3 DBIx::Class::Manual::Cookbook - Miscellaneous recipes
9 When you expect a large number of results, you can ask L<DBIx::Class> for a
10 paged resultset, which will fetch only a defined number of records at a time:
12 my $rs = $schema->resultset('Artist')->search(
15 page => 1, # page to return (defaults to 1)
16 rows => 10, # number of results per page
20 return $rs->all(); # all records for page 1
22 You can get a L<Data::Page> object for the resultset (suitable for use
23 in e.g. a template) using the C<pager> method:
27 =head2 Complex WHERE clauses
29 Sometimes you need to formulate a query using specific operators:
31 my @albums = $schema->resultset('Album')->search({
32 artist => { 'like', '%Lamb%' },
33 title => { 'like', '%Fear of Fours%' },
36 This results in something like the following C<WHERE> clause:
38 WHERE artist LIKE '%Lamb%' AND title LIKE '%Fear of Fours%'
40 Other queries might require slightly more complex logic:
42 my @albums = $schema->resultset('Album')->search({
45 artist => { 'like', '%Smashing Pumpkins%' },
46 title => 'Siamese Dream',
48 artist => 'Starchildren',
52 This results in the following C<WHERE> clause:
54 WHERE ( artist LIKE '%Smashing Pumpkins%' AND title = 'Siamese Dream' )
55 OR artist = 'Starchildren'
57 For more information on generating complex queries, see
58 L<SQL::Abstract/WHERE CLAUSES>.
60 =head2 Retrieve one and only one row from a resultset
62 Sometimes you need only the first "top" row of a resultset. While this can be
63 easily done with L<< $rs->first|DBIx::Class::ResultSet/first >>, it is suboptimal,
64 as a full blown cursor for the resultset will be created and then immediately
65 destroyed after fetching the first row object.
66 L<< $rs->single|DBIx::Class::ResultSet/single >> is
67 designed specifically for this case - it will grab the first returned result
68 without even instantiating a cursor.
70 Before replacing all your calls to C<first()> with C<single()> please observe the
76 While single() takes a search condition just like search() does, it does
77 _not_ accept search attributes. However one can always chain a single() to
80 my $top_cd = $cd_rs -> search({}, { order_by => 'rating' }) -> single;
84 Since single() is the engine behind find(), it is designed to fetch a
85 single row per database query. Thus a warning will be issued when the
86 underlying SELECT returns more than one row. Sometimes however this usage
87 is valid: i.e. we have an arbitrary number of cd's but only one of them is
88 at the top of the charts at any given time. If you know what you are doing,
89 you can silence the warning by explicitly limiting the resultset size:
91 my $top_cd = $cd_rs -> search ({}, { order_by => 'rating', rows => 1 }) -> single;
95 =head2 Arbitrary SQL through a custom ResultSource
97 Sometimes you have to run arbitrary SQL because your query is too complex
98 (e.g. it contains Unions, Sub-Selects, Stored Procedures, etc.) or has to
99 be optimized for your database in a special way, but you still want to
100 get the results as a L<DBIx::Class::ResultSet>.
101 The recommended way to accomplish this is by defining a separate ResultSource
102 for your query. You can then inject complete SQL statements using a scalar
103 reference (this is a feature of L<SQL::Abstract>).
105 Say you want to run a complex custom query on your user data, here's what
106 you have to add to your User class:
108 package My::Schema::User;
110 use base qw/DBIx::Class/;
112 # ->load_components, ->table, ->add_columns, etc.
114 # Make a new ResultSource based on the User class
115 my $source = __PACKAGE__->result_source_instance();
116 my $new_source = $source->new( $source );
117 $new_source->source_name( 'UserFriendsComplex' );
119 # Hand in your query as a scalar reference
120 # It will be added as a sub-select after FROM,
121 # so pay attention to the surrounding brackets!
122 $new_source->name( \<<SQL );
123 ( SELECT u.* FROM user u
124 INNER JOIN user_friends f ON u.id = f.user_id
125 WHERE f.friend_user_id = ?
127 SELECT u.* FROM user u
128 INNER JOIN user_friends f ON u.id = f.friend_user_id
129 WHERE f.user_id = ? )
132 # Finally, register your new ResultSource with your Schema
133 My::Schema->register_extra_source( 'UserFriendsComplex' => $new_source );
135 Next, you can execute your complex query using bind parameters like this:
137 my $friends = [ $schema->resultset( 'UserFriendsComplex' )->search( {},
139 bind => [ 12345, 12345 ]
143 ... and you'll get back a perfect L<DBIx::Class::ResultSet> (except, of course,
144 that you cannot modify the rows it contains, ie. cannot call L</update>,
145 L</delete>, ... on it).
147 If you prefer to have the definitions of these custom ResultSources in separate
148 files (instead of stuffing all of them into the same resultset class), you can
149 achieve the same with subclassing the resultset class and defining the
152 package My::Schema::UserFriendsComplex;
154 use My::Schema::User;
155 use base qw/My::Schema::User/;
157 __PACKAGE__->table('dummy'); # currently must be called before anything else
159 # Hand in your query as a scalar reference
160 # It will be added as a sub-select after FROM,
161 # so pay attention to the surrounding brackets!
162 __PACKAGE__->name( \<<SQL );
163 ( SELECT u.* FROM user u
164 INNER JOIN user_friends f ON u.id = f.user_id
165 WHERE f.friend_user_id = ?
167 SELECT u.* FROM user u
168 INNER JOIN user_friends f ON u.id = f.friend_user_id
169 WHERE f.user_id = ? )
174 =head2 Using specific columns
176 When you only want specific columns from a table, you can use
177 C<columns> to specify which ones you need. This is useful to avoid
178 loading columns with large amounts of data that you aren't about to
181 my $rs = $schema->resultset('Artist')->search(
184 columns => [qw/ name /]
189 # SELECT artist.name FROM artist
191 This is a shortcut for C<select> and C<as>, see below. C<columns>
192 cannot be used together with C<select> and C<as>.
