Stored Procedures and Triggers

Advanced⏱120 minutes📖3 min readLesson 7 of 10

Stored Procedures and Triggers

Master CREATE PROCEDURE, parameters, variables, control flow, cursors, and triggers (BEFORE/AFTER, INSERT/UPDATE/DELETE)

Stored Procedures

Stored procedures are reusable database routines that can contain business logic, control flow, and transaction management.

sql

CREATE OR REPLACE PROCEDURE transfer_funds(
    p_from_account INT,
    p_to_account INT,
    p_amount NUMERIC
)
LANGUAGE plpgsql
AS $$
BEGIN
    -- Start implicit transaction
    UPDATE accounts SET balance = balance - p_amount
    WHERE id = p_from_account;

    UPDATE accounts SET balance = balance + p_amount
    WHERE id = p_to_account;

    COMMIT;
EXCEPTION
    WHEN OTHERS THEN
        ROLLBACK;
        RAISE;
END;
$$;

-- Call
CALL transfer_funds(1001, 2005, 500.00);

Procedures vs Functions

Aspect	Procedure	Function
Return value	None (or OUT params)	Must return a value
Transaction control	Yes (COMMIT/ROLLBACK)	No
In SELECT	No	Yes
Output parameters	Yes	Yes
Mutate data	Yes	Yes (volatile functions)

Parameters and Variables

sql

CREATE OR REPLACE PROCEDURE process_order(
    p_customer_id INT,
    p_items JSONB,
    p_discount NUMERIC DEFAULT 0
)
LANGUAGE plpgsql
AS $$
DECLARE
    v_order_id BIGINT;
    v_total NUMERIC := 0;
    v_item JSONB;
    v_price NUMERIC;
    v_quantity INT;
BEGIN
    -- Create the order
    INSERT INTO orders (customer_id, order_date, status)
    VALUES (p_customer_id, NOW(), 'pending')
    RETURNING id INTO v_order_id;

    -- Process line items
    FOR v_item IN SELECT * FROM jsonb_array_elements(p_items)
    LOOP
        v_price := (v_item->>'price')::NUMERIC;
        v_quantity := (v_item->>'quantity')::INT;

        INSERT INTO order_items (order_id, product_id, quantity, unit_price)
        VALUES (v_order_id, (v_item->>'product_id')::INT, v_quantity, v_price);

        v_total := v_total + (v_price * v_quantity);
    END LOOP;

    -- Apply discount
    v_total := v_total * (1 - p_discount / 100);

    UPDATE orders
    SET total = v_total, status = 'confirmed'
    WHERE id = v_order_id;

    RAISE NOTICE 'Order % created: total %', v_order_id, v_total;
END;
$$;

ℹ️Note

Use RAISE NOTICE for debugging stored procedures. In production, consider logging to a table for audit trails.

Control Flow

sql

CREATE OR REPLACE PROCEDURE apply_penalty()
LANGUAGE plpgsql
AS $$
DECLARE
    v_penalty NUMERIC := 25.00;
BEGIN
    -- IF/ELSIF/ELSE
    IF CURRENT_DATE >= '2024-12-25' THEN
        v_penalty := 0;  -- holiday waiver
    ELSIF CURRENT_DATE >= '2024-12-20' THEN
        v_penalty := v_penalty * 0.5;
    END IF;

    -- Simple CASE
    CASE EXTRACT(DOW FROM CURRENT_DATE)
        WHEN 0 THEN v_penalty := v_penalty * 0;  -- Sunday amnesty
        WHEN 6 THEN v_penalty := v_penalty * 0.5;  -- Saturday half
        ELSE NULL;  -- full penalty on weekdays
    END CASE;

    -- Loops
    <<process_loop>>
    LOOP
        -- process logic
        EXIT process_loop WHEN <condition>;
    END LOOP;

    -- WHILE loop
    WHILE v_counter > 0 LOOP
        -- logic
        v_counter := v_counter - 1;
    END LOOP;

    -- FOR loop over query
    FOR rec IN SELECT * FROM overdue_accounts LOOP
        UPDATE accounts SET balance = balance - v_penalty
        WHERE id = rec.id;
    END LOOP;

    -- FOR loop over range
    FOR i IN 1..10 LOOP
        RAISE NOTICE 'Iteration %', i;
    END LOOP;
END;
$$;

Cursors

For processing large result sets row-by-row without loading everything into memory.

