkind { my ($self, $data, $dispatch_table) = @_; - my $coderef = first {$_} map {$dispatch_table->{$_}} - $self->_try_refkind($data) - or puke "no dispatch entry for ".$self->_refkind($data); + my $coderef; + for (@{$self->_try_refkind($data)}) { + $coderef = $dispatch_table->{$_} + and last; + } + + puke "no dispatch entry for ".$self->_refkind($data) + unless $coderef; + $coderef->(); } @@ -1210,7 +1241,7 @@ sub values { foreach my $k ( sort keys %$data ) { my $v = $data->{$k}; $self->_SWITCH_refkind($v, { - ARRAYREF => sub { + ARRAYREF => sub { if ($self->{array_datatypes}) { # array datatype push @all_bind, $self->_bindtype($k, $v); } @@ -1257,7 +1288,7 @@ sub generate { } elsif ($r eq 'SCALAR') { # literal SQL without bind push @sqlq, "$label = $$v"; - } else { + } else { push @sqlq, "$label = ?"; push @sqlv, $self->_bindtype($k, $v); } @@ -1275,7 +1306,7 @@ sub generate { } elsif ($r eq 'SCALAR') { # literal SQL without bind # embedded literal SQL push @sqlq, $$v; - } else { + } else { push @sqlq, '?'; push @sqlv, $v; } @@ -1398,14 +1429,14 @@ These are then used directly in your DBI code: If your database has array types (like for example Postgres), activate the special option C<< array_datatypes => 1 >> -when creating the C object. +when creating the C object. Then you may use an arrayref to insert and update database array types: my $sql = SQL::Abstract->new(array_datatypes => 1); my %data = ( planets => [qw/Mercury Venus Earth Mars/] ); - + my($stmt, @bind) = $sql->insert('solar_system', \%data); This results in: @@ -1425,7 +1456,7 @@ say something like this: my %data = ( name => 'Bill', date_entered => \["to_date(?,'MM/DD/YYYY')", "03/02/2003"], - ); + ); The first value in the array is the actual SQL. Any other values are optional and would be included in the bind values array. This gives @@ -1433,7 +1464,7 @@ you: my($stmt, @bind) = $sql->insert('people', \%data); - $stmt = "INSERT INTO people (name, date_entered) + $stmt = "INSERT INTO people (name, date_entered) VALUES (?, to_date(?,'MM/DD/YYYY'))"; @bind = ('Bill', '03/02/2003'); @@ -1480,7 +1511,7 @@ Easy, eh? The functions are simple. There's one for each major SQL operation, and a constructor you use first. The arguments are specified in a -similar order to each function (table, then fields, then a where +similar order to each function (table, then fields, then a where clause) to try and simplify things. @@ -1537,8 +1568,8 @@ for arrays, and "and" for hashes. This means that a WHERE array of the form: @where = ( - event_date => {'>=', '2/13/99'}, - event_date => {'<=', '4/24/03'}, + event_date => {'>=', '2/13/99'}, + event_date => {'<=', '4/24/03'}, ); will generate SQL like this: @@ -1557,7 +1588,7 @@ Which will change the above C to: The logic can also be changed locally by inserting a modifier in front of an arrayref : - @where = (-and => [event_date => {'>=', '2/13/99'}, + @where = (-and => [event_date => {'>=', '2/13/99'}, event_date => {'<=', '4/24/03'} ]); See the L section for explanations. @@ -1635,7 +1666,7 @@ will expect the bind values in this format. =item quote_char This is the character that a table or column name will be quoted -with. By default this is an empty string, but you could set it to +with. By default this is an empty string, but you could set it to the character C<`>, to generate SQL like this: SELECT `a_field` FROM `a_table` WHERE `some_field` LIKE '%someval%' @@ -1647,7 +1678,7 @@ that generates SQL like this: SELECT [a_field] FROM [a_table] WHERE [some_field] LIKE '%someval%' -Quoting is useful if you have tables or columns names that are reserved +Quoting is useful if you have tables or columns names that are reserved words in your database's SQL dialect. =item name_sep @@ -1660,8 +1691,8 @@ so that tables and column names can be individually quoted like this: =item array_datatypes -When this option is true, arrayrefs in INSERT or UPDATE are -interpreted as array datatypes and are passed directly +When this option is true, arrayrefs in INSERT or UPDATE are +interpreted as array datatypes and are passed directly to the DBI layer. When this option is false, arrayrefs are interpreted as literal SQL, just like