194 =head2 Using database functions or stored procedures
196 The combination of C<select> and C<as> can be used to return the result of a
197 database function or stored procedure as a column value. You use C<select> to
198 specify the source for your column value (e.g. a column name, function, or
199 stored procedure name). You then use C<as> to set the column name you will use
200 to access the returned value:
202 my $rs = $schema->resultset('Artist')->search(
205 select => [ 'name', { LENGTH => 'name' } ],
206 as => [qw/ name name_length /],
211 # SELECT name name, LENGTH( name )
214 Note that the C< as > attribute has absolutely nothing to with the sql
215 syntax C< SELECT foo AS bar > (see the documentation in
216 L<DBIx::Class::ResultSet/ATTRIBUTES>). If your alias exists as a
217 column in your base class (i.e. it was added with C<add_columns>), you
218 just access it as normal. Our C<Artist> class has a C<name> column, so
219 we just use the C<name> accessor:
221 my $artist = $rs->first();
222 my $name = $artist->name();
224 If on the other hand the alias does not correspond to an existing column, you
225 have to fetch the value using the C<get_column> accessor:
227 my $name_length = $artist->get_column('name_length');
229 If you don't like using C<get_column>, you can always create an accessor for
230 any of your aliases using either of these:
232 # Define accessor manually:
233 sub name_length { shift->get_column('name_length'); }
235 # Or use DBIx::Class::AccessorGroup:
236 __PACKAGE__->mk_group_accessors('column' => 'name_length');
238 =head2 SELECT DISTINCT with multiple columns
240 my $rs = $schema->resultset('Foo')->search(
244 { distinct => [ $source->columns ] }
246 as => [ $source->columns ] # remember 'as' is not the same as SQL AS :-)
250 =head2 SELECT COUNT(DISTINCT colname)
252 my $rs = $schema->resultset('Foo')->search(
256 { count => { distinct => 'colname' } }
262 my $count = $rs->next->get_column('count');
264 =head2 Grouping results
266 L<DBIx::Class> supports C<GROUP BY> as follows:
268 my $rs = $schema->resultset('Artist')->search(
272 select => [ 'name', { count => 'cds.id' } ],
273 as => [qw/ name cd_count /],
274 group_by => [qw/ name /]
279 # SELECT name, COUNT( cd.id ) FROM artist
280 # LEFT JOIN cd ON artist.id = cd.artist
283 Please see L<DBIx::Class::ResultSet/ATTRIBUTES> documentation if you
284 are in any way unsure about the use of the attributes above (C< join
285 >, C< select >, C< as > and C< group_by >).
287 =head2 Subqueries (EXPERIMENTAL)
289 You can write subqueries relatively easily in DBIC.
291 my $inside_rs = $schema->resultset('Artist')->search({
292 name => [ 'Billy Joel', 'Brittany Spears' ],
295 my $rs = $schema->resultset('CD')->search({
296 artist_id => { 'IN' => $inside_rs->get_column('id')->as_query },
299 The usual operators ( =, !=, IN, NOT IN, etc) are supported.
301 B<NOTE>: You have to explicitly use '=' when doing an equality comparison.
302 The following will B<not> work:
304 my $rs = $schema->resultset('CD')->search({
305 artist_id => $inside_rs->get_column('id')->as_query,
310 Subqueries are supported in the where clause (first hashref), and in the
311 from, select, and +select attributes.
313 =head3 Correlated subqueries
315 my $cdrs = $schema->resultset('CD');
316 my $rs = $cdrs->search({
318 '=' => $cdrs->search(
319 { artistid => { '=' => \'me.artistid' } },
321 )->get_column('year')->max_rs->as_query,
325 That creates the following SQL:
327 SELECT me.cdid, me.artist, me.title, me.year, me.genreid, me.single_track
330 SELECT MAX(inner.year)
332 WHERE artistid = me.artistid
337 Please note that subqueries are considered an experimental feature.
339 =head2 Predefined searches
341 You can write your own L<DBIx::Class::ResultSet> class by inheriting from it
342 and define often used searches as methods:
344 package My::DBIC::ResultSet::CD;
347 use base 'DBIx::Class::ResultSet';
349 sub search_cds_ordered {
352 return $self->search(
354 { order_by => 'name DESC' },
360 To use your resultset, first tell DBIx::Class to create an instance of it
361 for you, in your My::DBIC::Schema::CD class:
363 # class definition as normal
364 __PACKAGE__->load_components(qw/ Core /);
365 __PACKAGE__->table('cd');
367 # tell DBIC to use the custom ResultSet class
368 __PACKAGE__->resultset_class('My::DBIC::ResultSet::CD');
370 Note that C<resultset_class> must be called after C<load_components> and C<table>, or you will get errors about missing methods.
372 Then call your new method in your code:
374 my $ordered_cds = $schema->resultset('CD')->search_cds_ordered();
376 =head2 Using SQL functions on the left hand side of a comparison
378 Using SQL functions on the left hand side of a comparison is generally
379 not a good idea since it requires a scan of the entire table. However,
380 it can be accomplished with C<DBIx::Class> when necessary.
382 If you do not have quoting on, simply include the function in your search
383 specification as you would any column:
385 $rs->search({ 'YEAR(date_of_birth)' => 1979 });
387 With quoting on, or for a more portable solution, use the C<where>
390 $rs->search({}, { where => \'YEAR(date_of_birth) = 1979' });
394 (When the bind args ordering bug is fixed, this technique will be better
395 and can replace the one above.)
397 With quoting on, or for a more portable solution, use the C<where> and
401 where => \'YEAR(date_of_birth) = ?',
407 =head1 JOINS AND PREFETCHING
409 =head2 Using joins and prefetch
411 You can use the C<join> attribute to allow searching on, or sorting your
412 results by, one or more columns in a related table. To return all CDs matching
413 a particular artist name:
415 my $rs = $schema->resultset('CD')->search(
417 'artist.name' => 'Bob Marley'
420 join => 'artist', # join the artist table
425 # SELECT cd.* FROM cd
426 # JOIN artist ON cd.artist = artist.id
427 # WHERE artist.name = 'Bob Marley'
429 If required, you can now sort on any column in the related tables by including
430 it in your C<order_by> attribute:
432 my $rs = $schema->resultset('CD')->search(
434 'artist.name' => 'Bob Marley'
438 order_by => [qw/ artist.name /]
443 # SELECT cd.* FROM cd
444 # JOIN artist ON cd.artist = artist.id
445 # WHERE artist.name = 'Bob Marley'
446 # ORDER BY artist.name
448 Note that the C<join> attribute should only be used when you need to search or
449 sort using columns in a related table. Joining related tables when you only
450 need columns from the main table will make performance worse!
