X-Git-Url: http://git.shadowcat.co.uk/gitweb/gitweb.cgi?a=blobdiff_plain;f=lib%2FSQL%2FAbstract.pm;h=532d70aa65730b4350978a1dc60bd2a6b27d82d3;hb=bd6a65cac4599702fc4a717772934ff28421c6be;hp=3e6e5cfd0a75b8897badcd202f1568cbe28520dc;hpb=190bbeb7fe176719eb8cd67527d4800fa3faf6ab;p=dbsrgits%2FSQL-Abstract.git diff --git a/lib/SQL/Abstract.pm b/lib/SQL/Abstract.pm index 3e6e5cf..532d70a 100644 --- a/lib/SQL/Abstract.pm +++ b/lib/SQL/Abstract.pm @@ -15,17 +15,27 @@ use Scalar::Util qw/blessed/; # GLOBALS #====================================================================== -our $VERSION = '1.49_02'; -$VERSION = eval $VERSION; # numify for warning-free dev releases +our $VERSION = '1.58'; +# This would confuse some packagers +#$VERSION = eval $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' ?? @@ -85,6 +95,10 @@ sub new { $opt{special_ops} ||= []; push @{$opt{special_ops}}, @BUILTIN_SPECIAL_OPS; + # unary operators + $opt{unary_ops} ||= []; + push @{$opt{unary_ops}}, @BUILTIN_UNARY_OPS; + return bless \%opt, $class; } @@ -110,18 +124,7 @@ sub _insert_HASHREF { # explicit list of fields and then values my @fields = sort keys %$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 +140,48 @@ 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 ($$data); +} + +sub _insert_values { + my ($self, $data) = @_; + my (@values, @all_bind); - for my $v (@$data) { + foreach my $column (sort keys %$data) { + my $v = $data->{$column}; $self->_SWITCH_refkind($v, { ARRAYREF => sub { if ($self->{array_datatypes}) { # if array datatype are activated push @values, '?'; - push @all_bind, $v; + push @all_bind, $self->_bindtype($column, $v); } else { # else literal SQL with bind my ($sql, @bind) = @$v; + $self->_assert_bindval_matches_bindtype(@bind); push @values, $sql; push @all_bind, @bind; } @@ -156,6 +189,7 @@ sub _insert_ARRAYREF { # just generate values(?,?) part (no list of fields) ARRAYREFREF => sub { # literal SQL with bind my ($sql, @bind) = @${$v}; + $self->_assert_bindval_matches_bindtype(@bind); push @values, $sql; push @all_bind, @bind; }, @@ -165,7 +199,7 @@ sub _insert_ARRAYREF { # just generate values(?,?) part (no list of fields) #TODO in SQLA >= 2.0 it will die instead belch "HASH ref as bind value in insert is not supported"; push @values, '?'; - push @all_bind, $v; + push @all_bind, $self->_bindtype($column, $v); }, SCALARREF => sub { # literal SQL without bind @@ -174,7 +208,7 @@ sub _insert_ARRAYREF { # just generate values(?,?) part (no list of fields) SCALAR_or_UNDEF => sub { push @values, '?'; - push @all_bind, $v; + push @all_bind, $self->_bindtype($column, $v); }, }); @@ -186,19 +220,6 @@ sub _insert_ARRAYREF { # just generate values(?,?) part (no list of fields) } -sub _insert_ARRAYREFREF { # literal SQL with bind - my ($self, $data) = @_; - return @${$data}; -} - - -sub _insert_SCALARREF { # literal SQL without bind - my ($self, $data) = @_; - - return ($$data); -} - - #====================================================================== # UPDATE methods @@ -229,14 +250,16 @@ sub update { } 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"; @@ -357,17 +380,7 @@ 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) { @@ -377,6 +390,8 @@ 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}}}, + 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 @@ -401,7 +416,16 @@ 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}}; + + return ($sql, @bind); +} #====================================================================== # WHERE: top-level HASHREF @@ -411,12 +435,11 @@ 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) { 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) + 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); @@ -431,47 +454,90 @@ sub _where_HASHREF { sub _where_op_in_hash { - my ($self, $op, $v) = @_; + my ($self, $orig_op, $v) = @_; + + # 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 - $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..."); + 