Saturday, June 17, 2017

Relational Algebra - Joins - Theta Join, Equijoin, Natural Join, Outer Join, Semijoin

I wrote a post on Relational Algebra that discusses most of operations related to it. This is the continuation of it and this explains Join Operations related to Relational Algebra. You may find that it is different with Joins we do now but it is the foundation for all type of joins we do with our relational databases.

Join Operations

Join is one of the main operations we perform against relations (tables) for retrieving data. It is done over the Cartesian Product of the two operand relations, using a SELECT statement with a Predicate. We are familiar with Joins like INNER JOIN, OUTER JOIN and CROSS JOIN but initially there were introduced with types like Theta Join, Equijoin, Natural Join, Outer Join and Semijoin. Modern DBMSs have enhanced these and have different implementations and that is why we do not see these types with mentioned names. But let's try to understand each of these types and how they are represented with Relational Algebra. Note that I have used different SQL Statement but it can be implemented using many techniques. Remember, almost all Joins are based on Cartesian Products.

Theta Join

This is based on a Predicate added to a Cartesian Product. In simple term, if you have joined two tables using CROSS JOIN, then you can add a filter to the result using one of the comparison operators. See the example given. Note that it can be implemented using SELECTION over a Cartesian Product as well.


Equijoin

This is same as Theta Join but the comparison operator is equal. Generally, if the operator of the Theta Join is equal operator (=), then the join is called as Equijoin instead of Theta Join, Here are two examples;



Natural Join

Natural Join is an Equijoin of two relations over all common attributes. In other words, when joining two tables, join is done using all common columns. Therefore, explicit Predicate is not required. See the sample given. I have used NATURAL JOIN which is not available with some DBMSs. Note that Common Attributes are not duplicated.


Outer Join

This join type includes both matching and no matching values from one relation and matching values from the other relation when two relations are joined. The relation that returns all tuples is determined using the Symbol used for the operation. If the Symbol is opened for the Left Relation, it is considered as the relation that returns all tuples. This is implemented using either LEFT or RIGHT in SQL.


Semijoin

Here is the last one. This join performs a join operations over two relations and projects over the attributes of first operand (or the relation). With this join, tuples can be limited for the join operation by adding a predicate, increasing the performance of the join operation.

 

No comments: