Multi-joins and left joins

Author: jfarmer

left_join.sql

@@ -0,0 +1,87 @@
+-- Outline
+--
+-- So far, we've been talking about joins using the JOIN keyword. This is
+-- just one type of join, called an INNER JOIN.  In fact, wherever we've
+-- been using the "JOIN" keyword we could instead replace it with "INNER JOIN"
+-- without changing anything.
+--
+-- There's a second type of join, called a left join or left outer join.
+-- It's a bit difficult to explain, but here are some blog posts:
+--   http://stackoverflow.com/questions/448023/what-is-the-difference-between-left-right-outer-and-inner-joins
+--   http://blog.codinghorror.com/a-visual-explanation-of-sql-joins/
+
+-- Compare the output of these queries:
+-- JOIN
+SELECT employees.id, employees.first_name, employees.last_name,
+       COUNT(DISTINCT(customers.id)) AS customers_served
+FROM employees
+JOIN customers
+  ON (employees.id = customers.support_rep_id)
+GROUP BY employees.id;
+
+-- LEFT JOIN
+SELECT employees.id, employees.first_name, employees.last_name,
+       COUNT(DISTINCT(customers.id)) AS customers_served
+FROM employees
+LEFT JOIN customers
+  ON (employees.id = customers.support_rep_id)
+GROUP BY employees.id;
+
+-- Note: Unlike JOIN, with a LEFT JOIN it does matter which table comes first.
+-- That is,
+--   FROM TableA LEFT JOIN TableB
+-- **IS NOT THE SAME AS**
+--   FROM TableB LEFT JOIN TableA
+
+-- In any join (inner, left, or anything else) there are three components:
+--   1. The left table
+--   2. The right table
+--   3. The join condition
+-- An INNER JOIN returns only those rows from the left and right tables that
+-- match the join condition.  A LEFT join returns all rows from the left table
+-- no matter what, and any rows from the right table that match the join condition.
+-- In other words, in an INNER JOIN, if a table in the left row doesn't have
+-- a corresponding row in the right table then that row (in the left table)
+-- won't be returned, whereas it WILL be returned in the result set using
+-- a LEFT JOIN.
+
+-- You can see this in the COUNT queries above, but more clearly, here:
+-- This will return nothing since there are no customers who have Employee #1
+-- as a customer service rep.  Remember, "INNER JOIN" is the same as "JOIN"
+SELECT *
+FROM employees
+INNER JOIN customers
+  ON (employees.id = customers.support_rep_id)
+WHERE employees.id = 1;
+
+-- This will return a SINGLE row, even though there are no corresponding customers.
+-- The fields from the customers table are displayed as empty (or NULL), telling
+-- us that Employee #1 has no associated customers.
+SELECT *
+FROM employees
+LEFT JOIN customers
+  ON (employees.id = customers.support_rep_id)
+WHERE employees.id = 1;
+
+-- Employees is the left table, customers is the right table, and the join
+-- condition is employees.id = customers.support_rep_id.
+
+-- One handy thing is that this allows us to create queries like...
+-- Show me all employees who don't support any customers
+
+SELECT *
+FROM employees
+LEFT JOIN customers
+  ON (employees.id = customers.support_rep_id)
+WHERE customers.id IS NULL;
+
+-- There are no exercises, here, because the toy database doesn't have
+-- interesting enough data to merit the use of LEFT JOINs in many contexts.
+-- For example, this query would return all customers who had yet to make an
+-- order.  In the toy database, there are no such customers, alas!
+
+SELECT *
+FROM customers
+LEFT JOIN invoices
+  ON (invoices.customer_id = customers.id)
+WHERE invoices.id IS NULL;

multi_join.sql

@@ -0,0 +1,57 @@
+-- Outline
+--
+-- All of these JOINs requires more than two tables.
+-- Remember: then using the JOIN keyword it doesn't matter what order we join
+-- the tables in.  That is,
+--
+--   SELECT * FROM TableA JOIN TableB ON (TableA.id = TableB.table_a_id)
+--
+-- is the same as
+--
+--   SELECT * FROM TableB JOIN TableA ON (TableA.id = TableB.table_a_id)
+--
+-- Also remember: the syntax TableA.first_name means the "first_name" field
+-- from the table named TableA.  Field names have to be unique in a given table.
+-- That is, you can't have two fields / columns named "first_name" in the same
+-- table.
+--
+-- However, two *different* tables can both have a "first_name" field. For
+-- example, you might have a "customers" table and an "employees" table, both
+-- of which have "first_name", "last_name", and "email" fields.  If you have
+-- a query that involves *both* customers and employees, you'll need to refer
+-- to fields using their table name, i.e., customers.first_name,
+-- employees.first_name, customers.email, and so on.
+
+-- Remember, the "skeleton" for a SELECT query with JOINs looks like
+--
+--   SELECT <field names>
+--   FROM <table1>
+--   JOIN <table2>
+--     ON (<join condition>)
+--   JOIN <table3>
+--     ON (<join condition>)
+--   WHERE <filtering clauses>
+--   GROUP BY / ORDER BY / LIMIT / etc.
+--
+-- This is the general order, but a SELECT query *always* needs a FROM clause.
+
+-- Exercises
+--
+
+-- What are the five highest grossing hip hop tracks?
+
+-- What are the 3 least frequently-purchased tv shows?
+
+-- What are the 5 highest-grossing genres?
+
+-- Who are the 5 artists with the longest average track length?
+
+-- Write a query that sorts a specific album's tracks by sales volume (you'll have to pick an arbitrary album title)
+--   e.g., "Give me the the tracks from Alanis Morissette's Jagged Little Pilled sorted by sales volume"
+
+-- Write a query that sorts a specific album's tracks by gross sales (you'll have to pick an arbitrary album title)
+
+-- How would you handle the case where two different artists have albums with the same name?
+--   e.g., sometimes both Peter Gabriel's first album and Led Zeppelin's first album are referred to as "One" or "I"
+
+-- We're runnin' low on queries!  Come up with 3-5 of your own and write them.