PLAN_TABLE_DATA

PLAN_TABLE_DATA stores plan information collected by EXPLAIN PLAN. Different from the PLAN_TABLE view, the system catalog PLAN_TABLE_DATA stores EXPLAIN PLAN information collected by all sessions and users.

**Table 1** PLAN_TABLE_DATA columns
Name	Type	Description
session_id	text	Session that inserts the data. Its value consists of a service thread start timestamp and a service thread ID. Values are constrained by NOT NULL.
user_id	oid	User who inserts the data. Values are constrained by NOT NULL.
statement_id	character varying(30)	Query tag specified by a user
plan_id	bigint	Query plan ID. The ID is automatically generated in the plan generation phase and is used by kernel engineers for debugging.
id	integer	Node ID in a plan
operation	character varying(30)	Operation description
options	character varying(255)	Operation action
object_name	name	Name of an operated object. It is defined by users.
object_type	character varying(30)	Object type
object_owner	name	Schema to which an object belongs. It is defined by users.
projection	character varying(4000)	Returned column information
cost	double precision	Execution cost estimated by the optimizer for an operator
cardinality	double precision	Number of rows estimated by the optimizer for an operator

PLAN_TABLE_DATA records data of all users and sessions on the current node. Only administrators can access all the data. Common users can view their own data in the PLAN_TABLE view.
Data of inactive (exited) sessions is cleaned from PLAN_TABLE_DATA by gs_clean after being stored in this system catalog for a certain period of time (5 minutes by default). You can also manually run gs_clean -C to delete inactive session data from the table.
Data is automatically inserted into PLAN_TABLE_DATA after EXPLAIN PLAN is executed. Therefore, do not manually insert data into or update data in PLAN_TABLE_DATA. Otherwise, data in PLAN_TABLE_DATA may be disordered. When you need to delete data from a table, it is recommended that you use the PLAN_TABLE view.
Information in the statement_id, object_name, object_owner, and projection columns is stored in letter cases specified by users and information in other columns is stored in uppercase.