criteria

Can be a hash, hashref, or an arrayref. Takes the same options as the SQL::Abstract "where" method. If values contain any objects, they will be deflated before querying the database.

attributes

case, cmp, convert, and logic

These attributes are passed to SQL::Abstract's constuctor and alter the behavior of the criteria.

    { cmp => 'like' }
    

order_by

Specifies the sort order of the results.

    { order_by => 'created_on DESC' }
    

rows

Specifies the maximum number of rows to return. Currently supported RDBMs are Interbase, MaxDB, MySQL, PostgreSQL and SQLite. For other RDBMs, it will be emulated.

    { rows => 10 }
    

offset

Specifies the offset of the first row to return. Defaults to 0 if unspecified.

    { offset => 0 }
    

page

Specifies the current page in "page". Defaults to 1 if unspecified.

    { page => 1 }
    

prefetch

Specifies a listref of relationships to prefetch. These must be has_a or might_haves or Sweet will throw an error. This will cause Sweet to do a join across to the related tables in order to return the related object without a second trip to the database. All 'Essential' columns of the foreign table are retrieved.

    { prefetch => [ qw/some_rel some_other_rel/ ] }
    

Sweet constructs the joined SQL statement by aliasing the columns in each table and prefixing the column name with 'sweet__N_' where N is a counter starting at 1. Note that if your database has a column length limit (for example, Oracle's limit is 30) and you use long column names in your application, Sweet's addition of at least 9 extra characters to your column name may cause database errors.

use_resultset_cache

Enables the resultset cache. This is a little experimental and massive gotchas may rear their ugly head at some stage, but it does seem to work pretty well.

For best results, the resultset cache should only be used selectively on queries where you experience performance problems. Enabling it for every single query in your application will most likely cause a drop in performance as the cache overhead is greater than simply fetching the data from the database.

profile_cache

Records cache hits/misses and what keys they were for in ->profiling_data. Note that this is class metadata so if you don't want it to be global for Sweet you need to do

    __PACKAGE__->profiling_data({ });
    

in either your base class or your table classes to taste.

disable_sql_paging

Disables the use of paging in SQL statements if set, forcing Sweet to emulate paging by slicing the iterator at the end of ->search (which it normally only uses as a fallback mechanism). Useful for testing or for causing the entire query to be retrieved initially when the resultset cache is used.

This is also useful when using custom SQL via "set_sql" and setting "sql_method" (see below) where a COUNT(*) may not make sense (i.e. when the COUNT(*) might be as expensive as just running the full query and just slicing the iterator).

sql_method

This sets the name of the sql fragment to use as previously set by a "set_sql" call. The default name is "Join_Retrieve" and the associated default sql fragment set in this class is:

    __PACKAGE__->set_sql( Join_Retrieve => <<'SQL' );
    SELECT __ESSENTIAL(me)__%s
    FROM   %s
    WHERE  %s
    SQL
    

You may override this in your table or base class using the same name and CDBI::Sweet will use your custom fragment, instead.

If you need to use more than one sql fragment in a given class you may create a new sql fragment and then specify its name using the "sql_method" attribute.

The %s strings are replaced by sql parts as described in Ima::DBI. See "statement_order" for the sql part that replaces each instance of %s.

In addition, the associated statment for COUNT(*) statement has "_Count" appended to the sql_method name. Only "from" and "where" are passed to the sprintf function.

The default sql fragment used for "Join_Retrieve" is:

    __PACKAGE__->set_sql( Join_Retrieve_Count => <<'SQL' );
    SELECT COUNT(*)
    FROM   %s
    WHERE  %s
    SQL
    

If you create a custom sql method (and set the "sql_method" attribute) then you will likely need to also create an associated _Count fragment. If you do not have an associated _Count, and wish to call the "page" method, then set "disable_sql_paging" to true and your result set from the select will be spliced to return the page you request.

Here's an example.