refs to arrayrefs @@ -1672,13 +1703,13 @@ for literal SQL). =item special_ops -Takes a reference to a list of "special operators" +Takes a reference to a list of "special operators" to extend the syntax understood by L. See section L for details. =item unary_ops -Takes a reference to a list of "unary operators" +Takes a reference to a list of "unary operators" to extend the syntax understood by L. See section L for details. @@ -1723,14 +1754,14 @@ with those data types. =head2 select($source, $fields, $where, $order) -This returns a SQL SELECT statement and associated list of bind values, as +This returns a SQL SELECT statement and associated list of bind values, as specified by the arguments : =over =item $source -Specification of the 'FROM' part of the statement. +Specification of the 'FROM' part of the statement. The argument can be either a plain scalar (interpreted as a table name, will be quoted), or an arrayref (interpreted as a list of table names, joined by commas, quoted), or a scalarref @@ -1739,10 +1770,10 @@ of table names, joined by commas, quoted), or a scalarref =item $fields -Specification of the list of fields to retrieve from +Specification of the list of fields to retrieve from the source. The argument can be either an arrayref (interpreted as a list -of field names, will be joined by commas and quoted), or a +of field names, will be joined by commas and quoted), or a plain scalar (literal SQL, not quoted). Please observe that this API is not as flexible as for the first argument C<$table>, for backwards compatibility reasons. @@ -1751,13 +1782,13 @@ the first argument C<$table>, for backwards compatibility reasons. Optional argument to specify the WHERE part of the query. The argument is most often a hashref, but can also be -an arrayref or plain scalar -- +an arrayref or plain scalar -- see section L for details. =item $order Optional argument to specify the ORDER BY part of the query. -The argument can be a scalar, a hashref or an arrayref +The argument can be a scalar, a hashref or an arrayref -- see section L for details. @@ -1862,13 +1893,27 @@ an arrayref: ); This simple code will create the following: - + $stmt = "WHERE user = ? AND ( status = ? OR status = ? OR status = ? )"; @bind = ('nwiger', 'assigned', 'in-progress', 'pending'); -A field associated to an empty arrayref will be considered a +A field associated to an empty arrayref will be considered a logical false and will generate 0=1. +=head2 Tests for NULL values + +If the value part is C then this is converted to SQL + + my %where = ( + user => 'nwiger', + status => undef, + ); + +becomes: + + $stmt = "WHERE user = ? AND status IS NULL"; + @bind = ('nwiger'); + =head2 Specific comparison operators If you want to specify a different type of operator for your comparison, @@ -1952,7 +1997,7 @@ Because, in Perl you I do this: As the second C key will obliterate the first. The solution is to use the special C<-modifier> form inside an arrayref: - priority => [ -and => {'!=', 2}, + priority => [ -and => {'!=', 2}, {'!=', 1} ] @@ -1993,7 +2038,7 @@ Which would generate: $stmt = "WHERE status = ? AND reportid IN (?,?,?)"; @bind = ('completed', '567', '2335', '2'); -The reverse operator C<-not_in> generates SQL C and is used in +The reverse operator C<-not_in> generates SQL C and is used in the same way. If the argument to C<-in> is an empty array, 'sqlfalse' is generated @@ -2042,7 +2087,7 @@ are possible: start0 => { -between => [ 1, 2 ] }, start1 => { -between => \["? AND ?", 1, 2] }, start2 => { -between => \"lower(x) AND upper(y)" }, - start3 => { -between => [ + start3 => { -between => [ \"lower(x)", \["upper(?)", 'stuff' ], ] }, @@ -2059,7 +2104,7 @@ Would give you: @bind = (1, 2, 1, 2, 'stuff'); -These are the two builtin "special operators"; but the +These are the two builtin "special operators"; but the list can be expanded : see section L below. =head2 Unary operators: bool @@ -2067,7 +2112,7 @@ list can be expanded : see section L below. If you wish to test against boolean columns or functions within your database you can use the C<-bool> and C<-not_bool> operators. For example to test the column C being true and