sql

CREATE OR REPLACE PROCEDURE generate_monthly_statements()
LANGUAGE plpgsql
AS $$
DECLARE
    cur CURSOR FOR
        SELECT id, customer_id, balance
        FROM accounts
        WHERE active = true;

    v_account RECORD;
    v_statement_id BIGINT;
BEGIN
    OPEN cur;

    LOOP
        FETCH cur INTO v_account;
        EXIT WHEN NOT FOUND;

        -- Create statement for this account
        INSERT INTO statements (account_id, month, created_at)
        VALUES (v_account.id, DATE_TRUNC('month', CURRENT_DATE), NOW())
        RETURNING id INTO v_statement_id;

        -- Insert line items
        INSERT INTO statement_lines (statement_id, description, amount)
        SELECT v_statement_id, description, amount
        FROM transactions
        WHERE account_id = v_account.id
          AND transaction_date >= DATE_TRUNC('month', CURRENT_DATE - INTERVAL '1 month')
          AND transaction_date < DATE_TRUNC('month', CURRENT_DATE);

        RAISE NOTICE 'Statement % created for account %', v_statement_id, v_account.id;
    END LOOP;

    CLOSE cur;
END;
$$;

⚠️Warning

Cursors are slow compared to set-based operations. Use them only when you must process rows sequentially (e.g., external API calls, complex row-level calculations). Whenever possible, rewrite as a single INSERT ... SELECT.

Triggers

Triggers execute automatically in response to DML events.

Trigger Types

Type	Timing	Per	Event
`BEFORE INSERT`	Before row insert	Row	INSERT
`AFTER INSERT`	After row insert	Row/Statement	INSERT
`BEFORE UPDATE`	Before row update	Row	UPDATE
`AFTER UPDATE`	After row update	Row/Statement	UPDATE
`BEFORE DELETE`	Before row delete	Row	DELETE
`AFTER DELETE`	After row delete	Row/Statement	DELETE
`INSTEAD OF`	On views	Row	Any

BEFORE INSERT Trigger (Validation)

sql

CREATE OR REPLACE FUNCTION validate_order()
RETURNS TRIGGER
LANGUAGE plpgsql
AS $$
BEGIN
    -- Validate required fields
    IF NEW.customer_id IS NULL THEN
        RAISE EXCEPTION 'customer_id is required';
    END IF;

    IF NEW.total <= 0 THEN
        RAISE EXCEPTION 'total must be positive, got %', NEW.total;
    END IF;

    -- Set defaults
    IF NEW.status IS NULL THEN
        NEW.status := 'pending';
    END IF;

    IF NEW.order_date IS NULL THEN
        NEW.order_date := NOW();
    END IF;

    RETURN NEW;
END;
$$;

CREATE TRIGGER trg_before_insert_order
    BEFORE INSERT ON orders
    FOR EACH ROW
    EXECUTE FUNCTION validate_order();

AFTER INSERT Trigger (Audit Log)

sql

CREATE OR REPLACE FUNCTION log_order_insert()
RETURNS TRIGGER
LANGUAGE plpgsql
AS $$
BEGIN
    INSERT INTO audit_log (table_name, operation, record_id, old_data, new_data, changed_by, changed_at)
    VALUES (
        'orders',
        'INSERT',
        NEW.id,
        NULL,
        row_to_json(NEW),
        CURRENT_USER,
        NOW()
    );
    RETURN NEW;
END;
$$;

CREATE TRIGGER trg_after_insert_order
    AFTER INSERT ON orders
    FOR EACH ROW
    EXECUTE FUNCTION log_order_insert();

BEFORE UPDATE Trigger (Prevent Changes)

sql

CREATE OR REPLACE FUNCTION prevent_paid_order_update()
RETURNS TRIGGER
LANGUAGE plpgsql
AS $$
BEGIN
    IF OLD.status = 'paid' AND NEW.status != 'paid' THEN
        RAISE EXCEPTION 'Cannot modify a paid order';
    END IF;
    RETURN NEW;
END;
$$;

CREATE TRIGGER trg_before_update_order
    BEFORE UPDATE ON orders
    FOR EACH ROW
    EXECUTE FUNCTION prevent_paid_order_update();

