Version 4 supported
This version of Silverstripe CMS is still supported though will not receive any additional features. Go to documentation for the most recent stable version.

Indexes

Indexes are a great way to improve performance in your application, especially as it grows. By adding indexes to your data model you can reduce the time taken for the framework to find and filter data objects.

The addition of an indexes should be carefully evaluated as they can also increase the cost of other operations such as UPDATE/INSERT and DELETE. An index on a column whose data is non unique will actually cost you performance. For example, in most cases an index on boolean status flag, or ENUM state will not increase query performance.

It's important to find the right balance to achieve fast queries using the optimal set of indexes; For Silverstripe CMS applications it's a good practice to:

  • add indexes on columns which are frequently used in filter, where or orderBy statements
  • for these, only include indexes for columns which are the most restrictive (return the least number of rows)

The Silverstripe CMS framework already places certain indexes for you by default:

  • The primary key for each model has a PRIMARY KEY unique index
  • The ClassName column if your model inherits from DataObject
  • All relationships defined in the model have indexes for their has_one entity (for many_many relationships this index is present on the associative entity).

Defining an index

Indexes are represented on a DataObject through the DataObject::$indexes array which maps index names to a descriptor. There are several supported notations:

namespace App\Model;

use SilverStripe\ORM\DataObject;

class MyObject extends DataObject
{
    private static $indexes = [
        '<column-name>' => true,
        '<index-name>' => [
            'type' => '<type>',
            'columns' => ['<column-name>', '<other-column-name>'],
        ],
        '<index-name>' => ['<column-name>', '<other-column-name>'],
    ];
}

The <column-name> is used to put a standard non-unique index on the column specified. For complex or large tables we recommend building the index to suite the requirements of your data.

The <index-name> can be an arbitrary identifier in order to allow for more than one index on a specific database column. The "advanced" notation supports more <type> notations. These vary between database drivers, but all of them support the following:

  • index: Standard non unique index.
  • unique: Index plus uniqueness constraint on the value
  • fulltext: Fulltext content index
// app/src/MyTestObject.php
namespace App\Model;

use SilverStripe\ORM\DataObject;

class MyTestObject extends DataObject
{
    private static $db = [
        'MyField' => 'Varchar',
        'MyOtherField' => 'Varchar',
    ];

    private static $indexes = [
        'MyIndexName' => ['MyField', 'MyOtherField'],
    ];
}

Please note that if you have previously used the removed value key to define an index's contents, Silverstripe CMS will now throw an error. Use columns instead.

Complex/Composite indexes

For complex queries it may be necessary to define a complex or composite index on the supporting object. To create a composite index, define the fields in the index order as a comma separated list.

Note Most databases only use the leftmost prefix to optimise the query, try to ensure the order of the index and your query parameters are the same. For example

  • index (col1) - WHERE col1 = ?
  • index (col1, col2) = WHERE (col1 = ? AND col2 = ?)
  • index (col1, col2, col3) = WHERE (col1 = ? AND col2 = ? AND col3 = ?)

The index would not be used for a query WHERE col2 = ? or for WHERE col1 = ? OR col2 = ?

As an alternative to a composite index, you can also create a hashed column which is a combination of information from other columns. If this is indexed, smaller and reasonably unique it might be faster that an index on the whole column.

Index creation/destruction

Indexes are generated and removed automatically during a dev/build. Caution if you're working with large tables and modify an index as the next dev/build will DROP the index, and then ADD it.

As of 3.7.0 default_sort fields will automatically become database indexes as this provides significant performance benefits.

API documentation