X-Git-Url: http://git.shadowcat.co.uk/gitweb/gitweb.cgi?a=blobdiff_plain;f=lib%2FSQL%2FAbstract.pm;h=0e29467d00c1b9ce26438484a6e3f38fc50c6749;hb=d7d3d1584c95a2bc501e9b783678fae97ac0bb57;hp=6c7fc60a4e225dd6a412668da5c20270ebe9f297;hpb=c520207b444ed47a2a9fc1e68a804d42e17bef48;p=dbsrgits%2FSQL-Abstract.git diff --git a/lib/SQL/Abstract.pm b/lib/SQL/Abstract.pm index 6c7fc60..0e29467 100644 --- a/lib/SQL/Abstract.pm +++ b/lib/SQL/Abstract.pm @@ -15,7 +15,7 @@ use Scalar::Util (); # GLOBALS #====================================================================== -our $VERSION = '1.64_01'; +our $VERSION = '1.71'; # This would confuse some packagers $VERSION = eval $VERSION if $VERSION =~ /_/; # numify for warning-free dev releases @@ -25,17 +25,17 @@ 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 \s )? between $/ix, handler => '_where_field_BETWEEN'}, + {regex => qr/^ (?: not \s )? in $/ix, 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' }, + { 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 $/xi, handler => '_where_op_BOOL' }, ); #====================================================================== @@ -280,7 +280,19 @@ sub update { }, SCALARREF => sub { # literal SQL without bind push @set, "$label = $$v"; - }, + }, + HASHREF => sub { + my ($op, $arg, @rest) = %$v; + + puke 'Operator calls in update must be in the form { -op => $arg }' + if (@rest or not $op =~ /^\-(.+)/); + + local $self->{_nested_func_lhs} = $k; + my ($sql, @bind) = $self->_where_unary_op ($1, $arg); + + push @set, "$label = $sql"; + push @all_bind, @bind; + }, SCALAR_or_UNDEF => sub { push @set, "$label = ?"; push @all_bind, $self->_bindtype($k, $v); @@ -406,7 +418,11 @@ sub _where_ARRAYREF { # skip empty elements, otherwise get invalid trailing AND stuff ARRAYREF => sub {$self->_recurse_where($el) if @$el}, - ARRAYREFREF => sub { @{${$el}} if @{${$el}}}, + ARRAYREFREF => sub { + my ($s, @b) = @$$el; + $self->_assert_bindval_matches_bindtype(@b); + ($s, @b); + }, HASHREF => sub {$self->_recurse_where($el, 'and') if %$el}, # LDNOTE : previous SQLA code for hashrefs was creating a dirty @@ -438,8 +454,8 @@ sub _where_ARRAYREF { sub _where_ARRAYREFREF { my ($self, $where) = @_; - my ($sql, @bind) = @{${$where}}; - + my ($sql, @bind) = @$$where; + $self->_assert_bindval_matches_bindtype(@bind); return ($sql, @bind); } @@ -459,34 +475,24 @@ sub _where_HASHREF { 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 = substr $op, 1; # remove initial dash $op =~ s/^\s+|\s+$//g;# remove leading/trailing space + $op =~ s/\s+/ /g; # compress whitespace - $self->_debug("Unary OP(-$op) within hashref, recursing..."); + # so that -not_foo works correctly + $op =~ s/^not_/NOT /i; - 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 $self->{_nested_func_lhs} eq $k; # top level vs nested - ($sql, @bind); - } + $self->_debug("Unary OP(-$op) within hashref, recursing..."); + my ($s, @b) = $self->_where_unary_op ($op, $v); + + # top level vs nested + # we assume that handled unary ops will take care of their ()s + $s = "($s)" unless ( + List::Util::first {$op =~ $_->{regex}} @{$self->{unary_ops}} + or + defined($self->{_nested_func_lhs}) && ($self->{_nested_func_lhs} eq $k) + ); + ($s, @b); } else { my $method = $self->_METHOD_FOR_refkind("_where_hashpair", $v); @@ -501,9 +507,29 @@ sub _where_HASHREF { return $self->_join_sql_clauses('and', \@sql_clauses, \@all_bind); } -sub _where_func_generic { +sub _where_unary_op { my ($self, $op, $rhs) = @_; + if (my $op_entry = List::Util::first {$op =~ $_->{regex}} @{$self->{unary_ops}}) { + my $handler = $op_entry->{handler}; + + if (not ref $handler) { + if ($op =~ s/ [_\s]? \d+ $//x ) { + 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, $rhs); + } + elsif (ref $handler eq 'CODE') { + return $handler->($self, $op, $rhs); + } + else { + puke "Illegal handler for operator $op - expecting a method name or a coderef"; + } + } + + $self->debug("Generic unary OP: $op - recursing as function"); + my ($sql, @bind) = $self->_SWITCH_refkind ($rhs, { SCALAR => sub { puke "Illegal use of top-level '$op'" @@ -585,30 +611,22 @@ sub _where_op_NEST { 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)); - }, + my ($s, @b) = $self->_SWITCH_refkind($v, { + SCALAR => sub { # interpreted as SQL column + $self->_convert($self->_quote($v)); + }, - UNDEF => sub { - puke "-$op => undef not supported"; - }, + UNDEF => sub { + puke "-$op => undef not supported"; + }, - FALLBACK => sub { - $self->_recurse_where ($v); - }, - }); - }; + FALLBACK => sub { + $self->_recurse_where ($v); + }, + }); - return ( - join ('', $prefix, $sql, $suffix), - @bind, - ); + $s = "(NOT $s)" if $op =~ /^not/i; + ($s, @b); } @@ -656,9 +674,14 @@ sub _where_hashpair_HASHREF { # 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 + + # FIXME - we need to phase out dash-less ops + $op =~ s/^-//; # remove possible initial dash $op =~ s/^\s+|\s+$//g;# remove leading/trailing space + $op =~ s/\s+/ /g; # compress whitespace + + # so that -not_foo works correctly + $op =~ s/^not_/NOT /i; my ($sql, @bind); @@ -710,7 +733,7 @@ sub _where_hashpair_HASHREF { # retain for proper column type bind $self->{_nested_func_lhs} ||= $k; - ($sql, @bind) = $self->_where_func_generic ($op, $val); + ($sql, @bind) = $self->_where_unary_op ($op, $val); $sql = join (' ', $self->_convert($self->_quote($k)), @@ -782,7 +805,7 @@ 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 ); @@ -852,7 +875,9 @@ 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 $$vals; @@ -865,15 +890,23 @@ sub _where_field_BETWEEN { foreach my $val (@$vals) { my ($sql, @bind) = $self->_SWITCH_refkind($val, { SCALAR => sub { - return ($placeholder, ($val)); + return ($placeholder, $val); }, SCALARREF => sub { - return ($self->_convert($$val), ()); + return $$val; }, ARRAYREFREF => sub { my ($sql, @bind) = @$$val; - return ($self->_convert($sql), @bind); + $self->_assert_bindval_matches_bindtype(@bind); + return ($sql, @bind); }, + HASHREF => sub { + my ($func, $arg, @rest) = %$val; + puke ("Only simple { -func => arg } functions accepted as sub-arguments to BETWEEN") + if (@rest or $func !~ /^ \- (.+)/x); + local $self->{_nested_func_lhs} = $k; + $self->_where_unary_op ($1 => $arg); + } }); push @all_sql, $sql; push @all_bind, @bind; @@ -907,11 +940,41 @@ sub _where_field_IN { my ($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); + my (@all_sql, @all_bind); + + for my $val (@$vals) { + my ($sql, @bind) = $self->_SWITCH_refkind($val, { + SCALAR => sub { + return ($placeholder, $val); + }, + SCALARREF => sub { + return $$val; + }, + ARRAYREFREF => sub { + my ($sql, @bind) = @$$val; + $self->_assert_bindval_matches_bindtype(@bind); + return ($sql, @bind); + }, + HASHREF => sub { + my ($func, $arg, @rest) = %$val; + puke ("Only simple { -func => arg } functions accepted as sub-arguments to IN") + if (@rest or $func !~ /^ \- (.+)/x); + local $self->{_nested_func_lhs} = $k; + $self->_where_unary_op ($1 => $arg); + } + }); + push @all_sql, $sql; + push @all_bind, @bind; + } - return ($sql, @bind); + return ( + sprintf ('%s %s ( %s )', + $label, + $op, + join (', ', @all_sql) + ), + $self->_bindtype($k, @all_bind), + ); } else { # empty list : some databases won't understand "IN ()", so DWIM my $sql = ($op =~ /\bnot\b/i) ? $self->{sqltrue} : $self->{sqlfalse}; @@ -983,7 +1046,11 @@ 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($arg)}, @@ -993,11 +1060,11 @@ sub _order_by_chunks { HASHREF => sub { # get first pair in hash - my ($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)"; } @@ -1047,42 +1114,38 @@ sub _table { # UTILITY FUNCTIONS #====================================================================== +# highly optimized, as it's called way too often sub _quote { - my $self = shift; - my $label = shift; + # my ($self, $label) = @_; - $label or puke "can't quote an empty label"; + return '' unless defined $_[1]; + return ${$_[1]} if ref($_[1]) eq 'SCALAR'; - # left and right quote characters - my ($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; - - # no quoting for literal SQL - return $$label if ref($label) eq 'SCALAR'; + return $_[1] unless $_[0]->{quote_char}; - # separate table / column (if applicable) - my $sep = $self->{name_sep} || ''; - my @to_quote = $sep ? split /\Q$sep\E/, $label : ($label); - - # 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, @@ -1093,34 +1156,36 @@ sub _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 # 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]" } } } @@ -1145,11 +1210,9 @@ 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]); } @@ -1159,38 +1222,37 @@ sub _sqlcase { sub _refkind { my ($self, $data) = @_; - my $suffix = ''; - my $ref; - my $n_steps = 0; - while (1) { - # blessed objects are treated like scalars - $ref = (Scalar::Util::blessed $data) ? '' : ref $data; - $n_steps += 1 if $ref; - last if $ref ne 'REF'; - $data = $$data; - } + return 'UNDEF' unless defined $data; - my $base = $ref || (defined $data ? 'SCALAR' : 'UNDEF'); + # blessed objects are treated like scalars + my $ref = (Scalar::Util::blessed $data) ? '' : ref $data; - return $base . ('REF' x $n_steps); -} + 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 $method; - for ($self->_try_refkind($data)) { + for (@{$self->_try_refkind($data)}) { $method = $self->can($meth_prefix."_".$_) and last; } @@ -1203,7 +1265,7 @@ sub _SWITCH_refkind { my ($self, $data, $dispatch_table) = @_; my $coderef; - for ($self->_try_refkind($data)) { + for (@{$self->_try_refkind($data)}) { $coderef = $dispatch_table->{$_} and last; } @@ -1894,6 +1956,20 @@ This simple code will create the following: 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, @@ -2092,7 +2168,7 @@ 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', @@ -2255,6 +2331,17 @@ 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) @@ -2578,6 +2665,12 @@ 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 @@ -2608,6 +2701,15 @@ 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. +=head1 REPO + +=over + +=item * gitweb: L + +=item * git: L + +=back =head1 CHANGES