The motivating factor for writing this study was an article . An interesting article published 4 years ago also caught my attention — .
What was interesting is that the same idea-"to implement business logic in the database".
came to my mind not only to me.
I also wanted to preserve, primarily for myself, interesting developments that arose during the implementation. Especially considering that not long ago a strategic decision was made to change the architecture and move the business logic to the backend level. So, everything that was developed will soon not be needed by anyone and will not be of interest to anyone.
The described methods are not any kind of discovery or exceptional know how, everything follows classical approaches and has been implemented repeatedly (I applied a similar approach 20 years ago on Oracle, for example). I just decided to gather everything in one place. Maybe it will be useful for someone. As practice has shown, quite often the same idea comes to different people independently. And it is useful to leave it as a memory for myself.
Of course, nothing in this world is perfect; errors and typos are unfortunately possible. Criticism and comments are welcome and expected. And one more small detail — specific implementation details are omitted. After all, everything is still used in a functioning project. So, the article serves as a study and a description of the general concept, nothing more. I hope the details are sufficient for understanding the overall picture.
The general idea is "divide and conquer, hide and own"
The idea is classic — a separate schema for tables, a separate schema for stored functions.
The client does not have direct access to the data. All the client can do is call a stored function and process the received response.
Roles
CREATE ROLE store;
CREATE ROLE sys_functions;
CREATE ROLE loc_audit_functions;
CREATE ROLE service_functions;
CREATE ROLE business_functions;
Schemas
Schema for storing tables
Target tables that implement subject entities.
CREATE SCHEMA store AUTHORIZATION store;
Schema for system functions
System functions, particularly for logging table changes.
CREATE SCHEMA sys_functions AUTHORIZATION sys_functions;
Schema for local audit
Functions and tables for implementing local auditing of stored function execution and target table modifications.
CREATE SCHEMA loc_audit_functions AUTHORIZATION loc_audit_functions;
Service functions schema
Functions for service and DML functions.
CREATE SCHEMA service_functions AUTHORIZATION service_functions;
Business functions schema
Functions for terminal business functions called by the client.
CREATE SCHEMA business_functions AUTHORIZATION business_functions;
Access rights
Role — DBA has full access to all schemas (separate from the DB Owner role).
CREATE ROLE dba_role;
GRANT store TO dba_role;
GRANT sys_functions TO dba_role;
GRANT loc_audit_functions TO dba_role;
GRANT service_functions TO dba_role;
GRANT business_functions TO dba_role;
Role — USER has the privilege EXECUTE in the schema business_functions.
CREATE ROLE user_role;
Privileges between schemas
GRANT
Since all functions are created with the attribute SECURITY DEFINER a command is required REVOKE EXECUTE ON ALL FUNCTION… FROM public;
REVOKE EXECUTE ON ALL FUNCTION IN SCHEMA sys_functions FROM public;
REVOKE EXECUTE ON ALL FUNCTION IN SCHEMA loc_audit_functions FROM public;
REVOKE EXECUTE ON ALL FUNCTION IN SCHEMA service_functions FROM public;
REVOKE EXECUTE ON ALL FUNCTION IN SCHEMA business_functions FROM public;
GRANT USAGE ON SCHEMA sys_functions TO dba_role;
GRANT EXECUTE ON ALL FUNCTIONS IN SCHEMA sys_functions TO dba_role;
GRANT USAGE ON SCHEMA loc_audit_functions TO dba_role;
GRANT EXECUTE ON ALL FUNCTIONS IN SCHEMA loc_audit_functions TO dba_role;
GRANT USAGE ON SCHEMA service_functions TO dba_role;
GRANT EXECUTE ON ALL FUNCTIONS IN SCHEMA service_functions TO dba_role;
GRANT USAGE ON SCHEMA business_functions TO dba_role;
GRANT EXECUTE ON ALL FUNCTIONS IN SCHEMA business_functions TO dba_role;
GRANT EXECUTE ON ALL FUNCTIONS IN SCHEMA business_functions TO user_role;
GRANT ALL PRIVILEGES ON SCHEMA store TO GROUP business_functions;
GRANT ALL PRIVILEGES ON ALL TABLES IN SCHEMA store TO business_functions;
GRANT USAGE ON ALL SEQUENCES IN SCHEMA store TO business_functions;
Thus, the database schema is ready. You can start populating it with data.
