Skip to content

Schema Reflection

AltibaseDialect implements reflection queries against Altibase system catalogs (SYSTEM_.*).

Reflection flow

mermaid flowchart TD A[Inspector / MetaData.reflect] --> B[_effective_schema(schema)] B --> C[get_table_names / get_view_names] C --> D[get_columns] D --> E[get_pk_constraint] E --> F[get_foreign_keys] F --> G[get_indexes] G --> H[get_table_comment]

Catalog and schema handling

  • User-facing schema names are normalized to uppercase via _effective_schema().
  • If no schema is passed, dialect queries default schema via:
  • SELECT USER_NAME() FROM DUAL
  • Reflection SQL joins SYSTEM_.SYS_USERS_ with object catalogs (SYS_TABLES_, SYS_COLUMNS_, etc.).

Table and view reflection

  • get_table_names() returns tables where TABLE_TYPE = 'T'
  • get_view_names() returns views where TABLE_TYPE = 'V'
  • get_view_definition() reads SYSTEM_.SYS_VIEWS_.VIEW_TEXT

Column reflection

get_columns() returns SQLAlchemy column dictionaries including:

  • name
  • type (resolved by _resolve_column_type())
  • nullable
  • default
  • autoincrement (true when reflected type is SERIAL)

Type resolution supports both textual types and integer type codes, with fallback to NullType and warning on unknown types.

Column defaults are normalized by _normalize_default() which strips vendor-specific quoting (e.g., 'value' → value, NULL → None).

Primary key reflection

get_pk_constraint() reads SYS_CONSTRAINTS_ (CONSTRAINT_TYPE = 3) and ordered columns from SYS_CONSTRAINT_COLUMNS_.

Return shape:

{"name": "PK_TABLE", "constrained_columns": ["ID"]}

Foreign key reflection

get_foreign_keys():

  1. Collects FK constraints (CONSTRAINT_TYPE = 0)
  2. Collects constrained columns in order
  3. Resolves referred table/schema by REFERENCED_TABLE_ID
  4. Resolves referred columns from REFERENCED_INDEX_ID

Return shape:

{
    "name": "FK_ORDERS_USERS",
    "constrained_columns": ["USER_ID"],
    "referred_schema": "APP",
    "referred_table": "USERS",
    "referred_columns": ["ID"],
}

Index reflection

get_indexes() returns index dictionaries with:

  • name
  • column_names
  • unique

Uniqueness is derived from IS_UNIQUE values such as Y, 1, TRUE, UNIQUE.

Unique/check/sequence reflection notes

The current dialect exposes:

  • has_sequence(sequence_name, schema=None) for existence checks
  • has_index(...) for index existence

It does not currently implement dedicated public reflection methods like:

  • get_unique_constraints()
  • get_check_constraints()
  • get_sequence_names()

In practice, unique index metadata is available from get_indexes() (unique=True).

Practical reflection example

from sqlalchemy import create_engine, inspect

engine = create_engine("altibase://sys:password@localhost:20300/mydb")
insp = inspect(engine)

print(insp.get_schema_names())
print(insp.get_table_names(schema="APP"))
print(insp.get_columns("USERS", schema="APP"))
print(insp.get_pk_constraint("USERS", schema="APP"))
print(insp.get_foreign_keys("ORDERS", schema="APP"))
print(insp.get_indexes("USERS", schema="APP"))