Dari supports MySQL, PostgreSQL, and Oracle, all of which are implemented using the same underlying table structure and indexing strategy. Dari’s SQL database schema is implemented using multiple tables (see below). The primary table, Record, stores the raw information about the objects that are saved. Objects are serialized to JSON format when they are saved and stored in the data column of the Record table. Several other tables are used to implement field indexes.
Dari creates unique IDs for every object by generating a UUID. The id and typeId fields in the Record table store these UUIDs in the most appropriate datatype for the underlying database. For MySQL and Oracle, this is binary(16). On PostgreSQL, the native UUID is used.
Because SQL databases generally provide fast primary key lookups, Brightspot projects typically use an SQL implementation as the primary database. However, SQL database performance is limited for text matching and compound-predicate queries. Therefore, Brightspot projects generally supplement the SQL implementation with the Solr database, which Dari uses for full-text searches.
The following table lists the tables that Dari creates in SQL-based databases. An integer in the table name indicates a version, which may be different in your database.
|Record||Primary storage table. All objects are stored in this table as serialized JSON blobs.|
|RecordLocation3||Stores spatial indexes. Supported on MySQL and PostgreSQL only.|
|RecordNumber3||Stores number and time stamp indexes.|
|RecordRegion2||Stores indexes for localized data.|
|RecordSearch||Tracks indexed fields and methods.|
|RecordString4||Stores string and enumeration indexes.|
|RecordUpdate||Tracks when objects were last updated.|
|RecordUuid3||Stores relationship indexes.|
|Symbol||Stores symbols such as index names. It is referenced by the other index tables.|