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"; + } +} + +sub _where_op_ANDOR { + my ($self, $op, $v) = @_; + + $self->_SWITCH_refkind($v, { + ARRAYREF => sub { + return $self->_where_ARRAYREF($v, $op); + }, + + HASHREF => sub { + return ( $op =~ /^or/i ) + ? $self->_where_ARRAYREF( [ map { $_ => $v->{$_} } ( sort keys %$v ) ], $op ) + : $self->_where_HASHREF($v); + }, + + SCALARREF => sub { + puke "-$op => \\\$scalar not supported, use -nest => ..."; + }, + + ARRAYREFREF => sub { + puke "-$op => \\[..] not supported, use -nest => ..."; + }, + + SCALAR => sub { # permissively interpreted as SQL + 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, { 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, ''); }, 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 $self->_where_HASHREF($v); }, SCALARREF => sub { # literal SQL - $op eq 'NEST' - or puke "-$op => \\\$scalar not supported, use -nest => ..."; return ($$v); }, ARRAYREFREF => sub { # literal SQL - $op eq 'NEST' - or puke "-$op => \\[..] not supported, use -nest => ..."; return @{${$v}}; }, SCALAR => sub { # permissively interpreted as SQL - $op eq 'NEST' - or puke "-$op => 'scalar' not supported, use -nest => \\'scalar'"; belch "literal SQL should be -nest => \\'scalar' " . "instead of -nest => 'scalar' "; return ($v); @@ -484,6 +550,43 @@ sub _where_op_in_hash { } +sub _where_op_BOOL { + my ($self, $op, $v) = @_; + + 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 . $suffix), @bind ); + }, + + SCALARREF => sub { # literal SQL + return ($prefix . $$v . $suffix); + }, + + SCALAR => sub { # interpreted as SQL column + return ($prefix . $self->_convert($self->_quote($v)) . $suffix); + }, + + UNDEF => sub { + puke "-$op => undef not supported"; + }, + }); +} + + sub _where_hashpair_ARRAYREF { my ($self, $k, $v) = @_; @@ -492,13 +595,21 @@ 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? @@ -508,26 +619,38 @@ sub _where_hashpair_ARRAYREF { } sub _where_hashpair_HASHREF { - my ($self, $k, $v) = @_; + my ($self, $k, $v, $logic) = @_; + $logic ||= 'and'; - my (@all_sql, @all_bind); + my ($all_sql, @all_bind); - for my $op (sort keys %$v) { - my $val = $v->{$op}; + 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: 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); + 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, { @@ -536,27 +659,32 @@ sub _where_hashpair_HASHREF { ($sql, @bind) = $self->_where_field_op_ARRAYREF($k, $op, $val); }, - SCALARREF => sub { # CASE: col => {op => \$scalar} + 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]} + 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; - @bind = $self->_bindtype($k, @sub_bind); + @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' : - 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), @@ -566,11 +694,10 @@ sub _where_hashpair_HASHREF { }); } - 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); } @@ -578,20 +705,32 @@ 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 + # 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 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 { @@ -613,15 +752,17 @@ 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}; + $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"); @@ -677,16 +818,39 @@ 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"; + (ref $vals eq 'ARRAY' && @$vals == 2) or + (ref $vals eq 'REF' && (@$$vals == 1 || @$$vals == 2 || @$$vals == 3)) + or puke "special op 'between' requires an arrayref of two values (or a scalarref or arrayrefref for literal SQL)"; - my ($label) = $self->_convert($self->_quote($k)); - my ($placeholder) = $self->_convert('?'); - my $and = $self->_sqlcase('and'); + my ($clause, @bind, $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); + if (ref $vals eq 'REF') { + ($clause, @bind) = @$$vals; + } + else { + my (@all_sql, @all_bind); + + foreach my $val (@$vals) { + my ($sql, @bind) = $self->_SWITCH_refkind($val, { + SCALAR => sub { + return ($placeholder, ($val)); + }, + SCALARREF => sub { + return ($self->_convert($$val), ()); + }, + }); + push @all_sql, $sql; + push @all_bind, @bind; + } + + $clause = (join $and, @all_sql); + @bind = $self->_bindtype($k, @all_bind); + } + my $sql = "( $label $op $clause )"; return ($sql, @bind) } @@ -718,6 +882,7 @@ sub _where_field_IN { ARRAYREFREF => sub { # literal SQL with bind my ($sql, @bind) = @$$vals; + $self->_assert_bindval_matches_bindtype(@bind); return ("$label $op ( $sql )", @bind); }, @@ -732,8 +897,6 @@ sub _where_field_IN { - - #====================================================================== # ORDER BY #====================================================================== @@ -741,47 +904,76 @@ sub _where_field_IN { sub _order_by { my ($self, $arg) = @_; - # construct list of ordering instructions - my @order = $self->_SWITCH_refkind($arg, { + my (@sql, @bind); + for my $c ($self->_order_by_chunks ($arg) ) { + $self->_SWITCH_refkind ($c, { + SCALAR => sub { push @sql, $c }, + ARRAYREF => sub { push @sql, shift @$c; push @bind, @$c }, + }); + } + + my $sql = @sql + ? sprintf ('%s %s', + $self->_sqlcase(' order by'), + join (', ', @sql) + ) + : '' + ; + + return wantarray ? ($sql, @bind) : $sql; +} + +sub _order_by_chunks { + my ($self, $arg) = @_; + + return $self->_SWITCH_refkind($arg, { ARRAYREF => sub { - map {$self->_SWITCH_refkind($_, { - SCALAR => sub {$self->_quote($_)}, - UNDEF => sub {}, - SCALARREF => sub {$$_}, # literal SQL, no quoting - HASHREF => sub {$self->_order_by_hash($_)} - }) } @$arg; + map { $self->_order_by_chunks ($_ ) } @$arg; }, + ARRAYREFREF => sub { [ @$$arg ] }, + SCALAR => sub {$self->_quote($arg)}, - UNDEF => sub {}, + + UNDEF => sub {return () }, + SCALARREF => sub {$$arg}, # literal SQL, no quoting - HASHREF => sub {$self->_order_by_hash($arg)}, - }); + HASHREF => sub { + # get first pair in hash + my ($key, $val) = each %$arg; - # build SQL - my $order = join ', ', @order; - return $order ? $self->_sqlcase(' order by')." $order" : ''; -} + return () unless $key; + if ( (keys %$arg) > 1 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; + }, + }); +} #====================================================================== @@ -869,6 +1061,20 @@ sub _bindtype (@) { return $self->{bindtype} eq 'columns' ? map {[$col, $_]} @vals : @vals; } +# Dies if any element of @bind is not in [colname => value] format +# if bindtype is 'columns'. +sub _assert_bindval_matches_bindtype { + my ($self, @bind) = @_; + + if ($self->{bindtype} eq 'columns') { + foreach my $val (@bind) { + if (!defined $val || ref($val) ne 'ARRAY' || @$val != 2) { + die "bindtype 'columns' selected, you need to pass: [column_name => bind_value]" + } + } + } +} + sub _join_sql_clauses { my ($self, $logic, $clauses_aref, $bind_aref) = @_; @@ -963,7 +1169,35 @@ 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 { @@ -979,13 +1213,13 @@ 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 = $$v"; } else { push @sqlq, "$label = ?"; @@ -997,11 +1231,12 @@ sub generate { # unlike insert(), assume these are ONLY the column names, i.e. for SQL for my $v (@$_) { my $r = ref $v; - if ($r eq 'ARRAY') { - my @val = @$v; - push @sqlq, shift @val; - push @sqlv, @val; - } elsif ($r eq 'SCALAR') { + if ($r eq 'ARRAY') { # literal SQL with bind + my ($sql, @bind) = @$v; + $self->_assert_bindval_matches_bindtype(@bind); + push @sqlq, $sql; + push @sqlv, @bind; + } elsif ($r eq 'SCALAR') { # literal SQL without bind # embedded literal SQL push @sqlq, $$v; } else { @@ -1254,12 +1489,15 @@ 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 = ( @@ -1267,7 +1505,7 @@ array of the form: 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' @@ -1281,10 +1519,10 @@ 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. @@ -1354,6 +1592,10 @@ 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 @@ -1398,6 +1640,12 @@ 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 @@ -1565,10 +1813,10 @@ This simple code will create the following: $stmt = "WHERE user = ? AND ( status = ? OR status = ? OR status = ? )"; @bind = ('nwiger', 'assigned', 'in-progress', 'pending'); -An empty arrayref will be considered a logical false and -will generate 0=1. +A field associated to an empty arrayref will be considered a +logical false and will generate 0=1. -=head2 Key-value pairs +=head2 Specific comparison operators If you want to specify a different type of operator for your comparison, you can use a hashref for a given column: @@ -1585,19 +1833,9 @@ Which would generate: To test against multiple values, just enclose the values in an arrayref: - status => { '!=', ['assigned', 'in-progress', 'pending'] }; - -Which would give you: - - "WHERE status != ? AND status != ? AND status != ?" - -Notice that since the operator was recognized as being a 'negative' -operator, the arrayref was interpreted with 'AND' logic (because -of Morgan's laws). By contrast, the reverse - status => { '=', ['assigned', 'in-progress', 'pending'] }; -would generate : +Which would give you: "WHERE status = ? OR status = ? OR status = ?" @@ -1685,29 +1923,6 @@ Here is a quick list of equivalencies, since there is some overlap: status => [ -or => {'=', 'assigned'}, {'=', 'in-progress'}] status => [ {'=', 'assigned'}, {'=', 'in-progress'} ] -In addition to C<-and> and C<-or>, there is also a special C<-nest> -operator which adds an additional set of parens, to create a subquery. -For example, to get something like this: - - $stmt = "WHERE user = ? AND ( workhrs > ? OR geo = ? )"; - @bind = ('nwiger', '20', 'ASIA'); - -You would do: - - my %where = ( - user => 'nwiger', - -nest => [ workhrs => {'>', 20}, geo => 'ASIA' ], - ); - -If you need several nested subexpressions, you can number -the C<-nest> branches : - - my %where = ( - user => 'nwiger', - -nest1 => ..., - -nest2 => ..., - ... - ); =head2 Special operators : IN, BETWEEN, etc. @@ -1728,6 +1943,12 @@ Which would generate: 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 +(by default : C<1=0>). Similarly, C<< -not_in => [] >> generates +'sqltrue' (by default : C<1=1>). + + + Another pair of operators is C<-between> and C<-not_between>, used with an arrayref of two values: @@ -1745,7 +1966,40 @@ Would give you: These are the two builtin "special operators"; but the list can be expanded : see section L below. -=head2 Nested conditions +=head2 Unary operators: bool + +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:- + + my %where = ( + -bool => 'is_user', + -not_bool => 'is_enabled', + ); + +Would give you: + + WHERE is_user AND NOT is_enabled + +If a more complex combination is required, testing more conditions, +then you should use the and/or operators:- + + my %where = ( + -and => [ + -bool => 'one', + -bool => 'two', + -bool => 'three', + -not_bool => 'four', + ], + ); + +Would give you: + + WHERE one AND two AND three AND NOT four + + +=head2 Nested conditions, -and/-or prefixes So far, we've seen how multiple conditions are joined with a top-level C. We can change this by putting the different conditions we want in @@ -1768,15 +2022,32 @@ This data structure would create the following: OR ( user = ? AND status = ? ) )"; @bind = ('nwiger', 'pending', 'dispatched', 'robot', 'unassigned'); -This can be combined with the C<-nest> operator to properly group -SQL statements: + +There is also a special C<-nest> +operator which adds an additional set of parens, to create a subquery. +For example, to get something like this: + + $stmt = "WHERE user = ? AND ( workhrs > ? OR geo = ? )"; + @bind = ('nwiger', '20', 'ASIA'); + +You would do: + + my %where = ( + user => 'nwiger', + -nest => [ workhrs => {'>', 20}, geo => 'ASIA' ], + ); + + +Finally, clauses in hashrefs or arrayrefs can be +prefixed with an C<-and> or C<-or> to change the logic +inside : my @where = ( -and => [ user => 'nwiger', -nest => [ - ["-and", workhrs => {'>', 