452 Now let's say you want to display a list of CDs, each with the name of the
453 artist. The following will work fine:
455 while (my $cd = $rs->next) {
456 print "CD: " . $cd->title . ", Artist: " . $cd->artist->name;
459 There is a problem however. We have searched both the C<cd> and C<artist> tables
460 in our main query, but we have only returned data from the C<cd> table. To get
461 the artist name for any of the CD objects returned, L<DBIx::Class> will go back
464 SELECT artist.* FROM artist WHERE artist.id = ?
466 A statement like the one above will run for each and every CD returned by our
467 main query. Five CDs, five extra queries. A hundred CDs, one hundred extra
470 Thankfully, L<DBIx::Class> has a C<prefetch> attribute to solve this problem.
471 This allows you to fetch results from related tables in advance:
473 my $rs = $schema->resultset('CD')->search(
475 'artist.name' => 'Bob Marley'
479 order_by => [qw/ artist.name /],
480 prefetch => 'artist' # return artist data too!
484 # Equivalent SQL (note SELECT from both "cd" and "artist"):
485 # SELECT cd.*, artist.* FROM cd
486 # JOIN artist ON cd.artist = artist.id
487 # WHERE artist.name = 'Bob Marley'
488 # ORDER BY artist.name
490 The code to print the CD list remains the same:
492 while (my $cd = $rs->next) {
493 print "CD: " . $cd->title . ", Artist: " . $cd->artist->name;
496 L<DBIx::Class> has now prefetched all matching data from the C<artist> table,
497 so no additional SQL statements are executed. You now have a much more
500 Note that as of L<DBIx::Class> 0.05999_01, C<prefetch> I<can> be used with
501 C<has_many> relationships.
503 Also note that C<prefetch> should only be used when you know you will
504 definitely use data from a related table. Pre-fetching related tables when you
505 only need columns from the main table will make performance worse!
507 =head2 Multiple joins
509 In the examples above, the C<join> attribute was a scalar. If you
510 pass an array reference instead, you can join to multiple tables. In
511 this example, we want to limit the search further, using
514 # Relationships defined elsewhere:
515 # CD->belongs_to('artist' => 'Artist');
516 # CD->has_one('liner_notes' => 'LinerNotes', 'cd');
517 my $rs = $schema->resultset('CD')->search(
519 'artist.name' => 'Bob Marley'
520 'liner_notes.notes' => { 'like', '%some text%' },
523 join => [qw/ artist liner_notes /],
524 order_by => [qw/ artist.name /],
529 # SELECT cd.*, artist.*, liner_notes.* FROM cd
530 # JOIN artist ON cd.artist = artist.id
531 # JOIN liner_notes ON cd.id = liner_notes.cd
532 # WHERE artist.name = 'Bob Marley'
533 # ORDER BY artist.name
535 =head2 Multi-step joins
537 Sometimes you want to join more than one relationship deep. In this example,
538 we want to find all C<Artist> objects who have C<CD>s whose C<LinerNotes>
539 contain a specific string:
541 # Relationships defined elsewhere:
542 # Artist->has_many('cds' => 'CD', 'artist');
543 # CD->has_one('liner_notes' => 'LinerNotes', 'cd');
545 my $rs = $schema->resultset('Artist')->search(
547 'liner_notes.notes' => { 'like', '%some text%' },
551 'cds' => 'liner_notes'
557 # SELECT artist.* FROM artist
558 # LEFT JOIN cd ON artist.id = cd.artist
559 # LEFT JOIN liner_notes ON cd.id = liner_notes.cd
560 # WHERE liner_notes.notes LIKE '%some text%'
562 Joins can be nested to an arbitrary level. So if we decide later that we
563 want to reduce the number of Artists returned based on who wrote the liner
566 # Relationship defined elsewhere:
567 # LinerNotes->belongs_to('author' => 'Person');
569 my $rs = $schema->resultset('Artist')->search(
571 'liner_notes.notes' => { 'like', '%some text%' },
572 'author.name' => 'A. Writer'
577 'liner_notes' => 'author'
584 # SELECT artist.* FROM artist
585 # LEFT JOIN cd ON artist.id = cd.artist
586 # LEFT JOIN liner_notes ON cd.id = liner_notes.cd
587 # LEFT JOIN author ON author.id = liner_notes.author
588 # WHERE liner_notes.notes LIKE '%some text%'
589 # AND author.name = 'A. Writer'
591 =head2 Multi-step and multiple joins
593 With various combinations of array and hash references, you can join
594 tables in any combination you desire. For example, to join Artist to
595 CD and Concert, and join CD to LinerNotes:
597 # Relationships defined elsewhere:
598 # Artist->has_many('concerts' => 'Concert', 'artist');
600 my $rs = $schema->resultset('Artist')->search(
613 # SELECT artist.* FROM artist
614 # LEFT JOIN cd ON artist.id = cd.artist
615 # LEFT JOIN liner_notes ON cd.id = liner_notes.cd
616 # LEFT JOIN concert ON artist.id = concert.artist
618 =head2 Multi-step prefetch
620 From 0.04999_05 onwards, C<prefetch> can be nested more than one relationship
621 deep using the same syntax as a multi-step join:
623 my $rs = $schema->resultset('Tag')->search(
633 # SELECT tag.*, cd.*, artist.* FROM tag
634 # JOIN cd ON tag.cd = cd.id
635 # JOIN artist ON cd.artist = artist.id
637 Now accessing our C<cd> and C<artist> relationships does not need additional
640 my $tag = $rs->first;
641 print $tag->cd->artist->name;
643 =head1 ROW-LEVEL OPERATIONS
645 =head2 Retrieving a row object's Schema
647 It is possible to get a Schema object from a row object like so:
649 my $schema = $cd->result_source->schema;
650 # use the schema as normal:
651 my $artist_rs = $schema->resultset('Artist');
653 This can be useful when you don't want to pass around a Schema object to every
656 =head2 Getting the value of the primary key for the last database insert
658 AKA getting last_insert_id
660 If you are using PK::Auto (which is a core component as of 0.07), this is
663 my $foo = $rs->create(\%blah);
665 my $id = $foo->id; # foo->my_primary_key_field will also work.
667 If you are not using autoincrementing primary keys, this will probably
668 not work, but then you already know the value of the last primary key anyway.
670 =head2 Stringification
672 Employ the standard stringification technique by using the C<overload>
675 To make an object stringify itself as a single column, use something
676 like this (replace C<name> with the column/method of your choice):
678 use overload '""' => sub { shift->name}, fallback => 1;
680 For more complex stringification, you can use an anonymous subroutine:
682 use overload '""' => sub { $_[0]->name . ", " .
