CREATE ROW LEVEL SECURITY POLICY¶
Function¶
Creates a row-level access control policy for a table.
The policy takes effect only after row-level access control is enabled (by running ALTER TABLE... ENABLE ROW LEVEL SECURITY).
Currently, row-level access control affects the read (SELECT, UPDATE, DELETE) of data tables and does not affect the write (INSERT and MERGE INTO) of data tables. The table owner or system administrators can create an expression in the USING clause. When the client reads the data table, the database server combines the expressions that meet the condition and applies it to the execution plan in the statement rewriting phase of a query. For each tuple in a data table, if the expression returns TRUE, the tuple is visible to the current user; if the expression returns FALSE or NULL, the tuple is invisible to the current user.
A row-level access control policy name is specific to a table. A data table cannot have row-level access control policies with the same name. Different data tables can have the same row-level access control policy.
Row-level access control policies can be applied to specified operations (SELECT, UPDATE, DELETE, and ALL). ALL indicates that SELECT, UPDATE, and DELETE will be affected. For a new row-level access control policy, the default value ALL will be used if you do not specify the operations that will be affected.
Row-level access control policies can be applied to a specified user (role) or to all users (PUBLIC). For a new row-level access control policy, the default value PUBLIC will be used if you do not specify the user that will be affected.
Precautions¶
Row-level access control policies can be defined for row-store tables, row-store partitioned tables, column-store tables, column-store partitioned tables, replication tables, unlogged tables, and hash tables.
Row-level access control policies cannot be defined for HDFS tables, foreign tables, and temporary tables.
Row-level access control policies cannot be defined for views.
A maximum of 100 row-level access control policies cannot be defined for a table.
System administrators and table owners are not affected by row-level access control and can view full data of tables.
Tables queried by using SQL statements, views, functions, and stored procedures are affected by row-level access control policies.
The type of a column on which a row-level access control policy depends cannot be changed. For example, the following modifications are not supported:
ALTER TABLE public.all_data ALTER COLUMN role TYPE text;
Syntax¶
CREATE [ ROW LEVEL SECURITY ] POLICY policy_name ON table_name
[ AS { PERMISSIVE | RESTRICTIVE } ]
[ FOR { ALL | SELECT | UPDATE | DELETE } ]
[ TO { role_name | PUBLIC } [, ...] ]
USING ( using_expression )
Parameter Description¶
policy_name
Specifies the name of a row-level access control policy to be created. The names of row-level access control policies for a table must be unique.
table_name
Specifies the name of a table to which a row-level access control policy is applied.
PERMISSIVE
Specifies that the row-level access control policy is to be created as a permissive policy. For a given query, all applicable permissive policies are combined using the OR operator. Row-level access control policies are permissive by default.
RESTRICTIVE
Specifies that the row-level access control policy is to be created as a restrictive policy. For a given query, all applicable restrictive policies are combined using the AND operator.
Important
At least one permissive policy is required to grant access to data records. If only restrictive policies are used, no records will be accessible. When both permissive and restrictive policies are used, a record is accessible only when it passes at least one permissive policy and all restrictive policies.
command
Specifies the SQL operations affected by a row-level access control policy, including ALL, SELECT, UPDATE, and DELETE. If this parameter is not specified, the default value ALL will be used, covering SELECT, UPDATE, and DELETE.
If command is set to SELECT, only tuple data that meets the condition (the return value of using_expression is TRUE) can be queried. The operations that are affected include SELECT, UPDATE.... RETURNING, and DELETE... RETURNING.
If command is set to UPDATE, only tuple data that meets the condition (the return value of using_expression is TRUE) can be updated. The operations that are affected include UPDATE, UPDATE ... RETURNING, and SELECT ... FOR UPDATE/SHARE.
If command is set to DELETE, only tuple data that meets the condition (the return value of using_expression is TRUE) can be deleted. The operations that are affected include DELETE and DELETE ... RETURNING.
The following table describes the relationship between row-level access control policies and SQL statements.
Table 1 Relationship between row-level security policies and SQL statements¶ Command
SELECT/ALL Policy
UPDATE/ALL Policy
DELETE/ALL Policy
SELECT
Existing row
No
No
SELECT FOR UPDATE/SHARE
Existing row
Existing row
No
UPDATE
No
Existing row
No
UPDATE RETURNING
Existing row
Existing row
No
DELETE
No
No
Existing row
DELETE RETURNING
Existing row
No
Existing row
role_name
Specifies database users affected by a row-level access control policy.
If this parameter is not specified, the default value PUBLIC will be used, indicating that all database users will be affected. You can specify multiple affected database users.
