The table list in the FROM clause actually defines all the candidate rows which may become part of the result. Conceptually, the first processing step of a query specification is to establish an intermediate resultant table containing all columns and all rows as defined in the table list. If only one table is involved, then the resultant table will be equivalent to the base table. However, should more than one table be listed, then all the tables in the list must be conceptually joined.

The next processing step concerns the WHERE clause. Each row in the intermediate resultant table is conceptually subjected to the search condition specified in the WHERE clause. If the condition equates to true, then the candidate row proceeds to the next stage. Otherwise, it is eliminated from further consideration, thus reducing the size of the final resultant table. Should no WHERE clause have been specified or the condition equate to true for all candidate rows then the subsequent resultant table will contain all rows as illustrated by the intermediate resultant table No.1 in the above step.

The next possible processing step concerns the GROUP BY clause. This step actually splits into two phases resulting in Tables No. 3 and No. 4. If built-in functions are used within a query, it is called a grouped query. The query is also grouped if a GROUP BY clause is specified, even if no functions are given. Built-in functions are aggregate operators which operate on a set of values in order to produce a single value as a result. These functions can be applied to the whole intermediate resultant table in order to produce a final resultant table of one row. In such a case, no GROUP BY clause is specified but the query is still grouped, as it uses built-in functions. Any column referenced within a grouped query must be an operand of a function, a grouping column or appear anywhere in the WHERE clause. This is because outside of the WHERE clause, the query is concerned with groups instead of mere rows. The converse, however, is not true. A grouping column may appear in a function. In the case of a special register, it must be specified exactly the same way as it appears in the SELECT list.

It is possible to divide the intermediate resultant table into groups. Groups are partitioned by specifying at least one grouping column in the GROUP BY list. A group is then established by extracting all candidate rows from the intermediate resultant table No. 2, where the value of the grouping column/s is/are equal. As many groups are established as there are differing values of the grouping column. There is no predetermined ordering of these groups. Groups are established as follows:

At this point the second phase is initiated. The query is examined in order to produce a list of the columns required for intermediate resultant table No. 4. These new columns are either grouping columns or columns derived from functions applied to columns in intermediate resultant table No. 3. In either case, only columns or functions appearing in the derived column list or the HAVING clause have to be considered. Thus, aggregate functions are applied to each group in turn resulting in one candidate row per group in intermediate resultant table No. 4.

The aggregate functions can now be applied to each group in turn resulting in one candidate row per group for the next conceptual intermediate table.

In conclusion, the GROUP BY clause establishes candidate groups which, when operated upon by the aggregate functions, are transformed into candidate rows, one per group, which form the next intermediate resultant table No. 4.

The next possible processing step concerns the HAVING clause . Each row in the intermediate resultant table is conceptually subjected to the search condition specified in the HAVING clause. If the condition equates to true, then the candidate row proceeds to the next stage, otherwise it is eliminated from further consideration. As such, it is analogous to the WHERE clause except it eliminates candidate groups rather than candidate rows. It is therefore permissible to use functions in the search conditions. In fact, columns which are not contained in a function must be specified in the GROUP BY list.

The final stage can now be executed, namely the production of the final resultant table. This is a derivation of the previous intermediate resultant table and is conceptually the same, regardless of whether it came from the HAVING, GROUP BY, WHERE or FROM clause. A resultant row is processed by evaluating each derived column in turn, based on the values contained in the corresponding row of the intermediate resultant table. This evaluation may be quite complex, depending on the nature of the expressions contained in the derived column's specification. Step 5 finalizes the processing of this query by producing the final resultant table no. 6.