683 $_[0]->address }, fallback => 1;
685 =head3 Stringification Example
687 Suppose we have two tables: C<Product> and C<Category>. The table
690 Product(id, Description, category)
691 Category(id, Description)
693 C<category> is a foreign key into the Category table.
695 If you have a Product object C<$obj> and write something like
699 things will not work as expected.
701 To obtain, for example, the category description, you should add this
702 method to the class defining the Category table:
704 use overload "" => sub {
707 return $self->Description;
710 =head2 Want to know if find_or_create found or created a row?
712 Just use C<find_or_new> instead, then check C<in_storage>:
714 my $obj = $rs->find_or_new({ blah => 'blarg' });
715 unless ($obj->in_storage) {
717 # do whatever else you wanted if it was a new row
720 =head2 Dynamic Sub-classing DBIx::Class proxy classes
722 AKA multi-class object inflation from one table
724 L<DBIx::Class> classes are proxy classes, therefore some different
725 techniques need to be employed for more than basic subclassing. In
726 this example we have a single user table that carries a boolean bit
727 for admin. We would like like to give the admin users
728 objects(L<DBIx::Class::Row>) the same methods as a regular user but
729 also special admin only methods. It doesn't make sense to create two
730 seperate proxy-class files for this. We would be copying all the user
731 methods into the Admin class. There is a cleaner way to accomplish
734 Overriding the C<inflate_result> method within the User proxy-class
735 gives us the effect we want. This method is called by
736 L<DBIx::Class::ResultSet> when inflating a result from storage. So we
737 grab the object being returned, inspect the values we are looking for,
738 bless it if it's an admin object, and then return it. See the example
745 use base qw/DBIx::Class::Schema/;
747 __PACKAGE__->load_classes(qw/User/);
750 B<Proxy-Class definitions>
752 package DB::Schema::User;
756 use base qw/DBIx::Class/;
758 ### Defined what our admin class is for ensure_class_loaded
759 my $admin_class = __PACKAGE__ . '::Admin';
761 __PACKAGE__->load_components(qw/Core/);
763 __PACKAGE__->table('users');
765 __PACKAGE__->add_columns(qw/user_id email password
766 firstname lastname active
769 __PACKAGE__->set_primary_key('user_id');
773 my $ret = $self->next::method(@_);
774 if( $ret->admin ) {### If this is an admin rebless for extra functions
775 $self->ensure_class_loaded( $admin_class );
776 bless $ret, $admin_class;
782 print "I am a regular user.\n";
787 package DB::Schema::User::Admin;
791 use base qw/DB::Schema::User/;
795 print "I am an admin.\n";
801 print "I am doing admin stuff\n";
811 my $user_data = { email => 'someguy@place.com',
815 my $admin_data = { email => 'someadmin@adminplace.com',
819 my $schema = DB::Schema->connection('dbi:Pg:dbname=test');
821 $schema->resultset('User')->create( $user_data );
822 $schema->resultset('User')->create( $admin_data );
824 ### Now we search for them
825 my $user = $schema->resultset('User')->single( $user_data );
826 my $admin = $schema->resultset('User')->single( $admin_data );
828 print ref $user, "\n";
829 print ref $admin, "\n";
831 print $user->password , "\n"; # pass1
832 print $admin->password , "\n";# pass2; inherited from User
833 print $user->hello , "\n";# I am a regular user.
834 print $admin->hello, "\n";# I am an admin.
836 ### The statement below will NOT print
837 print "I can do admin stuff\n" if $user->can('do_admin_stuff');
838 ### The statement below will print
839 print "I can do admin stuff\n" if $admin->can('do_admin_stuff');
841 =head2 Skip row object creation for faster results
843 DBIx::Class is not built for speed, it's built for convenience and
844 ease of use, but sometimes you just need to get the data, and skip the
847 To do this simply use L<DBIx::Class::ResultClass::HashRefInflator>.
849 my $rs = $schema->resultset('CD');
851 $rs->result_class('DBIx::Class::ResultClass::HashRefInflator');
853 my $hash_ref = $rs->find(1);
857 =head2 Get raw data for blindingly fast results
859 If the L<HashRefInflator|DBIx::Class::ResultClass::HashRefInflator> solution
860 above is not fast enough for you, you can use a DBIx::Class to return values
861 exactly as they come out of the data base with none of the convenience methods
864 This is used like so:
866 my $cursor = $rs->cursor
867 while (my @vals = $cursor->next) {
868 # use $val[0..n] here
871 You will need to map the array offsets to particular columns (you can
872 use the I<select> attribute of C<search()> to force ordering).
874 =head1 RESULTSET OPERATIONS
876 =head2 Getting Schema from a ResultSet
878 To get the schema object from a result set, do the following:
880 $rs->result_source->schema
882 =head2 Getting Columns Of Data
886 If you want to find the sum of a particular column there are several
887 ways, the obvious one is to use search:
889 my $rs = $schema->resultset('Items')->search(
892 select => [ { sum => 'Cost' } ],
893 as => [ 'total_cost' ], # remember this 'as' is for DBIx::Class::ResultSet not SQL
896 my $tc = $rs->first->get_column('total_cost');
898 Or, you can use the L<DBIx::Class::ResultSetColumn>, which gets
899 returned when you ask the C<ResultSet> for a column using
902 my $cost = $schema->resultset('Items')->get_column('Cost');
905 With this you can also do:
907 my $minvalue = $cost->min;
908 my $maxvalue = $cost->max;
910 Or just iterate through the values of this column only:
912 while ( my $c = $cost->next ) {
916 foreach my $c ($cost->all) {
920 C<ResultSetColumn> only has a limited number of built-in functions, if
921 you need one that it doesn't have, then you can use the C<func> method
924 my $avg = $cost->func('AVERAGE');
926 This will cause the following SQL statement to be run:
928 SELECT AVERAGE(Cost) FROM Items me
930 Which will of course only work if your database supports this function.
931 See L<DBIx::Class::ResultSetColumn> for more documentation.
933 =head2 Creating a result set from a set of rows
935 Sometimes you have a (set of) row objects that you want to put into a
936 resultset without the need to hit the DB again. You can do that by using the
937 L<set_cache|DBIx::Class::Resultset/set_cache> method:
939 my @uploadable_groups;
940 while (my $group = $groups->next) {
941 if ($group->can_upload($self)) {
942 push @uploadable_groups, $group;
945 my $new_rs = $self->result_source->resultset;
946 $new_rs->set_cache(\@uploadable_groups);
950 =head1 USING RELATIONSHIPS
952 =head2 Create a new row in a related table
954 my $author = $book->create_related('author', { name => 'Fred'});
956 =head2 Search in a related table
958 Only searches for books named 'Titanic' by the author in $author.
960 my $books_rs = $author->search_related('books', { name => 'Titanic' });
962 =head2 Delete data in a related table
964 Deletes only the book named Titanic by the author in $author.
966 $author->delete_related('books', { name => 'Titanic' });
968 =head2 Ordering a relationship result set
970 If you always want a relation to be ordered, you can specify this when you
971 create the relationship.