Important
System administrators are not affected by row access control.
using_expression
Specifies an expression defined for a row-level access control policy (return type: boolean).
The expression cannot contain aggregate functions and window functions. In the statement rewriting phase of a query, if row-level access control for a data table is enabled, the expressions that meet the specified conditions will be added to the plan tree. The expression is calculated for each tuple in the data table. For SELECT, UPDATE, and DELETE, row data is visible to the current user only when the return value of the expression is TRUE. If the expression returns FALSE, the tuple is invisible to the current user. In this case, the user cannot view the tuple through the SELECT statement, update the tuple through the UPDATE statement, or delete the tuple through the DELETE statement.
Example 1: Create a Row-level Access Control Policy That the Current User Can Only View Its Own Data.¶
Create users alice and bob.
CREATE ROLE alice PASSWORD '{password}'; CREATE ROLE bob PASSWORD '{password}';
Create data table public.all_data:
CREATE TABLE public.all_data(id int, role varchar(100), data varchar(100));
Insert data into the data table:
INSERT INTO all_data VALUES(1, 'alice', 'alice data'); INSERT INTO all_data VALUES(2, 'bob', 'bob data'); INSERT INTO all_data VALUES(3, 'peter', 'peter data');
Grant the read permission for the all_data table to users alice and bob:
GRANT SELECT ON all_data TO alice, bob;
Enable row-level access control.
ALTER TABLE all_data ENABLE ROW LEVEL SECURITY;
Create a row-level access control policy to specify that the current user can view only their own data:
CREATE ROW LEVEL SECURITY POLICY all_data_rls ON all_data USING(role = CURRENT_USER);
View information about the all_data table.
\d+ all_data Table "public.all_data" Column | Type | Modifiers | Storage | Stats target | Description --------+------------------------+-----------+----------+--------------+------------- id | integer | | plain | | role | character varying(100) | | extended | | data | character varying(100) | | extended | | Row Level Security Policies: POLICY "all_data_rls" USING (((role)::name = "current_user"())) Has OIDs: no Distribute By: HASH(id) Location Nodes: ALL DATANODES Options: orientation=row, compression=no, enable_rowsecurity=true
Run SELECT.
SELECT * FROM all_data; id | role | data ----+-------+------------ 1 | alice | alice data 2 | bob | bob data 3 | peter | peter data (3 rows) EXPLAIN(COSTS OFF) SELECT * FROM all_data; QUERY PLAN ---------------------------- Streaming (type: GATHER) Node/s: All datanodes -> Seq Scan on all_data (3 rows)
Switch to the alice user.
set role alice password '{password}';
Perform the SELECT operation.
SELECT * FROM all_data; id | role | data ----+-------+------------ 1 | alice | alice data (1 row) EXPLAIN(COSTS OFF) SELECT * FROM all_data; QUERY PLAN ---------------------------------------------------------------- Streaming (type: GATHER) Node/s: All datanodes -> Seq Scan on all_data Filter: ((role)::name = 'alice'::name) Notice: This query is influenced by row level security feature (5 rows)
Example 2: Partition Permission Management Through Row-Level Control¶
Create user alice.
CREATE ROLE alice PASSWORD '{password1}';
Create range partitioned table web_returns_p1, and insert data into the table.
CREATE TABLE web_returns_p1 ( wr_returned_date_sk integer, wr_returned_time_sk integer, wr_item_sk integer NOT NULL, wr_refunded_customer_sk integer ) WITH (orientation = column) DISTRIBUTE BY HASH (wr_item_sk) PARTITION BY RANGE(wr_returned_date_sk) ( PARTITION p2016 START(800) END(830) EVERY(1) ); INSERT INTO web_returns_p1 values (801,17,11,102); INSERT INTO web_returns_p1 values (802,18,12,103);
Grant the read permission on the web_returns_p1 table to user alice.
GRANT SELECT ON web_returns_p1 TO alice;
Enable row-level access control.
ALTER TABLE web_returns_p1 ENABLE ROW LEVEL SECURITY;
Create row-level access control policy web_returns_rsl. In the command, wr_returned_date_sk is a partition name of the web_returns_p1 partition table, and 801 is the partition value.
CREATE ROW LEVEL SECURITY POLICY web_returns_rsl ON web_returns_p1 USING('wr_returned_date_sk' = '801');
Impose the row-level access control policy web_returns_rsl on user alice.
ALTER ROW LEVEL SECURITY POLICY web_returns_rsl ON web_returns_p1 TO alice;
Switch to the alice user.
set role alice password '{password1}';
Query the web_returns_p1 table.
select * from web_returns_p1;