Assume a CD has_a Artist (and thus Artists have_many CDs), and you wish to return a list of artists and how many CDs each have:

In package MyDB::Artist

    __PACKAGE__->columns( TEMP => 'cd_count');
    __PACKAGE__->set_sql( 'count_by_cd', <<'');
        SELECT      __ESSENTIAL(me)__, COUNT(cds.cdid) as cd_count
        FROM        %s                  -- ("from")
        WHERE       %s                  -- ("where")
        GROUP BY    __ESSENTIAL(me)__
        %s %s                           -- ("limit" and "order_by")
    

Then in your application code:

    my ($pager, $iterator) = MyDB::Artist->page(
        {
            'cds.title'    => { '!=', undef },
        },
        {
            sql_method          => 'count_by_cd',
            statement_order     => [qw/ from where limit order_by / ],
            disable_sql_paging  => 1,
            order_by            => 'cd_count desc',
            rows                => 10,
            page                => 1,
        } );
    

The above generates the following SQL:

    SELECT      me.artistid, me.name, COUNT(cds.cdid) as cd_count
    FROM        artist me, cd cds
    WHERE       ( cds.title IS NOT NULL ) AND me.artistid = cds.artist
    GROUP BY    me.artistid, me.name
    ORDER BY    cd_count desc
    

The one caveat is that Sweet cannot figure out the has_many joins unless you specify them in the $criteria. In the previous example that's done by asking for all cd titles that are not null (which should be all).

To fetch a list like above but limited to cds that were created before the year 2000, you might do:

    my ($pager, $iterator) = MyDB::Artist->page(
        {
            'cds.year'  => { '<', 2000 },
        },
        {
            sql_method          => 'count_by_cd',
            statement_order     => [qw/ from where limit order_by / ],
            disable_sql_paging  => 1,
            order_by            => 'cd_count desc',
            rows                => 10,
            page                => 1,
        } );
    

statement_order

Specifies a list reference of SQL parts that are replaced in the SQL fragment (which is defined with "sql_method" above). The available SQL parts are:

    prefetch_cols from where order_by limit sql prefetch_names
    

The "sql" part is shortcut notation for these three combined:

    where order_by limit
    

Prefecch_cols are the columns selected when a prefetch is speccified -- use in the SELECT. Prefetch_names are just the column names for use in GROUP BY.

This is useful when statement order needs to be changed, such as when using a GROUP BY:

count

Returns a count of the number of rows matching the criteria. "count" will discard "offset", "order_by", and "rows".

    $count = MyApp::Article->count(%criteria);

search

Returns an iterator in scalar context, or an array of objects in list context.

    @objects  = MyApp::Article->search(%criteria);
    $iterator = MyApp::Article->search(%criteria);

search_like

As search but adds the attribute { cmp => 'like' }.

page

Retuns a page object and an iterator. The page object is an instance of Data::Page.

    ( $page, $iterator )
        = MyApp::Article->page( $criteria, { rows => 10, page => 2 );
    printf( "Results %d - %d of %d Found\n",
        $page->first, $page->last, $page->total_entries );

pager

An alias to page.

retrieve_all

Same as "Class::DBI" with addition that it takes "attributes" as arguments, "attributes" can be a hash or a hashref.

    $iterator = MyApp::Article->retrieve_all( order_by => 'created_on' );

retrieve_next

Returns the next record after the current one according to the order_by attribute (or primary key if no order_by specified) matching the criteria. Must be called as an object method.

retrieve_previous

As retrieve_next but retrieves the previous record.

cache

Class method: if this is set caching is enabled. Any cache object that has a "get", "set", and "remove" method is supported.

    __PACKAGE__->cache(
        Cache::FastMmap->new(
            share_file => '/tmp/cdbi',
            expire_time => 3600
        )
    );

cache_key

Returns a cache key for an object consisting of class and primary keys.

Overloaded methods

_init

Overrides "Class::DBI"'s internal cache. On a cache hit, it will return a cached object; on a cache miss it will create an new object and store it in the cache.

create

insert

All caches for this table are marked stale and will be re-cached on next retrieval. create is an alias kept for backwards compability.

retrieve

On a cache hit the object will be inflated by the "select" trigger and then served.

update

Object is removed from the cache and will be cached on next retrieval.

delete

Object is removed from the cache.