X-Git-Url: http://git.shadowcat.co.uk/gitweb/gitweb.cgi?a=blobdiff_plain;f=lib%2FSQL%2FAbstract.pm;h=bf0022875f502fe139acae9016c6e0619d17f311;hb=4ee5e99cdf609a1ab37e8a9de8fcf222f7fd5010;hp=6fb68e9a6b48e68fed421572488b2093d66e610d;hpb=5db47f9fe6d06048b35092378a04e3a292d03a19;p=dbsrgits%2FSQL-Abstract.git diff --git a/lib/SQL/Abstract.pm b/lib/SQL/Abstract.pm index 6fb68e9..bf00228 100644 --- a/lib/SQL/Abstract.pm +++ b/lib/SQL/Abstract.pm @@ -8,24 +8,34 @@ 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 $VERSION = '1.49_01'; -$VERSION = eval $VERSION; # numify for warning-free dev releases +our $VERSION = '1.68'; +# This would confuse some packagers +$VERSION = eval $VERSION if $VERSION =~ /_/; # numify for warning-free dev releases our $AUTOLOAD; # special operators (-in, -between). May be extended/overridden by user. # See section WHERE: BUILTIN SPECIAL OPERATORS below for implementation my @BUILTIN_SPECIAL_OPS = ( - {regex => qr/^(not )?between$/i, handler => \&_where_field_BETWEEN}, - {regex => qr/^(not )?in$/i, handler => \&_where_field_IN}, + {regex => qr/^(not )?between$/i, handler => '_where_field_BETWEEN'}, + {regex => qr/^(not )?in$/i, handler => '_where_field_IN'}, +); + +# unaryish operators - key maps to handler +my @BUILTIN_UNARY_OPS = ( + # the digits are backcompat stuff + { regex => qr/^and (?: \s? \d+ )? $/xi, handler => '_where_op_ANDOR' }, + { regex => qr/^or (?: \s? \d+ )? $/xi, handler => '_where_op_ANDOR' }, + { regex => qr/^nest (?: \s? \d+ )? $/xi, handler => '_where_op_NEST' }, + { regex => qr/^ (?: not \s )? bool $/xi, handler => '_where_op_BOOL' }, ); #====================================================================== @@ -62,7 +72,7 @@ sub new { delete $opt{case} if $opt{case} && $opt{case} ne 'lower'; # default logic for interpreting arrayrefs - $opt{logic} = uc $opt{logic} || 'OR'; + $opt{logic} = $opt{logic} ? uc $opt{logic} : 'OR'; # how to return bind vars # LDNOTE: changed nwiger code : why this 'delete' ?? @@ -74,17 +84,21 @@ 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 + $opt{unary_ops} ||= []; + push @{$opt{unary_ops}}, @BUILTIN_UNARY_OPS; + return bless \%opt, $class; } @@ -95,33 +109,39 @@ sub new { #====================================================================== sub insert { - my $self = shift; - my $table = $self->_table(shift); - my $data = shift || return; + my $self = shift; + my $table = $self->_table(shift); + my $data = shift || return; + my $options = shift; my $method = $self->_METHOD_FOR_refkind("_insert", $data); - my ($sql, @bind) = $self->$method($data); + my ($sql, @bind) = $self->$method($data); $sql = join " ", $self->_sqlcase('insert into'), $table, $sql; + + 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 {$$fields}, + }); + return join (' ', $self->_sqlcase(' returning'), $f); +} + sub _insert_HASHREF { # explicit list of fields and then values my ($self, $data) = @_; my @fields = sort keys %$data; - my ($sql, @bind); - { # get values (need temporary override of bindtype to avoid an error) - local $self->{bindtype} = 'normal'; - ($sql, @bind) = $self->_insert_ARRAYREF([@{$data}{@fields}]); - } - - # if necessary, transform values according to 'bindtype' - if ($self->{bindtype} eq 'columns') { - for my $i (0 .. $#fields) { - ($bind[$i]) = $self->_bindtype($fields[$i], $bind[$i]); - } - } + my ($sql, @bind) = $self->_insert_values($data); # assemble SQL $_ = $self->_quote($_) foreach @fields; @@ -137,18 +157,48 @@ sub _insert_ARRAYREF { # just generate values(?,?) part (no list of fields) $self->{bindtype} ne 'columns' or belch "can't do 'columns' bindtype when called with arrayref"; + # fold