973 To order C<< $book->pages >> by descending page_number, create the relation
976 __PACKAGE__->has_many('pages' => 'Page', 'book', { order_by => \'page_number DESC'} );
978 =head2 Filtering a relationship result set
980 If you want to get a filtered result set, you can just add add to $attr as follows:
982 __PACKAGE__->has_many('pages' => 'Page', 'book', { where => { scrap => 0 } } );
984 =head2 Many-to-many relationships
986 This is straightforward using L<ManyToMany|DBIx::Class::Relationship/many_to_many>:
989 use base 'DBIx::Class';
990 __PACKAGE__->load_components('Core');
991 __PACKAGE__->table('user');
992 __PACKAGE__->add_columns(qw/id name/);
993 __PACKAGE__->set_primary_key('id');
994 __PACKAGE__->has_many('user_address' => 'My::UserAddress', 'user');
995 __PACKAGE__->many_to_many('addresses' => 'user_address', 'address');
997 package My::UserAddress;
998 use base 'DBIx::Class';
999 __PACKAGE__->load_components('Core');
1000 __PACKAGE__->table('user_address');
1001 __PACKAGE__->add_columns(qw/user address/);
1002 __PACKAGE__->set_primary_key(qw/user address/);
1003 __PACKAGE__->belongs_to('user' => 'My::User');
1004 __PACKAGE__->belongs_to('address' => 'My::Address');
1006 package My::Address;
1007 use base 'DBIx::Class';
1008 __PACKAGE__->load_components('Core');
1009 __PACKAGE__->table('address');
1010 __PACKAGE__->add_columns(qw/id street town area_code country/);
1011 __PACKAGE__->set_primary_key('id');
1012 __PACKAGE__->has_many('user_address' => 'My::UserAddress', 'address');
1013 __PACKAGE__->many_to_many('users' => 'user_address', 'user');
1015 $rs = $user->addresses(); # get all addresses for a user
1016 $rs = $address->users(); # get all users for an address
1020 As of version 0.04001, there is improved transaction support in
1021 L<DBIx::Class::Storage> and L<DBIx::Class::Schema>. Here is an
1022 example of the recommended way to use it:
1024 my $genus = $schema->resultset('Genus')->find(12);
1026 my $coderef2 = sub {
1031 my $coderef1 = sub {
1032 $genus->add_to_species({ name => 'troglodyte' });
1035 $schema->txn_do($coderef2); # Can have a nested transaction. Only the outer will actualy commit
1036 return $genus->species;
1041 $rs = $schema->txn_do($coderef1);
1044 if ($@) { # Transaction failed
1045 die "the sky is falling!" #
1046 if ($@ =~ /Rollback failed/); # Rollback failed
1048 deal_with_failed_transaction();
1051 Nested transactions will work as expected. That is, only the outermost
1052 transaction will actually issue a commit to the $dbh, and a rollback
1053 at any level of any transaction will cause the entire nested
1054 transaction to fail. Support for savepoints and for true nested
1055 transactions (for databases that support them) will hopefully be added
1060 =head2 Creating Schemas From An Existing Database
1062 L<DBIx::Class::Schema::Loader> will connect to a database and create a
1063 L<DBIx::Class::Schema> and associated sources by examining the database.
1065 The recommend way of achieving this is to use the
1066 L<make_schema_at|DBIx::Class::Schema::Loader/make_schema_at> method:
1068 perl -MDBIx::Class::Schema::Loader=make_schema_at,dump_to_dir:./lib \
1069 -e 'make_schema_at("My::Schema", { debug => 1 }, [ "dbi:Pg:dbname=foo","postgres" ])'
1071 This will create a tree of files rooted at C<./lib/My/Schema/> containing
1072 source definitions for all the tables found in the C<foo> database.
1074 =head2 Creating DDL SQL
1076 The following functionality requires you to have L<SQL::Translator>
1077 (also known as "SQL Fairy") installed.
1079 To create a set of database-specific .sql files for the above schema:
1081 my $schema = My::Schema->connect($dsn);
1082 $schema->create_ddl_dir(['MySQL', 'SQLite', 'PostgreSQL'],
1087 By default this will create schema files in the current directory, for
1088 MySQL, SQLite and PostgreSQL, using the $VERSION from your Schema.pm.
1090 To create a new database using the schema:
1092 my $schema = My::Schema->connect($dsn);
1093 $schema->deploy({ add_drop_tables => 1});
1095 To import created .sql files using the mysql client:
1097 mysql -h "host" -D "database" -u "user" -p < My_Schema_1.0_MySQL.sql
1099 To create C<ALTER TABLE> conversion scripts to update a database to a
1100 newer version of your schema at a later point, first set a new
1101 C<$VERSION> in your Schema file, then:
1103 my $schema = My::Schema->connect($dsn);
1104 $schema->create_ddl_dir(['MySQL', 'SQLite', 'PostgreSQL'],
1110 This will produce new database-specific .sql files for the new version
1111 of the schema, plus scripts to convert from version 0.1 to 0.2. This
1112 requires that the files for 0.1 as created above are available in the
1113 given directory to diff against.