Trigger for Denormalized Aggregate

sql

CREATE OR REPLACE FUNCTION update_account_balance()
RETURNS TRIGGER
LANGUAGE plpgsql
AS $$
BEGIN
    IF TG_OP = 'INSERT' THEN
        UPDATE accounts SET balance = balance + NEW.amount
        WHERE id = NEW.account_id;
    ELSIF TG_OP = 'DELETE' THEN
        UPDATE accounts SET balance = balance - OLD.amount
        WHERE id = OLD.account_id;
    ELSIF TG_OP = 'UPDATE' THEN
        UPDATE accounts SET balance = balance - OLD.amount + NEW.amount
        WHERE id = NEW.account_id;
    END IF;
    RETURN NEW;
END;
$$;

CREATE TRIGGER trg_transaction_balance
    AFTER INSERT OR UPDATE OR DELETE ON transactions
    FOR EACH ROW
    EXECUTE FUNCTION update_account_balance();

💡Tip

TG_OP is a special variable containing the operation: 'INSERT', 'UPDATE', 'DELETE', or 'TRUNCATE'. Use it to create multi-operation triggers.

Trigger Best Practices

Practice	Why
Keep triggers fast	They run in the same transaction: slow triggers block everything
Avoid recursive triggers	Trigger on A → inserts into B → trigger on B → inserts into A
Document side effects	Triggers are invisible to application code; document them
Test thoroughly	Triggers can break bulk operations and imports
Use `pg_trigger_depth()`	Prevent infinite recursion

sql

-- Prevent recursive triggers
CREATE OR REPLACE FUNCTION safe_update_log()
RETURNS TRIGGER
LANGUAGE plpgsql
AS $$
BEGIN
    IF pg_trigger_depth() > 1 THEN
        RETURN NEW;  -- called by another trigger, skip logging
    END IF;

    INSERT INTO audit_log (...) VALUES (...);
    RETURN NEW;
END;
$$;

Practical Examples

Example 1: Soft Delete Trigger

sql

CREATE OR REPLACE FUNCTION soft_delete()
RETURNS TRIGGER
LANGUAGE plpgsql
AS $$
BEGIN
    NEW.deleted_at := NOW();
    NEW.deleted_by := CURRENT_USER;
    RETURN NEW;
END;
$$;

CREATE TRIGGER trg_soft_delete
    BEFORE DELETE ON sensitive_data
    FOR EACH ROW
    EXECUTE FUNCTION soft_delete();

Example 2: Materialized View Refresh Trigger

sql

CREATE OR REPLACE FUNCTION refresh_sales_summary()
RETURNS TRIGGER
LANGUAGE plpgsql
AS $$
BEGIN
    REFRESH MATERIALIZED VIEW CONCURRENTLY sales_summary;
    RETURN NULL;
END;
$$;

CREATE TRIGGER trg_refresh_sales
    AFTER INSERT OR UPDATE OR DELETE ON orders
    FOR EACH STATEMENT
    EXECUTE FUNCTION refresh_sales_summary();

Practice Questions

Write a stored procedure that transfers inventory between two warehouse locations, including validation and error handling.
What is the difference between a BEFORE and AFTER trigger? Give a use case for each.
Create a trigger that prevents deleting a customer who has active orders.
Write a procedure that accepts a JSONB parameter containing order items, inserts the order, and returns the order ID.
What is pg_trigger_depth() and when would you use it?
Create a trigger that logs all UPDATE operations on the employees table, recording old and new salary values.
Write a cursor-based procedure to process a batch of 10,000 pending email notifications and mark them as sent.
What are the risks of using triggers for business logic? How would you mitigate them?
Create a FOR EACH STATEMENT trigger vs a FOR EACH ROW trigger. When would you choose one over the other?
Write a stored procedure with COMMIT inside a loop to batch-process 1M rows without exhausting transaction logs.

Progress70%

Table Partitioning and Sharding

JSON and JSONB Operations

Tip: use ← and → to switch lessons

Course Content

1. ROW_NUMBER(), RANK(), DENSE_RANK(), NTILE()2. LAG(), LEAD(), FIRST_VALUE(), LAST_VALUE(), NTH_VALUE()3. PIVOT, UNPIVOT, and Crosstab Queries 4. Recursive CTEs 5. Advanced Query Tuning 6. Table Partitioning and Sharding 7. Stored Procedures and Triggers 8. JSON and JSONB Operations 9. User Roles, Permissions, and Security 10. Backup, Recovery, and High Availability