the list of values into a hash of column name - value pairs + # (where the column names are artificially generated, and their + # lexicographical ordering keep the ordering of the original list) + my $i = "a"; # incremented values will be in lexicographical order + my $data_in_hash = { map { ($i++ => $_) } @$data }; + + return $self->_insert_values($data_in_hash); +} + +sub _insert_ARRAYREFREF { # literal SQL with bind + my ($self, $data) = @_; + + my ($sql, @bind) = @${$data}; + $self->_assert_bindval_matches_bindtype(@bind); + + return ($sql, @bind); +} + + +sub _insert_SCALARREF { # literal SQL without bind + my ($self, $data) = @_; + + return ($$data); +} + +sub _insert_values { + my ($self, $data) = @_; + my (@values, @all_bind); - for my $v (@$data) { + foreach my $column (sort keys %$data) { + my $v = $data->{$column}; $self->_SWITCH_refkind($v, { - ARRAYREF => sub { + ARRAYREF => sub { if ($self->{array_datatypes}) { # if array datatype are activated push @values, '?'; - push @all_bind, $v; + push @all_bind, $self->_bindtype($column, $v); } else { # else literal SQL with bind my ($sql, @bind) = @$v; + $self->_assert_bindval_matches_bindtype(@bind); push @values, $sql; push @all_bind, @bind; } @@ -156,16 +206,17 @@ sub _insert_ARRAYREF { # just generate values(?,?) part (no list of fields) ARRAYREFREF => sub { # literal SQL with bind my ($sql, @bind) = @${$v}; + $self->_assert_bindval_matches_bindtype(@bind); push @values, $sql; 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"; push @values, '?'; - push @all_bind, $v; + push @all_bind, $self->_bindtype($column, $v); }, SCALARREF => sub { # literal SQL without bind @@ -174,7 +225,7 @@ sub _insert_ARRAYREF { # just generate values(?,?) part (no list of fields) SCALAR_or_UNDEF => sub { push @values, '?'; - push @all_bind, $v; + push @all_bind, $self->_bindtype($column, $v); }, }); @@ -186,19 +237,6 @@ sub _insert_ARRAYREF { # just generate values(?,?) part (no list of fields) } -sub _insert_ARRAYREFREF { # literal SQL with bind - my ($self, $data) = @_; - return @${$data}; -} - - -sub _insert_SCALARREF { # literal SQL without bind - my ($self, $data) = @_; - - return ($$data); -} - - #====================================================================== # UPDATE methods @@ -222,21 +260,23 @@ sub update { my $label = $self->_quote($k); $self->_SWITCH_refkind($v, { - ARRAYREF => sub { + ARRAYREF => sub { if ($self->{array_datatypes}) { # array datatype push @set, "$label = ?"; push @all_bind, $self->_bindtype($k, $v); } else { # literal SQL with bind my ($sql, @bind) = @$v; + $self->_assert_bindval_matches_bindtype(@bind); push @set, "$label = $sql"; - push @all_bind, $self->_bindtype($k, @bind); + push @all_bind, @bind; } }, ARRAYREFREF => sub { # literal SQL with bind my ($sql, @bind) = @${$v}; + $self->_assert_bindval_matches_bindtype(@bind); push @set, "$label = $sql"; - push @all_bind, $self->_bindtype($k, @bind); + push @all_bind, @bind; }, SCALARREF => sub { # literal SQL without bind push @set, "$label = $$v"; @@ -280,11 +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) : $sql; } #====================================================================== @@ -301,7 +341,7 @@ sub delete { my($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; } @@ -324,7 +364,7 @@ sub where { $sql .= $self->_order_by($order); } - return wantarray ? ($sql, @bind) : $sql; + return wantarray ? ($sql, @bind) : $sql; } @@ -334,12 +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) : $sql; } @@ -357,19 +396,9 @@ sub _where_ARRAYREF { my @clauses = @$where; - # if the array starts with [-and|or => ...], recurse with that logic - my $first = $clauses[0] || ''; - if ($first =~ /^-(and|or)/i) { - $logic = $1; - shift @clauses; - return $self->_where_ARRAYREF(\@clauses, $logic); - } - - #otherwise.. 