1115 =head2 Select from dual
1117 Dummy tables are needed by some databases to allow calling functions
1118 or expressions that aren't based on table content, for examples of how
1119 this applies to various database types, see:
1120 L<http://troels.arvin.dk/db/rdbms/#other-dummy_table>.
1122 Note: If you're using Oracles dual table don't B<ever> do anything
1123 other than a select, if you CRUD on your dual table you *will* break
1126 Make a table class as you would for any other table
1128 package MyAppDB::Dual;
1131 use base 'DBIx::Class';
1132 __PACKAGE__->load_components("Core");
1133 __PACKAGE__->table("Dual");
1134 __PACKAGE__->add_columns(
1136 { data_type => "VARCHAR2", is_nullable => 0, size => 1 },
1139 Once you've loaded your table class select from it using C<select>
1140 and C<as> instead of C<columns>
1142 my $rs = $schema->resultset('Dual')->search(undef,
1143 { select => [ 'sydate' ],
1148 All you have to do now is be careful how you access your resultset, the below
1149 will not work because there is no column called 'now' in the Dual table class
1151 while (my $dual = $rs->next) {
1152 print $dual->now."\n";
1154 # Can't locate object method "now" via package "MyAppDB::Dual" at headshot.pl line 23.
1156 You could of course use 'dummy' in C<as> instead of 'now', or C<add_columns> to
1157 your Dual class for whatever you wanted to select from dual, but that's just
1158 silly, instead use C<get_column>
1160 while (my $dual = $rs->next) {
1161 print $dual->get_column('now')."\n";
1166 my $cursor = $rs->cursor;
1167 while (my @vals = $cursor->next) {
1168 print $vals[0]."\n";
1171 In case you're going to use this "trick" together with L<DBIx::Class::Schema/deploy> or
1172 L<DBIx::Class::Schema/create_ddl_dir> a table called "dual" will be created in your
1173 current schema. This would overlap "sys.dual" and you could not fetch "sysdate" or
1174 "sequence.nextval" anymore from dual. To avoid this problem, just tell
1175 L<SQL::Translator> to not create table dual:
1178 add_drop_table => 1,
1179 parser_args => { sources => [ grep $_ ne 'Dual', schema->sources ] },
1181 $schema->create_ddl_dir( [qw/Oracle/], undef, './sql', undef, $sqlt_args );
1183 Or use L<DBIx::Class::ResultClass::HashRefInflator>
1185 $rs->result_class('DBIx::Class::ResultClass::HashRefInflator');
1186 while ( my $dual = $rs->next ) {
1187 print $dual->{now}."\n";
1190 Here are some example C<select> conditions to illustrate the different syntax
1191 you could use for doing stuff like
1192 C<oracles.heavily(nested(functions_can('take', 'lots'), OF), 'args')>
1194 # get a sequence value
1195 select => [ 'A_SEQ.nextval' ],
1197 # get create table sql
1198 select => [ { 'dbms_metadata.get_ddl' => [ "'TABLE'", "'ARTIST'" ]} ],
1200 # get a random num between 0 and 100
1201 select => [ { "trunc" => [ { "dbms_random.value" => [0,100] } ]} ],
1204 select => [ { 'extract' => [ \'year from sysdate' ] } ],
1207 select => [ {'round' => [{'cos' => [ \'180 * 3.14159265359/180' ]}]}],
1209 # which day of the week were you born on?
1210 select => [{'to_char' => [{'to_date' => [ "'25-DEC-1980'", "'dd-mon-yyyy'" ]}, "'day'"]}],
1212 # select 16 rows from dual
1213 select => [ "'hello'" ],
1215 group_by => [ 'cube( 1, 2, 3, 4 )' ],
1219 =head2 Adding Indexes And Functions To Your SQL
1221 Often you will want indexes on columns on your table to speed up searching. To
1222 do this, create a method called C<sqlt_deploy_hook> in the relevant source
1223 class (refer to the advanced
1224 L<callback system|DBIx::Class::ResultSource/sqlt_deploy_callback> if you wish
1225 to share a hook between multiple sources):
1227 package My::Schema::Artist;
1229 __PACKAGE__->table('artist');
1230 __PACKAGE__->add_columns(id => { ... }, name => { ... })
1232 sub sqlt_deploy_hook {
1233 my ($self, $sqlt_table) = @_;
1235 $sqlt_table->add_index(name => 'idx_name', fields => ['name']);
1240 Sometimes you might want to change the index depending on the type of the
1241 database for which SQL is being generated:
1243 my ($db_type = $sqlt_table->schema->translator->producer_type)
1244 =~ s/^SQL::Translator::Producer:://;
1246 You can also add hooks to the schema level to stop certain tables being
1253 sub sqlt_deploy_hook {
1254 my ($self, $sqlt_schema) = @_;
1256 $sqlt_schema->drop_table('table_name');
1259 You could also add views, procedures or triggers to the output using
1260 L<SQL::Translator::Schema/add_view>,
1261 L<SQL::Translator::Schema/add_procedure> or
1262 L<SQL::Translator::Schema/add_trigger>.
1265 =head2 Schema versioning
1267 The following example shows simplistically how you might use DBIx::Class to
1268 deploy versioned schemas to your customers. The basic process is as follows:
1274 Create a DBIx::Class schema
1286 Modify schema to change functionality
1290 Deploy update to customers
1294 B<Create a DBIx::Class schema>
1296 This can either be done manually, or generated from an existing database as
1297 described under L</Creating Schemas From An Existing Database>
1301 Call L<DBIx::Class::Schema/create_ddl_dir> as above under L</Creating DDL SQL>.
1303 B<Deploy to customers>
1305 There are several ways you could deploy your schema. These are probably
1306 beyond the scope of this recipe, but might include:
1312 Require customer to apply manually using their RDBMS.
1316 Package along with your app, making database dump/schema update/tests
1317 all part of your install.
1321 B<Modify the schema to change functionality>
1323 As your application evolves, it may be necessary to modify your schema
1324 to change functionality. Once the changes are made to your schema in
1325 DBIx::Class, export the modified schema and the conversion scripts as
1326 in L</Creating DDL SQL>.
1328 B<Deploy update to customers>
1330 Add the L<DBIx::Class::Schema::Versioned> schema component to your
1331 Schema class. This will add a new table to your database called
1332 C<dbix_class_schema_vesion> which will keep track of which version is installed
1333 and warn if the user trys to run a newer schema version than the
1334 database thinks it has.
1336 Alternatively, you can send the conversion sql scripts to your
1339 =head2 Setting quoting for the generated SQL.