the column - being false you would use:- +C being false you would use:- my %where = ( -bool => 'is_user', @@ -2150,7 +2195,7 @@ inside : That would yield: - WHERE ( user = ? AND + WHERE ( user = ? AND ( ( workhrs > ? AND geo = ? ) OR ( workhrs < ? AND geo = ? ) ) ) @@ -2170,18 +2215,18 @@ C the arrayref. Here is an example combining both features : yielding - WHERE ( ( ( a = ? AND b = ? ) - OR ( c = ? OR d = ? ) + WHERE ( ( ( a = ? AND b = ? ) + OR ( c = ? OR d = ? ) OR ( e LIKE ? AND e LIKE ? ) ) ) This difference in syntax is unfortunate but must be preserved for historical reasons. So be careful : the two examples below would seem algebraically equivalent, but they are not - {col => [-and => {-like => 'foo%'}, {-like => '%bar'}]} + {col => [-and => {-like => 'foo%'}, {-like => '%bar'}]} # yields : WHERE ( ( col LIKE ? AND col LIKE ? ) ) - [-and => {col => {-like => 'foo%'}, {col => {-like => '%bar'}}]] + [-and => {col => {-like => 'foo%'}, {col => {-like => '%bar'}}]] # yields : WHERE ( ( col LIKE ? OR col LIKE ? ) ) @@ -2230,6 +2275,17 @@ which yields $stmt = "WHERE priority < ? AND is_ready"; @bind = ('2'); +Literal SQL is also the only way to compare 2 columns to one another: + + my %where = ( + priority => { '<', 2 }, + requestor => \'= submittor' + ); + +which creates: + + $stmt = "WHERE priority < ? AND requestor = submitter"; + @bind = ('2'); =head2 Literal SQL with placeholders and bind values (subqueries) @@ -2271,17 +2327,17 @@ main SQL query. Here is a first example : This yields : - $stmt = "WHERE (foo = ? AND bar IN (SELECT c1 FROM t1 + $stmt = "WHERE (foo = ? AND bar IN (SELECT c1 FROM t1 WHERE c2 < ? AND c3 LIKE ?))"; @bind = (1234, 100, "foo%"); -Other subquery operators, like for example C<"E ALL"> or C<"NOT IN">, +Other subquery operators, like for example C<"E ALL"> or C<"NOT IN">, are expressed in the same way. Of course the C<$sub_stmt> and -its associated bind values can be generated through a former call +its associated bind values can be generated through a former call to C : my ($sub_stmt, @sub_bind) - = $sql->select("t1", "c1", {c2 => {"<" => 100}, + = $sql->select("t1", "c1", {c2 => {"<" => 100}, c3 => {-like => "foo%"}}); my %where = ( foo => 1234, @@ -2289,10 +2345,10 @@ to C : ); In the examples above, the subquery was used as an operator on a column; -but the same principle also applies for a clause within the main C<%where> +but the same principle also applies for a clause within the main C<%where> hash, like an EXISTS subquery : - my ($sub_stmt, @sub_bind) + my ($sub_stmt, @sub_bind) = $sql->select("t1", "*", {c1 => 1, c2 => \"> t0.c0"}); my %where = ( foo => 1234, @@ -2301,21 +2357,21 @@ hash, like an EXISTS subquery : which yields - $stmt = "WHERE (foo = ? AND EXISTS (SELECT * FROM t1 + $stmt = "WHERE (foo = ? AND EXISTS (SELECT * FROM t1 WHERE c1 = ? AND c2 > t0.c0))"; @bind = (1234, 1); -Observe that the condition on C in the subquery refers to -column C of the main query : this is I a bind -value, so we have to express it through a scalar ref. +Observe that the condition on C in the subquery refers to +column C of the main query : this is I a bind +value, so we have to express it through a scalar ref. Writing C<< c2 => {">" => "t0.c0"} >> would have generated C<< c2 > ? >> with bind value C<"t0.c0"> ... not exactly what we wanted here. Another use of the subquery technique is when some SQL clauses need parentheses, as it often occurs with some proprietary SQL extensions -like for example fulltext expressions, geospatial expressions, +like for example fulltext expressions, geospatial expressions, NATIVE clauses, etc. Here is an example of a fulltext query in MySQL : my %where = ( @@ -2325,7 +2381,7 @@ NATIVE clauses, etc. Here is an example of a fulltext query in MySQL : Finally, here is an example where a subquery is used for expressing unary negation: - my ($sub_stmt, @sub_bind) + my ($sub_stmt, @sub_bind) = $sql->where({age => [{"<" => 10}, {">" => 20}]}); $sub_stmt =~ s/^ where //i; # don't want "WHERE" in the subclause my %where = ( @@ -2360,7 +2416,7 @@ script. =head1 ORDER BY CLAUSES -Some functions take an order by clause. This can either be a scalar (just a +Some functions take an order by clause. This can either be a scalar (just a column name,) a hash of C<< { -desc => 'col' } >> or C<< { -asc => 'col' } >>, or an array of either of the two previous forms. Examples: @@ -2379,7 +2435,7 @@ or an array of either of the two previous forms. Examples: | ['colA', {-asc => 'colB'}] | ORDER BY colA, colB ASC | - { -asc => [qw/colA colB] } | ORDER BY colA ASC, colB ASC + { -asc => [qw/colA colB/] } | ORDER BY colA ASC, colB ASC | [ | { -asc => 'colA' }, | ORDER BY colA ASC, colB DESC, @@ -2406,9 +2462,9 @@ or an array of either of the two previous forms. Examples: }, ]); -A "special operator" is a SQL syntactic clause that can be +A "special operator" is a SQL syntactic clause that can be applied to a field, instead of a usual binary operator. -For example : +For example : WHERE field IN (?, ?, ?) WHERE field BETWEEN ? AND ? @@ -2452,13 +2508,13 @@ When supplied with a coderef, it is called as: =back -For example, here is an implementation +For example, here is an implementation of the MATCH .. AGAINST syntax for MySQL my $sqlmaker = SQL::Abstract->new(special_ops => [ - + # special op for MySql MATCH (field) AGAINST(word1, word2, ...) - {regex => qr/^match$/i, + {regex => qr/^match$/i, handler => sub { my ($self, $field, $op, $arg) = @_; $arg = [$arg] if not ref $arg; @@ -2471,7 +2527,7 @@ of the MATCH .. AGAINST syntax for MySQL return ($sql, @bind); } }, - + ]); @@ -2491,7 +2547,7 @@ of the MATCH .. AGAINST syntax for MySQL }, ]); -A "unary operator" is a SQL syntactic clause that can be +A "unary operator" is a SQL syntactic clause that can be applied to a field - the operator goes before the field You can write your own operator handlers - supply a C @@ -2553,6 +2609,12 @@ the same structure, you only have to generate the SQL the first time around. On subsequent queries, simply use the C function provided by this module to return your values in the correct order. +However this depends on the values having the same type - if, for +example, the values of a where clause may either have values +(resulting in sql of the form C with a single bind +value), or alternatively the values might be C (resulting in +sql of the form C with no bind value) then the +caching technique suggested will not work. =head1 FORMBUILDER @@ -2579,7 +2641,7 @@ query, but the point is that if you make your form look like your table, the actual query script can be extremely simplistic. If you're B lazy (I am), check out C for -a fast interface to returning and formatting data. I frequently +a fast interface to returning and formatting data. I frequently use these three modules together to write complex database query apps in under 50 lines. @@ -2589,17 +2651,17 @@ apps in under 50 lines. Version 1.50 was a major internal refactoring of C. Great care has been taken to preserve the I behavior documented in previous versions in the 1.* family; however, -some features that were previously undocumented, or behaved +some features that were previously undocumented, or behaved differently from the documentation, had to be changed in order to clarify the semantics. Hence, client code that was relying -on some dark areas of C v1.* +on some dark areas of C v1.* B in v1.50. The main changes are : =over -=item * +=item * support for literal SQL through the C<< \ [$sql, bind] >> syntax. @@ -2615,7 +2677,7 @@ support for the { operator => \["...", @bind] } construct (to embed literal SQL optional support for L -=item * +=item * defensive programming : check arguments @@ -2633,7 +2695,7 @@ as C<< "(cond1 AND cond2) OR (cond3 OR cond4)" >>. fixed semantics of _bindtype on array args -=item * +=item * dropped the C<_anoncopy> of the %where tree. No longer necessary, we just avoid shifting arrays within that tree. @@ -2652,7 +2714,7 @@ There are a number of individuals that have really helped out with this module. Unfortunately, most of them submitted bugs via CPAN so I have no idea who they are! But the people I do know are: - Ash Berlin (order_by hash term support) + Ash Berlin (order_by hash term support) Matt Trout (DBIx::Class support) Mark Stosberg (benchmarking) Chas Owens (initial "IN" operator support)