Target tables
Creating tables is trivial. No special features, except for the decision to refrain from using SERIAL and to generate sequences explicitly. Plus, of course, maximum utilization of the statement
COMMENT ON ...Comments for all objects, without exception.
Local audit
To log the execution of stored functions and modifications to target tables, a local audit table is used, which includes details of the client connection, the label of the invoked module, and the actual values of input and output parameters in JSON format.
System functions
Designed for logging changes in target tables. They are trigger functions.
Template — system function
---------------------------------------------------------
-- INSERT
CREATE OR REPLACE FUNCTION sys_functions.table_insert_log ()
RETURNS TRIGGER AS $$
BEGIN
PERFORM loc_audit_functions.make_log( ' '||'table' , 'insert' , json_build_object('id', NEW.id) );
RETURN NULL ;
END
$$ LANGUAGE plpgsql SECURITY DEFINER;
CREATE TRIGGER table_after_insert AFTER INSERT ON storage.table FOR EACH ROW EXECUTE PROCEDURE sys_functions.table_insert_log();
---------------------------------------------------------
-- UPDATE
CREATE OR REPLACE FUNCTION sys_functions.table_update_log ()
RETURNS TRIGGER AS $$
BEGIN
IF OLD.column != NEW.column
THEN
PERFORM loc_audit_functions.make_log( ' '||'table' , 'update' , json_build_object('OLD.column', OLD.column , 'NEW.column' , NEW.column ) );
END IF ;
RETURN NULL ;
END
$$ LANGUAGE plpgsql SECURITY DEFINER;
CREATE TRIGGER table_after_update AFTER UPDATE ON storage.table FOR EACH ROW EXECUTE PROCEDURE sys_functions.table_update_log ();
---------------------------------------------------------
-- DELETE
CREATE OR REPLACE FUNCTION sys_functions.table_delete_log ()
RETURNS TRIGGER AS $$
BEGIN
PERFORM loc_audit_functions.make_log( ' '||'table' , 'delete' , json_build_object('id', OLD.id ) );
RETURN NULL ;
END
$$ LANGUAGE plpgsql SECURITY DEFINER;
CREATE TRIGGER table_after_delete AFTER DELETE ON storage.table FOR EACH ROW EXECUTE PROCEDURE sys_functions.table_delete_log ();Service Functions
Designed for implementing service and DML operations on target tables.
Template — Service Function
--INSERT
--RETURN id OF NEW ROW
CREATE OR REPLACE FUNCTION service_functions.table_insert ( new_column store.table.column%TYPE )
RETURNS integer AS $$
DECLARE
new_id integer ;
BEGIN
-- Generate new id
new_id = nextval('store.table.seq');
-- Insert into table
INSERT INTO store.table
(
id ,
column
)
VALUES
(
new_id ,
new_column
);
RETURN new_id ;
END
$$ LANGUAGE plpgsql SECURITY DEFINER;
--DELETE
--RETURN ROW NUMBERS DELETED
CREATE OR REPLACE FUNCTION service_functions.table_delete ( current_id integer )
RETURNS integer AS $$
DECLARE
rows_count integer ;
BEGIN
DELETE FROM store.table WHERE id = current_id;
GET DIAGNOSTICS rows_count = ROW_COUNT;
RETURN rows_count ;
END
$$ LANGUAGE plpgsql SECURITY DEFINER;
-- UPDATE DETAILS
-- RETURN ROW NUMBERS UPDATED
CREATE OR REPLACE FUNCTION service_functions.table_update_column
(
current_id integer
,new_column store.table.column%TYPE
)
RETURNS integer AS $$
DECLARE
rows_count integer ;
BEGIN
UPDATE store.table
SET
column = new_column
WHERE id = current_id;
GET DIAGNOSTICS rows_count = ROW_COUNT;
RETURN rows_count ;
END
$$ LANGUAGE plpgsql SECURITY DEFINER;Business Functions
Designed for final business functions called by the client. Always return — JSON. For intercepting and logging runtime errors, a block is used EXCEPTION.