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, { @@ -377,11 +406,17 @@ sub _where_ARRAYREF { # skip empty elements, otherwise get invalid trailing AND stuff ARRAYREF => sub {$self->_recurse_where($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); }, @@ -401,7 +436,16 @@ sub _where_ARRAYREF { return $self->_join_sql_clauses($logic, \@sql_clauses, \@all_bind); } +#====================================================================== +# WHERE: top-level ARRAYREFREF +#====================================================================== +sub _where_ARRAYREFREF { + my ($self, $where) = @_; + my ($sql, @bind) = @$$where; + $self->_assert_bindval_matches_bindtype(@bind); + return ($sql, @bind); +} #====================================================================== # WHERE: top-level HASHREF @@ -411,16 +455,48 @@ sub _where_HASHREF { my ($self, $where) = @_; my (@sql_clauses, @all_bind); - # LDNOTE : don't really know why we need to sort keys - 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; @@ -429,61 +505,117 @@ 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, $op, $v) = @_; + my ($sql, @bind) = $self->_SWITCH_refkind ($rhs, { + SCALAR => sub { + puke "Illegal use of top-level '$op'" + unless $self->{_nested_func_lhs}; - $op =~ /^(AND|OR|NEST)[_\d]*/i - or puke "unknown operator: -$op"; - $op = uc($1); # uppercase, remove trailing digits - $self->_debug("OP(-$op) within hashref, recursing..."); + return ( + $self->_convert('?'), + $self->_bindtype($self->{_nested_func_lhs}, $rhs) + ); + }, + FALLBACK => sub { + $self->_recurse_where ($rhs) + }, + }); - $self->_SWITCH_refkind($v, { + $sql = sprintf ('%s %s', + $self->_sqlcase($op), + $sql, + ); + + return ($sql, @bind); +} +sub _where_op_ANDOR { + my ($self, $op, $v) = @_; + + $self->_SWITCH_refkind($v, { ARRAYREF => sub { - # LDNOTE : should deprecate {-or => [...]} and {-and => [...]} - # because they are misleading; the only proper way would be - # -nest => [-or => ...], -nest => [-and ...] - return $self->_where_ARRAYREF($v, $op eq 'NEST' ? '' : $op); + return $self->_where_ARRAYREF($v, $op); }, HASHREF => sub { - if ($op eq 'OR') { - belch "-or => {...} should be -nest => [...]"; - return $self->_where_ARRAYREF([%$v], 'OR'); - } - else { # NEST | AND - return $self->_where_HASHREF($v); - } + return ( $op =~ /^or/i ) + ? $self->_where_ARRAYREF( [ map { $_ => $v->{$_} } ( sort keys %$v ) ], $op ) + : $self->_where_HASHREF($v); }, - SCALARREF => sub { # literal SQL - $op eq 'NEST' - or puke "-$op => \\\$scalar not supported, use -nest => ..."; - return ($$v); + SCALARREF => sub { + puke "-$op => \\\$scalar not supported, use -nest => ..."; }, - ARRAYREFREF => sub { # literal SQL - $op eq 'NEST' - or puke "-$op => \\[..] not supported, use -nest => ..."; - return @{${$v}}; + ARRAYREFREF => sub { + puke "-$op => \\[..] not supported, use -nest => ..."; }, SCALAR => sub { # permissively interpreted as SQL - $op eq 'NEST' - or puke "-$op => 'scalar' not supported, use -nest => \\'scalar'"; + puke "-$op => 'scalar' not supported, use -nest => \\'scalar'"; + }, + + UNDEF => sub { + puke "-$op => undef not supported"; + }, + }); +} + +sub _where_op_NEST { + my ($self, $op, $v) = @_; + + $self->_SWITCH_refkind($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 ( $prefix, $suffix ) = ( $op =~ /\bnot\b/i ) + ? ( '(NOT ', ')' ) + : ( '', '' ); + + my ($sql, @bind) = do { + $self->_SWITCH_refkind($v, { + SCALAR => sub { # interpreted as SQL column + $self->_convert($self->_quote($v)); + }, + + UNDEF => sub { + puke "-$op => undef not supported"; + }, + + FALLBACK => sub { + $self->_recurse_where ($v); + }, + }); + }; + + return ( + join ('', $prefix, $sql, $suffix), + @bind, + ); +} + + sub _where_hashpair_ARRAYREF { my ($self, $k, $v) = @_; @@ -492,14 +624,22 @@ sub _where_hashpair_ARRAYREF { $self->_debug("ARRAY($k) means distribute over elements"); # put apart first element if it is an operator (-and, -or) - my $op = $v[0] =~ /^-/ ? shift @v : undef; - $self->_debug("OP($op) reinjected into the distributed array") if $op; - + my $op = ( + (defined $v[0] && $v[0] =~ /^ - (?: AND|OR ) $/ix) + ? shift @v + : '' + ); my @distributed = map { {$k => $_} } @v; - unshift @distributed, $op if $op; - return $self->_recurse_where(\@distributed); - } + if ($op) { + $self->_debug("OP($op) reinjected into the distributed array"); + unshift @distributed, $op; + } + + 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"); @@ -508,26 +648,43 @@ sub _where_hashpair_ARRAYREF { } sub _where_hashpair_HASHREF { - my ($self, $k, $v) = @_; + my ($self, $k, $v, $logic) = @_; + $logic ||= 'and'; - my (@all_sql, @all_bind); + local $self->{_nested_func_lhs} = $self->{_nested_func_lhs}; - for my $op (sort keys %$v) { - my $val = $v->{$op}; + my ($all_sql, @all_bind); + + for my $orig_op (sort keys %$v) { + my $val = $v->{$orig_op}; # put the operator in canonical form - $op =~ s/^-//; # remove initial dash - $op =~ tr/_/ /; # underscores become spaces - $op =~ s/^\s+//; # no initial space - $op =~ s/\s+$//; # no final space - $op =~ s/\s+/ /; # multiple spaces become one + 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 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 ($special_op) { - ($sql, @bind) = $special_op->{handler}->($self, $k, $op, $val); + 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"; + } + elsif (not ref $handler) { + ($sql, @bind) = $self->$handler ($k, $op, $val); + } + elsif (ref $handler eq 'CODE') { + ($sql, @bind) = $handler->($self, $k, $op, $val); + } + else { + puke "Illegal handler for special operator $orig_op - expecting a method name or a coderef"; + } } else { $self->_SWITCH_refkind($val, { @@ -536,14 +693,9 @@ sub _where_hashpair_HASHREF { ($sql, @bind) = $self->_where_field_op_ARRAYREF($k, $op, $val); }, - SCALARREF => sub { # CASE: col => {op => \$scalar} - $sql = join ' ', $self->_convert($self->_quote($k)), - $self->_sqlcase($op), - $$val; - }, - - ARRAYREFREF => sub { # CASE: col => {op => \[$sql, @bind]} + ARRAYREFREF => sub { # CASE: col => {op => \[$sql, @bind]} (literal SQL with bind) my ($sub_sql, @sub_bind) = @$$val; + $self->_assert_bindval_matches_bindtype(@sub_bind); $sql = join ' ', $self->_convert($self->_quote($k)), $self->_sqlcase($op), $sub_sql; @@ -553,24 +705,29 @@ sub _where_hashpair_HASHREF { UNDEF => sub { # CASE: col => {op => undef} : sql "IS (NOT)? NULL" my $is = ($op =~ $self->{equality_op}) ? 'is' : ($op =~ $self->{inequality_op}) ? 'is not' : - puke "unexpected operator '$op' with undef operand"; + puke "unexpected operator '$orig_op' with undef operand"; $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 + ); }, }); } - push @all_sql, $sql; + ($all_sql) = (defined $all_sql and $all_sql) ? $self->_join_sql_clauses($logic, [$all_sql, $sql], []) : $sql; push @all_bind, @bind; } - - return $self->_join_sql_clauses('and', \@all_sql, \@all_bind); + return ($all_sql, @all_bind); } @@ -578,24 +735,36 @@ sub _where_hashpair_HASHREF { sub _where_field_op_ARRAYREF { my ($self, $k, $op, $vals) = @_; - if(@$vals) { - $self->_debug("ARRAY($vals) means multiple elements: [ @$vals ]"); + my @vals = @$vals; #always work on a copy + + if(@vals) { + $self->_debug(sprintf '%s means multiple elements: [ %s ]', + $vals, + join (', ', map { defined $_ ? "'$_'" : 'NULL' } @vals ), + ); + # see if the first element is an -and/-or op + my $logic; + if (defined $vals[0] && $vals[0] =~ /^ - ( AND|OR ) $/ix) { + $logic = uc $1; + shift @vals; + } + # distribute $op over each remaining member of @vals, append logic if exists + return $self->_recurse_where([map { {$k => {$op, $_}} } @vals], $logic); - # LDNOTE : 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. - my $logic = ($op =~ $self->{inequality_op}) ? 'AND' : 'OR'; - - # distribute $op over each member of @$vals - return $self->_recurse_where([map { {$k => {$op, $_}} } @$vals], $logic); + # 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}; @@ -613,20 +782,22 @@ sub _where_hashpair_SCALARREF { return ($sql); } +# literal SQL with bind 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; - @bind = $self->_bindtype($k, @bind); return ($sql, @bind ); } +# literal SQL without bind sub _where_hashpair_SCALAR { my ($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 ( $sql, @bind); @@ -677,16 +848,55 @@ sub _where_UNDEF { sub _where_field_BETWEEN { my ($self, $k, $op, $vals) = @_; - ref $vals eq 'ARRAY' && @$vals == 2 - or puke "special op 'between' requires an arrayref of two values"; - - my ($label) = $self->_convert($self->_quote($k)); - my ($placeholder) = $self->_convert('?'); - my $and = $self->_sqlcase('and'); + my ($label, $and, $placeholder); + $label = $self->_convert($self->_quote($k)); + $and = ' ' . $self->_sqlcase('and') . ' '; + $placeholder = $self->_convert('?'); $op = $self->_sqlcase($op); - my $sql = "( $label $op $placeholder $and $placeholder )"; - my @bind = $self->_bindtype($k, @$vals); + my ($clause, @bind) = $self->_SWITCH_refkind($vals, { + ARRAYREFREF => sub { + my ($s, @b) = @$$vals; + $self->_assert_bindval_matches_bindtype(@b); + ($s, @b); + }, + SCALARREF => sub { + return $$vals; + }, + ARRAYREF => sub { + puke "special op 'between' accepts an arrayref with exactly two values" + if @$vals != 2; + + my (@all_sql, @all_bind); + foreach my $val (@$vals) { + my ($sql, @bind) = $self->_SWITCH_refkind($val, { + SCALAR => sub { + return ($placeholder, ($val)); + }, + SCALARREF => sub { + return ($self->_convert($$val), ()); + }, + ARRAYREFREF => sub { + my ($sql, @bind) = @$$val; + $self->_assert_bindval_matches_bindtype(@bind); + return ($self->_convert($sql), @bind); + }, + }); + push @all_sql, $sql; + push @all_bind, @bind; + } + + return ( + (join $and, @all_sql), + $self->_bindtype($k, @all_bind), + ); + }, + FALLBACK => sub { + puke "special op 'between' accepts an arrayref with two values, or a single literal scalarref/arrayref-ref"; + }, + }); + + my $sql = "( $label $op $clause )"; return ($sql, @bind) } @@ -697,30 +907,52 @@ sub _where_field_IN { # backwards compatibility : if scalar, force into an arrayref $vals = [$vals] if defined $vals && ! ref $vals; - ref $vals eq 'ARRAY' - or puke "special op 'in' requires an arrayref"; - my ($label) = $self->_convert($self->_quote($k)); my ($placeholder) = $self->_convert('?'); - my $and = $self->_sqlcase('and'); $op = $self->_sqlcase($op); - if (@$vals) { # nonempty list - my $placeholders = join ", ", (($placeholder) x @$vals); - my $sql = "$label $op ( $placeholders )"; - my @bind = $self->_bindtype($k, @$vals); + my ($sql, @bind) = $self->_SWITCH_refkind($vals, { + ARRAYREF => sub { # list of choices + if (@$vals) { # nonempty list + my $placeholders = join ", ", (($placeholder) x @$vals); + my $sql = "$label $op ( $placeholders )"; + my @bind = $self->_bindtype($k, @$vals); - return ($sql, @bind); - } - else { # empty list : some databases won't understand "IN ()", so DWIM - my $sql = ($op =~ /\bnot\b/i) ? $self->{sqltrue} : $self->{sqlfalse}; - return ($sql); - } -} + return ($sql, @bind); + } + else { # empty list : some databases won't understand "IN ()", so DWIM + my $sql = ($op =~ /\bnot\b/i) ? $self->{sqltrue} : $self->{sqlfalse}; + return ($sql); + } + }, + SCALARREF => sub { # literal SQL + my $sql = $self->_open_outer_paren ($$vals); + return ("$label $op ( $sql )"); + }, + ARRAYREFREF => sub { # literal SQL with bind + my ($sql, @bind) = @$$vals; + $self->_assert_bindval_matches_bindtype(@bind); + $sql = $self->_open_outer_paren ($sql); + return ("$label $op ( $sql )", @bind); + }, + FALLBACK => sub { + puke "special op 'in' requires an arrayref (or scalarref/arrayref-ref)"; + }, + }); + return ($sql, @bind); +} +# Some databases (SQLite) treat col IN (1, 2) different from +# col IN ( (1, 2) ). Use this to strip all outer parens while +# adding them back in the corresponding method +sub _open_outer_paren { + my ($self, $sql) = @_; + $sql = $1 while $sql =~ /^ \s* \( (.*) \) \s* $/xs; + return $sql; +} #====================================================================== @@ -730,47 +962,80 @@ sub _where_field_IN { sub _order_by { my ($self, $arg) = @_; - # construct list of ordering instructions - my @order = $self->_SWITCH_refkind($arg, { + my (@sql, @bind); + for my $c ($self->_order_by_chunks ($arg) ) { + $self->_SWITCH_refkind ($c, { + SCALAR => sub { push @sql, $c }, + ARRAYREF => sub { push @sql, shift @$c; push @bind, @$c }, + }); + } + + my $sql = @sql + ? sprintf ('%s %s', + $self->_sqlcase(' order by'), + join (', ', @sql) + ) + : '' + ; + + return wantarray ? ($sql, @bind) : $sql; +} + +sub _order_by_chunks { + my ($self, $arg) = @_; + + return $self->_SWITCH_refkind($arg, { ARRAYREF => sub { - map {$self->_SWITCH_refkind($_, { - SCALAR => sub {$self->_quote($_)}, - UNDEF => sub {}, - SCALARREF => sub {$$_}, # literal SQL, no quoting - HASHREF => sub {$self->_order_by_hash($_)} - }) } @$arg; + map { $self->_order_by_chunks ($_ ) } @$arg; + }, + + ARRAYREFREF => sub { + my ($s, @b) = @$$arg; + $self->_assert_bindval_matches_bindtype(@b); + [ $s, @b ]; }, SCALAR => sub {$self->_quote($arg)}, - UNDEF => sub {}, + + UNDEF => sub {return () }, + SCALARREF => sub {$$arg}, # literal SQL, no quoting - HASHREF => sub {$self->_order_by_hash($arg)}, - }); + HASHREF => sub { + # get first pair in hash + my ($key, $val, @rest) = %$arg; - # build SQL - my $order = join ', ', @order; - return $order ? $self->_sqlcase(' order by')." $order" : ''; -} + return () unless $key; + if ( @rest or not $key =~ /^-(desc|asc)/i ) { + puke "hash passed to _order_by must have exactly one key (-desc or -asc)"; + } -sub _order_by_hash { - my ($self, $hash) = @_; + my $direction = $1; - # get first pair in hash - my ($key, $val) = each %$hash; + my @ret; + for my $c ($self->_order_by_chunks ($val)) { + my ($sql, @bind); - # check if one pair was found and no other pair in hash - $key && !