Skip to main content

Validation

CSDB readers and writers validate both the file structure and relational model.

Structure​

  • The first section must be --- csdb.
  • Unknown section headers are rejected.
  • Human table sections must be grouped as schema then data.
  • Machine sections must appear after all human sections.
  • Table and database names must use letters, numbers, _, -, ., or $.

Metadata​

  • format must be CSDB.
  • version must be 1.
  • The metadata tables list must match table section order.
  • Duplicate table names are rejected.

Schema​

  • Every table schema needs name and ordered columns.
  • Column names must be valid names.
  • Primary key columns must exist and be listed in required.
  • Required fields, defaults, unique constraints, foreign keys, and indexes must reference known columns.
  • Unsupported runtime types are rejected unless represented as supported custom types.

Data​

  • CSV headers must match schema order exactly.
  • Optional columns may contain null values unless they are part of the primary key.
  • Required columns cannot be null.
  • Primary key columns cannot be null.
  • Values must decode according to their column type.
  • Primary keys, unique constraints, and unique indexes cannot duplicate non-null values.
  • Check constraints must evaluate truthy for each row.
  • Foreign keys must reference existing target rows unless their source values are null.