1341 If the database contains column names with spaces and/or reserved words, they
1342 need to be quoted in the SQL queries. This is done using:
1344 __PACKAGE__->storage->sql_maker->quote_char([ qw/[ ]/] );
1345 __PACKAGE__->storage->sql_maker->name_sep('.');
1347 The first sets the quote characters. Either a pair of matching
1348 brackets, or a C<"> or C<'>:
1350 __PACKAGE__->storage->sql_maker->quote_char('"');
1352 Check the documentation of your database for the correct quote
1353 characters to use. C<name_sep> needs to be set to allow the SQL
1354 generator to put the quotes the correct place.
1356 In most cases you should set these as part of the arguments passed to
1357 L<DBIx::Class::Schema/connect>:
1359 my $schema = My::Schema->connect(
1369 =head2 Setting limit dialect for SQL::Abstract::Limit
1371 In some cases, SQL::Abstract::Limit cannot determine the dialect of
1372 the remote SQL server by looking at the database handle. This is a
1373 common problem when using the DBD::JDBC, since the DBD-driver only
1374 know that in has a Java-driver available, not which JDBC driver the
1375 Java component has loaded. This specifically sets the limit_dialect
1376 to Microsoft SQL-server (See more names in SQL::Abstract::Limit
1379 __PACKAGE__->storage->sql_maker->limit_dialect('mssql');
1381 The JDBC bridge is one way of getting access to a MSSQL server from a platform
1382 that Microsoft doesn't deliver native client libraries for. (e.g. Linux)
1384 The limit dialect can also be set at connect time by specifying a
1385 C<limit_dialect> key in the final hash as shown above.
1387 =head2 Working with PostgreSQL array types
1389 You can also assign values to PostgreSQL array columns by passing array
1390 references in the C<\%columns> (C<\%vals>) hashref of the
1391 L<DBIx::Class::ResultSet/create> and L<DBIx::Class::Row/update> family of
1394 $resultset->create({
1395 numbers => [1, 2, 3]
1400 numbers => [1, 2, 3]
1407 In conditions (eg. C<\%cond> in the L<DBIx::Class::ResultSet/search> family of
1408 methods) you cannot directly use array references (since this is interpreted as
1409 a list of values to be C<OR>ed), but you can use the following syntax to force
1410 passing them as bind values:
1414 numbers => \[ '= ?', [numbers => [1, 2, 3]] ]
1418 See L<SQL::Abstract/array_datatypes> and L<SQL::Abstract/Literal SQL with
1419 placeholders and bind values (subqueries)> for more explanation. Note that
1420 L<DBIx::Class> sets L<SQL::Abstract/bindtype> to C<columns>, so you must pass
1421 the bind values (the C<[1, 2, 3]> arrayref in the above example) wrapped in
1422 arrayrefs together with the column name, like this: C<< [column_name => value]
1425 =head1 BOOTSTRAPPING/MIGRATING
1427 =head2 Easy migration from class-based to schema-based setup
1429 You want to start using the schema-based approach to L<DBIx::Class>
1430 (see L<SchemaIntro.pod>), but have an established class-based setup with lots
1431 of existing classes that you don't want to move by hand. Try this nifty script
1435 use SQL::Translator;
1437 my $schema = MyDB->schema_instance;
1439 my $translator = SQL::Translator->new(
1440 debug => $debug || 0,
1441 trace => $trace || 0,
1442 no_comments => $no_comments || 0,
1443 show_warnings => $show_warnings || 0,
1444 add_drop_table => $add_drop_table || 0,
1445 validate => $validate || 0,
1447 'DBIx::Schema' => $schema,
1450 'prefix' => 'My::Schema',
1454 $translator->parser('SQL::Translator::Parser::DBIx::Class');
1455 $translator->producer('SQL::Translator::Producer::DBIx::Class::File');
1457 my $output = $translator->translate(@args) or die
1458 "Error: " . $translator->error;
1462 You could use L<Module::Find> to search for all subclasses in the MyDB::*
1463 namespace, which is currently left as an exercise for the reader.
1465 =head1 OVERLOADING METHODS
1467 L<DBIx::Class> uses the L<Class::C3> package, which provides for redispatch of
1468 method calls, useful for things like default values and triggers. You have to
1469 use calls to C<next::method> to overload methods. More information on using
1470 L<Class::C3> with L<DBIx::Class> can be found in
1471 L<DBIx::Class::Manual::Component>.
1473 =head2 Setting default values for a row
1475 It's as simple as overriding the C<new> method. Note the use of
1479 my ( $class, $attrs ) = @_;
1481 $attrs->{foo} = 'bar' unless defined $attrs->{foo};
1483 my $new = $class->next::method($attrs);
1488 For more information about C<next::method>, look in the L<Class::C3>
1489 documentation. See also L<DBIx::Class::Manual::Component> for more
1490 ways to write your own base classes to do this.
1492 People looking for ways to do "triggers" with DBIx::Class are probably
1493 just looking for this.
1495 =head2 Changing one field whenever another changes
1497 For example, say that you have three columns, C<id>, C<number>, and
1498 C<squared>. You would like to make changes to C<number> and have
1499 C<squared> be automagically set to the value of C<number> squared.
1500 You can accomplish this by overriding C<store_column>:
1503 my ( $self, $name, $value ) = @_;
1504 if ($name eq 'number') {
1505 $self->squared($value * $value);
1507 $self->next::method($name, $value);
1510 Note that the hard work is done by the call to C<next::method>, which
1511 redispatches your call to store_column in the superclass(es).
1513 =head2 Automatically creating related objects
1515 You might have a class C<Artist> which has many C<CD>s. Further, if you
1516 want to create a C<CD> object every time you insert an C<Artist> object.
1517 You can accomplish this by overriding C<insert> on your objects:
1520 my ( $self, @args ) = @_;
1521 $self->next::method(@args);
1522 $self->cds->new({})->fill_from_artist($self)->insert;
1526 where C<fill_from_artist> is a method you specify in C<CD> which sets
1527 values in C<CD> based on the data in the C<Artist> object you pass in.
1529 =head2 Wrapping/overloading a column accessor
1533 Say you have a table "Camera" and want to associate a description
1534 with each camera. For most cameras, you'll be able to generate the description from
1535 the other columns. However, in a few special cases you may want to associate a
1536 custom description with a camera.
1540 In your database schema, define a description field in the "Camera" table that
1541 can contain text and null values.
1543 In DBIC, we'll overload the column accessor to provide a sane default if no
1544 custom description is defined. The accessor will either return or generate the
1545 description, depending on whether the field is null or not.