20}, geo => 'ASIA' ], - ["-and", workhrs => {'<', 50}, geo => 'EURO' ] + -and => [workhrs => {'>', 20}, geo => 'ASIA' ], + -and => [workhrs => {'<', 50}, geo => 'EURO' ] ], ], ); @@ -1787,6 +2058,37 @@ That would yield: ( ( workhrs > ? AND geo = ? ) OR ( workhrs < ? AND geo = ? ) ) ) + +=head2 Algebraic inconsistency, for historical reasons + +C: when connecting several conditions, the C<-and->|C<-or> +operator goes C of the nested structure; whereas when connecting +several constraints on one column, the C<-and> operator goes +C the arrayref. Here is an example combining both features : + + my @where = ( + -and => [a => 1, b => 2], + -or => [c => 3, d => 4], + e => [-and => {-like => 'foo%'}, {-like => '%bar'} ] + ) + +yielding + + 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'}]} + # yields : WHERE ( ( col LIKE ? AND col LIKE ? ) ) + + [-and => {col => {-like => 'foo%'}, {col => {-like => '%bar'}}]] + # yields : WHERE ( ( col LIKE ? OR col LIKE ? ) ) + + =head2 Literal SQL Finally, sometimes only literal SQL will do. If you want to include @@ -1817,8 +2119,10 @@ with this: TMTOWTDI. -Conditions on boolean columns can be expressed in the -same way, passing a reference to an empty string : +Conditions on boolean columns can be expressed in the same way, passing +a reference to an empty string, however using liternal SQL in this way +is deprecated - the preferred method is to use the boolean operators - +see L : my %where = ( priority => { '<', 2 }, @@ -1847,6 +2151,17 @@ This would create: $stmt = "WHERE ( date_column = date '2008-09-30' - ?::integer )" @bind = ('10'); +Note that you must pass the bind values in the same format as they are returned +by L. That means that if you set L to C, you must +provide the bind values in the C<< [ column_meta => value ] >> format, where +C is an opaque scalar value; most commonly the column name, but +you can use any scalar value (including references and blessed references), +L will simply pass it through intact. So if C is set +to C the above example will look like: + + my %where = ( + date_column => \[q/= date '2008-09-30' - ?::integer/, [ dummy => 10 ]/] + ) Literal SQL is especially useful for nesting parenthesized clauses in the main SQL query. Here is a first example : @@ -1953,30 +2268,45 @@ 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: - Given | Will Generate + Given | Will Generate ---------------------------------------------------------- - \'colA DESC' | ORDER BY colA DESC - 'colA' | ORDER BY colA - [qw/colA colB/] | ORDER BY colA, colB - {-asc => 'colA'} | ORDER BY colA ASC - {-desc => 'colB'} | ORDER BY colB DESC - [ | - {-asc => 'colA'}, | ORDER BY colA ASC, colB DESC - {-desc => 'colB'} | - ] | - [colA => {-asc => 'colB'}] | ORDER BY colA, colB ASC - ========================================================== + | + \'colA DESC' | ORDER BY colA DESC + | + 'colA' | ORDER BY colA + | + [qw/colA colB/] | ORDER BY colA, colB + | + {-asc => 'colA'} | ORDER BY colA ASC + | + {-desc => 'colB'} | ORDER BY colB DESC + | + ['colA', {-asc => 'colB'}] | ORDER BY colA, colB ASC + | + { -asc => [qw/colA colB] } | ORDER BY colA ASC, colB ASC + | + [ | + { -asc => 'colA' }, | ORDER BY colA ASC, colB DESC, + { -desc => [qw/colB/], | colC ASC, colD ASC + { -asc => [qw/colC colD/],| + ] | + =========================================================== =head1 SPECIAL OPERATORS my $sqlmaker = SQL::Abstract->new(special_ops => [ - {regex => qr/.../, + { + regex => qr/.../, handler => sub { my ($self, $field, $op, $arg) = @_; ... - }, + }, + }, + { + regex => qr/.../, + handler => 'method_name', }, ]); @@ -1989,12 +2319,13 @@ For example : WHERE MATCH(field) AGAINST (?, ?) Special operators IN and BETWEEN are fairly standard and therefore -are builtin within C. For other operators, -like the MATCH .. AGAINST example above which is -specific to MySQL, you can write your own operator handlers : -supply a C argument to the C method. -That argument