The HAVING clause filters group rows created by the GROUP BY clause. The HAVING clause is applied to each group of the grouped table, much as a WHERE clause is applied to a select list. If there is no GROUP BY clause, the HAVING clause is applied to the entire result as a single group. The SELECT clause cannot refer directly to any column that does not have a GROUP BY clause.
The HAVING clause supports the following syntax:
A boolean expression can include one or more of the following operators:
- IS NULL
- IS NOT NULL
- Comparison operators
- Quantified comparison operators
- Any column referenced in a HAVING clause must be either a grouping column or a column that refers to the result of an aggregate function.
- In a HAVING clause, you cannot specify:
- An alias that was defined in the select list. You must repeat the original, unaliased expression.
- An ordinal number that refers to a select list item. Only the GROUP BY and ORDER BY clauses accept ordinal numbers.
The following example query uses the HAVING clause to constrain an aggregate result. Drill queries the
dfs.clicks workspace and returns the total number of clicks for devices that indicate high click-throughs:
0: jdbc:drill:> select t.user_info.device, count(*) from \`clicks/clicks.json\` t group by t.user_info.device having count(*) > 1000; |------------|------------| | EXPR$0 | EXPR$1 | |------------|------------| | IOS5 | 11814 | | AOS4.2 | 5986 | | IOS6 | 4464 | | IOS7 | 3135 | | AOS4.4 | 1562 | | AOS4.3 | 3039 | |------------|------------|
The aggregate is a count of the records for each different mobile device in the clickstream data. Only the activity for the devices that registered more than 1000 transactions qualify for the result set.