Question

We’re using a “1 audit table for each monitored Table”; However, in our case emp(PARENT) table has a child table emp_address which also needs to be monitored, so we have emp_audit and emp_address_audit tables.

postgres audit SQL : how to join PARENT and CHILD tables for reporting purposes.

/* Employee table */    
create table emp (
 emp_id integer primary key,
 empnum  integer,
 empname varchar(50),
 loginid varchar(20),
 updatetime timestamp
);

/* Address table */    
create table emp_addr (
 addr_id integer primary key,
 emp_id integer, -- references table emp
 line1 varchar(30),
 line2 varchar(30),
 loginid varchar(20),
 updatetime timestamp
);

/* Audit table for emp table */    
create table emp_audit (
 operation   character(1),
 emp_id integer,
 empnum  integer,
 empname varchar(50),
 loginid varchar(20),
 updatetime timestamp,
 txid bigint
);

/* Audit table for emp_addr table */    
create table emp_addr_audit (
 operation   character(1),
 addr_id integer,
 emp_id integer,
 line1 varchar(30),
 line2 varchar(30),
 loginid varchar(20),
 updatetime timestamp,
 txid bigint
);

We’re using hibernate(java) for persistence and hibernate updates only those tables whose columns were modified in the update operation. Given this, I might have multiple(say, 5) records in the emp_addr_audit table for 1 emp_audit table.

The report needs 1 row for each transaction(modification).
The report will have the following columns

empnum, empname, line1, line2, operation(insert/delete/update), loginid, updatetime

Let’s consider 2 scenarios to understand what’s needed:

1. In the initial transaction only emp attributes are created. Then in a separate transaction, the corresponding row in emp_addr is created. So, now, we have 1 row in emp_audit table and 1 row in emp_addr_audit table. The report will have 2 rows (one each for each transaction).
2. Both emp and emp_addr attributes are created in a single transaction. This will ensure that there is 1 row in emp_audit and 1 row in emp_addr_audit. Now, the report will have ONLY 1 row (since both table rows were created in a single transaction).

What SQL will satisfy both the above scenarios?

UPDATE
Scenario :
Transaction #1 : I insert a row into both emp and emp_addr. This results in a row each in emp_audit and emp_addr_audit.(INSERT)
Transaction #2 : I update the above emp’ attribute. This results in a UPDATE row in emp_audit.
Transaction #3 : I update the above emp_addr’s attribute. This results in a UPDATE row in emp_addr_audit.

I tried the following SQL #1 and it returned 3 rows as expected;

SQL #1

SELECT emp.*, addr.*
 FROM  emp_audit emp 
 FULL OUTER JOIN emp_addr addr USING(emp_id, txid);

However, when I added a where clause to the SQL, it returns only 2 rows. The missing row was the result of Transaction #3, where only emp_addr table row was UPDATED and emp table row was untouched.
SQL #2

SELECT emp.*, addr.*
 FROM  emp_audit emp 
        FULL OUTER JOIN emp_addr addr USING(emp_id, txid);
WHERE  emp.empnum = 20;

What SQL will STILL be able to get me 3 rows for the 3 transactions so that I can still filter out based on empnum ?

Thank you,

Answer 1

Firstly add an additional column txid bigint to the audit tables, then modify the stored proc that does the audit to call txid_current() to store the current transaction id with the audit record.

CREATE OR REPLACE FUNCTION process_emp_audit() RETURNS TRIGGER AS $emp_audit$
    BEGIN
        --
        -- Create a row in emp_audit to reflect the operation performed on emp,
        -- make use of the special variable TG_OP to work out the operation.
        --
        IF (TG_OP = 'DELETE') THEN
            INSERT INTO emp_audit SELECT 'D', now(), user, txid_current(), OLD.*;
            RETURN OLD;
        ELSIF (TG_OP = 'UPDATE') THEN
            INSERT INTO emp_audit SELECT 'U', now(), user, txid_current(), NEW.*;
            RETURN NEW;
        ELSIF (TG_OP = 'INSERT') THEN
            INSERT INTO emp_audit SELECT 'I', now(), user, txid_current(), NEW.*;
            RETURN NEW;
        END IF;
        RETURN NULL; -- result is ignored since this is an AFTER trigger
    END;
$emp_audit$ LANGUAGE plpgsql;

Then when you need to report the audit records do an outer join between the 2 tables using the emp_id and the txid so that you can present the 2 separate inserts that occur within the same transaction on a single line.

SELECT emp_audit.*, emp_addr_audit.*
  FROM emp_audit
  FULL OUTER JOIN ON emp_audit.emp_id = emp_addr_audit.emp_id
                 AND emp_audit.txid = emp_addr_audit.txid;