takes an arrayref of operator definitions; -each operator definition is a hashref with two entries +are builtin within C (as the overridable methods +C<_where_field_IN> and C<_where_field_BETWEEN>). For other operators, +like the MATCH .. AGAINST example above which is specific to MySQL, +you can write your own operator handlers - supply a C +argument to the C method. That argument takes an arrayref of +operator definitions; each operator definition is a hashref with two +entries: =over @@ -2004,10 +2335,24 @@ the regular expression to match the operator =item handler -coderef that will be called when meeting that operator -in the input tree. The coderef will be called with -arguments C<< ($self, $field, $op, $arg) >>, and -should return a C<< ($sql, @bind) >> structure. +Either a coderef or a plain scalar method name. In both cases +the expected return is C<< ($sql, @bind) >>. + +When supplied with a method name, it is simply called on the +L object as: + + $self->$method_name ($field, $op, $arg) + + Where: + + $op is the part that matched the handler regex + $field is the LHS of the operator + $arg is the RHS + +When supplied with a coderef, it is called as: + + $coderef->($self, $field, $op, $arg) + =back @@ -2034,6 +2379,59 @@ of the MATCH .. AGAINST syntax for MySQL ]); +=head1 UNARY OPERATORS + + my $sqlmaker = SQL::Abstract->new(unary_ops => [ + { + regex => qr/.../, + handler => sub { + my ($self, $op, $arg) = @_; + ... + }, + }, + { + regex => qr/.../, + handler => 'method_name', + }, + ]); + +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 +argument to the C method. That argument takes an arrayref of +operator definitions; each operator definition is a hashref with two +entries: + +=over + +=item regex + +the regular expression to match the operator + +=item handler + +Either a coderef or a plain scalar method name. In both cases +the expected return is C<< $sql >>. + +When supplied with a method name, it is simply called on the +L object as: + + $self->$method_name ($op, $arg) + + Where: + + $op is the part that matched the handler regex + $arg is the RHS or argument of the operator + +When supplied with a coderef, it is called as: + + $coderef->($self, $op, $arg) + + +=back + + =head1 PERFORMANCE Thanks to some benchmarking by Mark Stosberg, it turns out that @@ -2119,10 +2517,6 @@ support for the { operator => \["...", @bind] } construct (to embed literal SQL =item * -added -nest1, -nest2 or -nest_1, -nest_2, ... - -=item * - optional support for L =item * @@ -2132,21 +2526,12 @@ defensive programming : check arguments =item * fixed bug with global logic, which was previously implemented -through global variables yielding side-effects. Prior versons would +through global variables yielding side-effects. Prior versions would interpret C<< [ {cond1, cond2}, [cond3, cond4] ] >> as C<< "(cond1 AND cond2) OR (cond3 AND cond4)" >>. Now this is interpreted as C<< "(cond1 AND cond2) OR (cond3 OR cond4)" >>. -=item * - -C<-and> / C<-or> operators are no longer accepted -in the middle of an arrayref : they are -only admitted if in first position. - -=item * - -changed logic for distributing an op over arrayrefs =item * @@ -2182,6 +2567,7 @@ so I have no idea who they are! But the people I do know are: Guillermo Roditi (patch to cleanup "IN" and "BETWEEN", fix and tests for _order_by) Laurent Dami (internal refactoring, multiple -nest, extensible list of special operators, literal SQL) Norbert Buchmuller (support for literal SQL in hashpair, misc. fixes & tests) + Peter Rabbitson (rewrite of SQLA::Test, misc. fixes & tests) Thanks! @@ -2200,6 +2586,8 @@ While not an official support venue, C makes heavy use of C, and as such list members there are very familiar with how to create queries. +=head1 LICENSE + This module is free software; you may copy this under the terms of the GNU General Public License, or the Artistic License, copies of which should have accompanied your Perl kit.