In order to generate fake data you need to provide schema description in yaml format. The structure of the configuration file relects the structure of relational database. Such structure allows the package to detect inner and inter-table dependencies, more to that, makes the simulated data preserving original database assumptions.
The schema can be automatically sourced from a database (see: Structure from DB) or you may configure such on your own.
Such configuration file should have the below structure:
public - schema name
└── tables: - tables list
├── table_a: - name of the table
│ ├── ... - additional table-wise parameters
│ ├── check_constraints: - list of table check constraints
│ │ └── constraint_name - unique name of the constraint
│ │ ├── column: column_a2 - column attached to the constraint (can be empty)
│ │ └── expression: !expr column_a2 == column_a1 - R expression describing constraint
│ ├── columns: - list of table columns
│ │ ├── column_a1: - name of the column
│ │ │ ├── type: char(8) - column type (obligatory, valid R class or sql type)
│ │ │ ├── not_null: true |
│ │ │ ├── unique: true | standard column parameters (optional)
│ │ │ └── ... - extra column parameters
│ │ └── column_a2:
│ │ └── type: numeric(4, 2)
│ └── primary_key: - list of primary keys
│ └── pk_name: - primary key unique name
│ └── columns: - array of primary key columns
│ └── - column_a1 - column name treated as primary key
└── table_b:
├── columns:
│ ├── column_b1:
│ │ └── type: char(8)
│ └── column_b2:
│ └── type: boolean
└── foreign_keys: - list of foreign keys
└── fk_name: - unique foreign key name
├── columns: - array of foreign key columns
│ └── - column_b1 - name of the table column beeing foreign key
└── references: - definition of foregin key reference
├── columns: - array of foreign key dependent columns
│ └── - column_a1 - name of dependent column
└── table: table_a - name of dependent foreign key table
Schema assumptions and limitations (possibly reduced in the future releases):
pk_name > columns list must
contain only a single column.fk_name > columns and
fk_name > references > columns lists must contain
only a single value.