1547 First, in your "Camera" schema class, define the description field as follows:
1549 __PACKAGE__->add_columns(description => { accessor => '_description' });
1551 Next, we'll define the accessor-wrapper subroutine:
1556 # If there is an update to the column, we'll let the original accessor
1558 return $self->_description(@_) if @_;
1560 # Fetch the column value.
1561 my $description = $self->_description;
1563 # If there's something in the description field, then just return that.
1564 return $description if defined $description && length $descripton;
1566 # Otherwise, generate a description.
1567 return $self->generate_description;
1570 =head1 DEBUGGING AND PROFILING
1572 =head2 DBIx::Class objects with Data::Dumper
1574 L<Data::Dumper> can be a very useful tool for debugging, but sometimes it can
1575 be hard to find the pertinent data in all the data it can generate.
1576 Specifically, if one naively tries to use it like so,
1580 my $cd = $schema->resultset('CD')->find(1);
1583 several pages worth of data from the CD object's schema and result source will
1584 be dumped to the screen. Since usually one is only interested in a few column
1585 values of the object, this is not very helpful.
1587 Luckily, it is possible to modify the data before L<Data::Dumper> outputs
1588 it. Simply define a hook that L<Data::Dumper> will call on the object before
1589 dumping it. For example,
1596 result_source => undef,
1604 local $Data::Dumper::Freezer = '_dumper_hook';
1606 my $cd = $schema->resultset('CD')->find(1);
1608 # dumps $cd without its ResultSource
1610 If the structure of your schema is such that there is a common base class for
1611 all your table classes, simply put a method similar to C<_dumper_hook> in the
1612 base class and set C<$Data::Dumper::Freezer> to its name and L<Data::Dumper>
1613 will automagically clean up your data before printing it. See
1614 L<Data::Dumper/EXAMPLES> for more information.
1618 When you enable L<DBIx::Class::Storage>'s debugging it prints the SQL
1619 executed as well as notifications of query completion and transaction
1620 begin/commit. If you'd like to profile the SQL you can subclass the
1621 L<DBIx::Class::Storage::Statistics> class and write your own profiling
1624 package My::Profiler;
1627 use base 'DBIx::Class::Storage::Statistics';
1629 use Time::HiRes qw(time);
1638 $self->print("Executing $sql: ".join(', ', @params)."\n");
1647 my $elapsed = sprintf("%0.4f", time() - $start);
1648 $self->print("Execution took $elapsed seconds.\n");
1654 You can then install that class as the debugging object:
1656 __PACKAGE__->storage->debugobj(new My::Profiler());
1657 __PACKAGE__->storage->debug(1);
1659 A more complicated example might involve storing each execution of SQL in an
1667 my $elapsed = time() - $start;
1668 push(@{ $calls{$sql} }, {
1674 You could then create average, high and low execution times for an SQL
1675 statement and dig down to see if certain parameters cause aberrant behavior.
1676 You might want to check out L<DBIx::Class::QueryLog> as well.
1678 =head1 STARTUP SPEED
1680 L<DBIx::Class|DBIx::Class> programs can have a significant startup delay
1681 as the ORM loads all the relevant classes. This section examines
1682 techniques for reducing the startup delay.
1684 These tips are are listed in order of decreasing effectiveness - so the
1685 first tip, if applicable, should have the greatest effect on your
1688 =head2 Statically Define Your Schema
1691 L<DBIx::Class::Schema::Loader|DBIx::Class::Schema::Loader> to build the
1692 classes dynamically based on the database schema then there will be a
1693 significant startup delay.
1695 For production use a statically defined schema (which can be generated
1696 using L<DBIx::Class::Schema::Loader|DBIx::Class::Schema::Loader> to dump
1697 the database schema once - see
1698 L<make_schema_at|DBIx::Class::Schema::Loader/make_schema_at> and
1699 L<dump_directory|DBIx::Class::Schema::Loader/dump_directory> for more
1700 details on creating static schemas from a database).
1702 =head2 Move Common Startup into a Base Class
1704 Typically L<DBIx::Class> result classes start off with
1706 use base qw/DBIx::Class/;
1707 __PACKAGE__->load_components(qw/InflateColumn::DateTime Core/);
1709 If this preamble is moved into a common base class:-
1713 use base qw/DBIx::Class/;
1714 __PACKAGE__->load_components(qw/InflateColumn::DateTime Core/);
1717 and each result class then uses this as a base:-
1719 use base qw/MyDBICbase/;
1721 then the load_components is only performed once, which can result in a
1722 considerable startup speedup for schemas with many classes.
1724 =head2 Explicitly List Schema Result Classes
1726 The schema class will normally contain
1728 __PACKAGE__->load_classes();
1730 to load the result classes. This will use L<Module::Find|Module::Find>
1731 to find and load the appropriate modules. Explicitly defining the
1732 classes you wish to load will remove the overhead of
1733 L<Module::Find|Module::Find> and the related directory operations:-
1735 __PACKAGE__->load_classes(qw/ CD Artist Track /);
1737 If you are instead using the L<load_namespaces|DBIx::Class::Schema/load_namespaces>
1738 syntax to load the appropriate classes there is not a direct alternative
1739 avoiding L<Module::Find|Module::Find>.
1743 =head2 Cached statements
1745 L<DBIx::Class> normally caches all statements with L<< prepare_cached()|DBI/prepare_cached >>.
1746 This is normally a good idea, but if too many statements are cached, the database may use too much
1747 memory and may eventually run out and fail entirely. If you suspect this may be the case, you may want
1748 to examine DBI's L<< CachedKids|DBI/CachedKidsCachedKids_(hash_ref) >> hash:
1750 # print all currently cached prepared statements
1751 print for keys %{$schema->storage->dbh->{CachedKids}};
1752 # get a count of currently cached prepared statements
1753 my $count = scalar keys %{$schema->storage->dbh->{CachedKids}};
1755 If it's appropriate, you can simply clear these statements, automatically deallocating them in the
1758 my $kids = $schema->storage->dbh->{CachedKids};
1759 delete @{$kids}{keys %$kids} if scalar keys %$kids > 100;
1761 But what you probably want is to expire unused statements and not those that are used frequently.
1762 You can accomplish this with L<Tie::Cache> or L<Tie::Cache::LRU>:
1766 my $schema = DB::Main->connect($dbi_dsn, $user, $pass, {
1767 on_connect_do => sub { tie %{shift->_dbh->{CachedKids}}, 'Tie::Cache', 100 },