Extended Statistics¶

Requires the dbwarden-pgsql-extensions plugin: dbwarden plugin add dbwarden-pgsql-extensions. The extended_statistics object handler ships in that plugin, not in core.

Handler: ExtendedStatisticsHandler (PREAMBLE phase, config-driven, PG 14+)

Extended statistics give the query planner better estimates for correlated columns.

pg_extended_statistics=[
    {
        "name": "stats_users_email_city",
        "table": "users",
        "kinds": ["d", "f"],
        "columns": "email, city",
    },
]

Kind Codes¶

Lifecycle¶

ALTER STATISTICS¶

The statistics target controls sample size. Higher values produce better estimates at the cost of longer ANALYZE time:

ALTER STATISTICS stats_users_email_city SET STATISTICS 1000;

Values range from -1 (use system default, typically 100) to 10000.

ANALYZE Requirements¶

CREATE STATISTICS defines the statistics object but does not populate it. Run ANALYZE on the table against the database directly to collect the first statistics sample:

ANALYZE users;

Examples¶

# ndistinct + dependencies on correlated columns
{"name": "stats_order_date_status", "table": "orders",
 "kinds": ["d", "f"], "columns": "order_date, status"}

# MCV on a high-cardinality group
{"name": "stats_product_category", "table": "products",
 "kinds": ["m"], "columns": "category_id, price"}

# Expression statistics (PG 14+)
{"name": "stats_user_email_domain", "table": "users",
 "kinds": ["d", "m"], "expressions": ["lower(email)"]}

Generated DDL:

CREATE STATISTICS stats_users_email_city (ndistinct, dependencies) ON email, city FROM users;

Schema Support¶

Extended statistics can be scoped to a schema:

{"name": "stats_order_date_status", "table": "orders",
 "schema": "app", "kinds": ["d", "f"], "columns": "order_date, status"}

Migration Safety¶

See Migration Safety for the full classification table.