(each %$hash) - or puke "hash passed to _order_by must have exactly one key (-desc or -asc)"; + $self->_SWITCH_refkind ($c, { + SCALAR => sub { + $sql = $c; + }, + ARRAYREF => sub { + ($sql, @bind) = @$c; + }, + }); - my ($order) = ($key =~ /^-(desc|asc)/i) - or puke "invalid key in _order_by hash : $key"; + $sql = $sql . ' ' . $self->_sqlcase($direction); - return $self->_quote($val) ." ". $self->_sqlcase($order); -} + push @ret, [ $sql, @bind]; + } + return @ret; + }, + }); +} #====================================================================== @@ -793,69 +1058,81 @@ sub _table { # UTILITY FUNCTIONS #====================================================================== +# highly optimized, as it's called way too often sub _quote { - my $self = shift; - my $label = shift; - - $label or puke "can't quote an empty label"; - - # left and right quote characters - my ($ql, $qr, @other) = $self->_SWITCH_refkind($self->{quote_char}, { - SCALAR => sub {($self->{quote_char}, $self->{quote_char})}, - ARRAYREF => sub {@{$self->{quote_char}}}, - UNDEF => sub {()}, - }); - not @other - or puke "quote_char must be an arrayref of 2 values"; - - # no quoting if no quoting chars - $ql or return $label; + # my ($self, $label) = @_; - # no quoting for literal SQL - return $$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 $self = shift; + if ($self->{bindtype} eq 'columns') { + for (@_) { + if (!defined $_ || ref($_) ne 'ARRAY' || @$_ != 2) { + puke "bindtype 'columns' selected, you need to pass: [column_name => bind_value]" + } + } + } } sub _join_sql_clauses { @@ -877,11 +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]); } @@ -891,48 +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->(); } @@ -952,7 +1236,35 @@ sub values { my $data = shift || return; puke "Argument to ", __PACKAGE__, "->values must be a \\%hash" unless ref $data eq 'HASH'; - return map { $self->_bindtype($_, $data->{$_}) } sort keys %$data; + + my @all_bind; + foreach my $k ( sort keys %$data ) { + my $v = $data->{$k}; + $self->_SWITCH_refkind($v, { + ARRAYREF => sub { + if ($self->{array_datatypes}) { # array datatype + push @all_bind, $self->_bindtype($k, $v); + } + else { # literal SQL with bind + my ($sql, @bind) = @$v; + $self->_assert_bindval_matches_bindtype(@bind); + push @all_bind, @bind; + } + }, + ARRAYREFREF => sub { # literal SQL with bind + my ($sql, @bind) = @${$v}; + $self->_assert_bindval_matches_bindtype(@bind); + push @all_bind, @bind; + }, + SCALARREF => sub { # literal SQL without bind + }, + SCALAR_or_UNDEF => sub { + push @all_bind, $self->_bindtype($k, $v); + }, + }); + } + + return @all_bind; } sub generate { @@ -968,15 +1280,15 @@ sub generate { my $r = ref $v; my $label = $self->_quote($k); if ($r eq 'ARRAY') { - # SQL included for values - my @bind = @$v; - my $sql = shift @bind; + # literal SQL with bind + my ($sql, @bind) = @$v; + $self->_assert_bindval_matches_bindtype(@bind); push @sqlq, "$label = $sql"; - push @sqlv, $self->_bindtype($k, @bind); + push @sqlv, @bind; } elsif ($r eq 'SCALAR') { - # embedded literal SQL + # literal SQL without bind push @sqlq, "$label = $$v"; - } else { + } else { push @sqlq, "$label = ?"; push @sqlv, $self->_bindtype($k, $v); } @@ -986,14 +1298,15 @@ sub generate { # unlike insert(), assume these are ONLY the column names, i.e. for SQL for my $v (@$_) { my $r = ref $v; - if ($r eq 'ARRAY') { - my @val = @$v; - push @sqlq, shift @val; - push @sqlv, @val; - } elsif ($r eq 'SCALAR') { + if ($r eq 'ARRAY') { # literal SQL with bind + my ($sql, @bind) = @$v; + $self->_assert_bindval_matches_bindtype(@bind); + push @sqlq, $sql; + push @sqlv, @bind; + } elsif ($r eq 'SCALAR') { # literal SQL without bind # embedded literal SQL push @sqlq, $$v; - } else { + } else { push @sqlq, '?'; push @sqlv, $v; } @@ -1116,14 +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: @@ -1143,7 +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 @@ -1151,7 +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'); @@ -1198,7 +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. @@ -1243,20 +1556,23 @@ the huge section on L at the bottom. =item sqltrue, sqlfalse Expressions for inserting