Template — Business Function
CREATE OR REPLACE FUNCTION business_functions.business_function_template(
--Input parameters
)
RETURNS JSON AS $$
DECLARE
------------------------
--for exception catching
error_message text ;
error_json json ;
result json ;
------------------------
BEGIN
--LOGGING
PERFORM loc_audit_functions.make_log
(
'business_function_template',
'STARTED',
json_build_object
(
--IN Parameters
)
);
PERFORM business_functions.notice('business_function_template');
--START BUSINESS PART
--END BUSINESS PART
-- SUCCESSFUL RESULT
PERFORM business_functions.notice('result');
PERFORM business_functions.notice(result);
PERFORM loc_audit_functions.make_log
(
'business_function_template',
'FINISHED',
json_build_object( 'result',result )
);
RETURN result ;
----------------------------------------------------------------------------------------------------------
-- EXCEPTION CATCHING
EXCEPTION
WHEN OTHERS THEN
PERFORM loc_audit_functions.make_log
(
'business_function_template',
'STARTED',
json_build_object
(
--IN Parameters
) , TRUE );
PERFORM loc_audit_functions.make_log
(
'business_function_template',
' ERROR',
json_build_object('SQLSTATE',SQLSTATE ), TRUE
);
PERFORM loc_audit_functions.make_log
(
'business_function_template',
' ERROR',
json_build_object('SQLERRM',SQLERRM ), TRUE
);
GET STACKED DIAGNOSTICS error_message = RETURNED_SQLSTATE ;
PERFORM loc_audit_functions.make_log
(
'business_function_template',
' ERROR-RETURNED_SQLSTATE',json_build_object('RETURNED_SQLSTATE',error_message ), TRUE );
GET STACKED DIAGNOSTICS error_message = COLUMN_NAME ;
PERFORM loc_audit_functions.make_log
(
'business_function_template',
' ERROR-COLUMN_NAME',
json_build_object('COLUMN_NAME',error_message ), TRUE );
GET STACKED DIAGNOSTICS error_message = CONSTRAINT_NAME ;
PERFORM loc_audit_functions.make_log
(
'business_function_template',
' ERROR-CONSTRAINT_NAME',
json_build_object('CONSTRAINT_NAME',error_message ), TRUE );
GET STACKED DIAGNOSTICS error_message = PG_DATATYPE_NAME ;
PERFORM loc_audit_functions.make_log
(
'business_function_template',
' ERROR-PG_DATATYPE_NAME',
json_build_object('PG_DATATYPE_NAME',error_message ), TRUE );
GET STACKED DIAGNOSTICS error_message = MESSAGE_TEXT ;
PERFORM loc_audit_functions.make_log
(
'business_function_template',
' ERROR-MESSAGE_TEXT',json_build_object('MESSAGE_TEXT',error_message ), TRUE );
GET STACKED DIAGNOSTICS error_message = SCHEMA_NAME ;
PERFORM loc_audit_functions.make_log
(
'business_function_template',
' ERROR-SCHEMA_NAME',json_build_object('SCHEMA_NAME',error_message ), TRUE );
GET STACKED DIAGNOSTICS error_message = PG_EXCEPTION_DETAIL ;
PERFORM loc_audit_functions.make_log
(
'business_function_template',
' ERROR-PG_EXCEPTION_DETAIL',
json_build_object('PG_EXCEPTION_DETAIL',error_message ), TRUE );
GET STACKED DIAGNOSTICS error_message = PG_EXCEPTION_HINT ;
PERFORM loc_audit_functions.make_log
(
'business_function_template',
' ERROR-PG_EXCEPTION_HINT',json_build_object('PG_EXCEPTION_HINT',error_message ), TRUE );
GET STACKED DIAGNOSTICS error_message = PG_EXCEPTION_CONTEXT ;
PERFORM loc_audit_functions.make_log
(
'business_function_template',
' ERROR-PG_EXCEPTION_CONTEXT',json_build_object('PG_EXCEPTION_CONTEXT',error_message ), TRUE );
RAISE WARNING 'ALARM: %' , SQLERRM ;
SELECT json_build_object
(
'isError' , TRUE ,
'errorMsg' , SQLERRM
) INTO error_json ;
RETURN error_json ;
END
$$ LANGUAGE plpgsql SECURITY DEFINER;Summary
To provide an overview, I think this is quite sufficient. If anyone is interested in the details and outcomes, feel free to leave comments, and I will gladly add more nuances to the picture.
P.S.
Logging a simple error — type of input parameter
-[ RECORD 1 ]-
date_trunc | 2020-08-19 13:15:46
id | 1072
usename | emp1
log_module | addKD
log_module_hash | 0b4c1529a89af3ddf6af3821dc790e8a
status | STARTED
jsonb_pretty | {
| "dko": {
| "id": 4,
| "type": "Type1",
| "title": "CREATED BY addKD",
| "Weight": 10,
| "Tr": "300",
| "reduction": 10,
| "isTrud": "TRUE",
| "description": "description",
| "lowerTr": "100",
| "measurement": "measurement1",
| "methodology": "m1",
| "passportUrl": "files",
| "upperTr": "200",
| "weightingFactor": 100.123,
| "actualTrValue": null,
| "upperTrCalcNumber": "120"
| },
| "CardId": 3
| }
-[ RECORD 2 ]-
date_trunc | 2020-08-19 13:15:46
id | 1073
usename | emp1
log_module | addKD
log_module_hash | 0b4c1529a89af3ddf6af3821dc790e8a
status | ERROR
jsonb_pretty | {
| "SQLSTATE": "22P02"
| }
-[ RECORD 3 ]-
date_trunc | 2020-08-19 13:15:46
id | 1074
usename | emp1
log_module | addKD
log_module_hash | 0b4c1529a89af3ddf6af3821dc790e8a
status | ERROR
jsonb_pretty | {
| "SQLERRM": "invalid input syntax for type numeric: "null""
| }
-[ RECORD 4 ]-
date_trunc | 2020-08-19 13:15:46
id | 1075
usename | emp1
log_module | addKD
log_module_hash | 0b4c1529a89af3ddf6af3821dc790e8a
status | ERROR-RETURNED_SQLSTATE
jsonb_pretty | {
| "RETURNED_SQLSTATE": "22P02"
| }
-[ RECORD 5 ]-
date_trunc | 2020-08-19 13:15:46
id | 1076
usename | emp1
log_module | addKD
log_module_hash | 0b4c1529a89af3ddf6af3821dc790e8a
status | ERROR-COLUMN_NAME
jsonb_pretty | {
| "COLUMN_NAME": ""
| }
-[ RECORD 6 ]-
date_trunc | 2020-08-19 13:15:46
id | 1077
usename | emp1
log_module | addKD
log_module_hash | 0b4c1529a89af3ddf6af3821dc790e8a
status | ERROR-CONSTRAINT_NAME
jsonb_pretty | {
| "CONSTRAINT_NAME": ""
| }
-[ RECORD 7 ]-
date_trunc | 2020-08-19 13:15:46
id | 1078
usename | emp1
log_module | addKD
log_module_hash | 0b4c1529a89af3ddf6af3821dc790e8a
status | ERROR-PG_DATATYPE_NAME
jsonb_pretty | {
| "PG_DATATYPE_NAME": ""
| }
-[ RECORD 8 ]-
date_trunc | 2020-08-19 13:15:46
id | 1079
usename | emp1
log_module | addKD
log_module_hash | 0b4c1529a89af3ddf6af3821dc790e8a
status | ERROR-MESSAGE_TEXT
jsonb_pretty | {
| "MESSAGE_TEXT": "invalid input syntax for type numeric: "null""
| }
-[ RECORD 9 ]-
date_trunc | 2020-08-19 13:15:46
id | 1080
usename | emp1
log_module | addKD
log_module_hash | 0b4c1529a89af3ddf6af3821dc790e8a
status | ERROR-SCHEMA_NAME
jsonb_pretty | {
| "SCHEMA_NAME": ""
| }
-[ RECORD 10 ]-
date_trunc | 2020-08-19 13:15:46
id | 1081
usename | emp1
log_module | addKD
log_module_hash | 0b4c1529a89af3ddf6af3821dc790e8a
status | ERROR-PG_EXCEPTION_DETAIL
jsonb_pretty | {
| "PG_EXCEPTION_DETAIL": ""
| }
-[ RECORD 11 ]-
date_trunc | 2020-08-19 13:15:46
id | 1082
usename | emp1
log_module | addKD
log_module_hash | 0b4c1529a89af3ddf6af3821dc790e8a
status | ERROR-PG_EXCEPTION_HINT
jsonb_pretty | {
| "PG_EXCEPTION_HINT": ""
| }
-[ RECORD 12 ]-
date_trunc | 2020-08-19 13:15:46
id | 1083
usename | emp1
log_module | addKD
log_module_hash | 0b4c1529a89af3ddf6af3821dc790e8a
status | ERROR-PG_EXCEPTION_CONTEXT
jsonb_pretty | {
usename | emp1
log_module | addKD
log_module_hash | 0b4c1529a89af3ddf6af3821dc790e8a
status | ERROR-MESSAGE_TEXT
jsonb_pretty | {
| "MESSAGE_TEXT": "invalid input syntax for type numeric: "null""
| }Source: habr.com