boolean values within SQL statements. -By default these are C<1=1> and C<1=0>. +By default these are C<1=1> and C<1=0>. They are used +by the special operators C<-in> and C<-not_in> for generating +correct SQL even when the argument is an empty array (see below). =item logic This determines the default logical operator for multiple WHERE -statements in arrays. By default it is "or", meaning that a WHERE +statements in arrays or hashes. If absent, the default logic is "or" +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: +will generate SQL like this: WHERE event_date >= '2/13/99' OR event_date <= '4/24/03' @@ -1270,10 +1586,10 @@ Which will change the above C to: WHERE event_date >= '2/13/99' AND event_date <= '4/24/03' The logic can also be changed locally by inserting -an extra first element in the array : +a modifier in front of an arrayref : - @where = (-and => event_date => {'>=', '2/13/99'}, - event_date => {'<=', '4/24/03'} ); + @where = (-and => [event_date => {'>=', '2/13/99'}, + event_date => {'<=', '4/24/03'} ]); See the L section for explanations. @@ -1343,10 +1659,14 @@ are or are not included. You could wrap that above C loop in a simple sub called C or something and reuse it repeatedly. You still get a layer of abstraction over manual SQL specification. +Note that if you set L to C, the C<\[$sql, @bind]> +construct (see L) +will expect the bind values in this format. + =item quote_char This is the character that a table or column name will be quoted -with. By default this is an empty string, but you could set it to +with. By default this is an empty string, but you could set it to the character C<`>, to generate SQL like this: SELECT `a_field` FROM `a_table` WHERE `some_field` LIKE '%someval%' @@ -1358,7 +1678,7 @@ that generates SQL like this: SELECT [a_field] FROM [a_table] WHERE [some_field] LIKE '%someval%' -Quoting is useful if you have tables or columns names that are reserved +Quoting is useful if you have tables or columns names that are reserved words in your database's SQL dialect. =item name_sep @@ -1371,8 +1691,8 @@ so that tables and column names can be individually quoted like this: =item array_datatypes -When this option is true, arrayrefs in INSERT or UPDATE are -interpreted as array datatypes and are passed directly +When this option is true, arrayrefs in INSERT or UPDATE are +interpreted as array datatypes and are passed directly to the DBI layer. When this option is false, arrayrefs are interpreted as literal SQL, just like refs to arrayrefs @@ -1383,15 +1703,21 @@ for literal SQL). =item special_ops -Takes a reference to a list of "special operators" +Takes a reference to a list of "special operators" to extend the syntax understood by L. See section L for details. +=item unary_ops + +Takes a reference to a list of "unary operators" +to extend the syntax understood by L. +See section L for details. + =back -=head2 insert($table, \@values || \%fieldvals) +=head2 insert($table, \@values || \%fieldvals, \%options) This is the simplest function. You simply give it a table name and either an arrayref of values or hashref of field/value pairs. @@ -1400,6 +1726,23 @@ See the sections on L and L for information on how to insert with those data types. +The optional C<\%options> hash reference may contain additional +options to generate the insert SQL. Currently supported options +are: + +=over 4 + +=item returning + +Takes either a scalar of raw SQL fields, or an array reference of +field names, and adds on an SQL C statement at the end. +This allows you to return data generated by the insert statement +(such as row IDs) without performing another C