SELECT query syntax

Firebolt supports running SELECT statements with the syntax described in this topic. You can run multiple queries in a single script. Separating them with a semicolon (;) is required. Firebolt also supports CREATE TABLE...AS SELECT (CTAS). For more information, see CREATE TABLE…AS SELECT.

• Syntax
• SELECT
  • SELECT expression
    • Example
  • SELECT wildcard
  • SELECT DISTINCT
  • SELECT ALL
• WITH
• FROM
• JOIN
• UNNEST
• WHERE
• GROUP BY
  • GROUP BY ALL
• HAVING
• UNION [ALL]
• ORDER BY
• LIMIT
• OFFSET

Syntax

[ WITH <with_query> [, ...n] ]
SELECT [ ALL | DISTINCT ] {<select_expr> [, ...]}
    [ FROM <from_item> [, ...] ]
    [ WHERE <condition> ]
    [ GROUP BY [ <grouping_element> [, ...] | ALL ] ]
    [ HAVING <condition> [, ...] ]
    [ UNION [ ALL ] <select_expr> [ ...n]
    [ ORDER BY <expression> [ ASC | DESC ] [ NULLS FIRST | NULLS LAST] [, ...] ]
    [ LIMIT <count> ]
    [ OFFSET <start> ]

SELECT

SELECT [ ALL | DISTINCT ] {<select_expr> [, ...]}

The SELECT list defines the columns that it returns. Each <select_expr> in the SELECT list can be either expression, or wildcards.

Selecting only partitioned or virtual columns is currently not supported in Firebolt. Selecting a combination of partitioned/virtual columns and regular columns is supported.

SELECT expression

<expression> [ AS <alias> ]

Expressions in the SELECT list evaluate to a single value and produce one output column. The output column names are defined either by an explicit alias in the AS clause, or, for expressions without explicit alias, the output column name is automatically generated. The expression can reference any column from the FROM clause, but cannot reference other columns produced by the same SELECT list. The expressions can use scalar functions, aggregate functions, window functions or subqueries if they return single element.

Example

SELECT price, quantity, price * quantity AS sales_amount FROM Sales

SELECT wildcard

[ <table_name>. ] * [ EXCLUDE { <column_name> | ( <column_name>, ... ) } ]

Wildcards are expanded to multiple output columns using the following rules:

• * is expanded to all columns in the FROM clause
• <table_name>.* is expanded to all columns in the FROM clause for the table named <table_name>
• EXCLUDE defines columns which are removed from the above expansion

SELECT DISTINCT

SELECT DISTINCT statement removes duplicate rows.

SELECT ALL

SELECT ALL statement returns all rows. SELECT ALL is the default behavior.

WITH

The WITH clause is used for subquery refactoring so that you can specify subqueries and then reference them as part of the main query. This simplifies the hierarchy of the main query, enabling you to avoid using multiple nested sub-queries.

In order to reference the data from the WITH clause, a name must be specified for it. This name is then treated as a temporary relation table during query execution.

The primary query and the queries included in the WITH clause are all executed at the same time; WITH queries are evaluated only once every time the main query is executed, even if the clause is referred to by the main query more than once.

Materialized common table expressions (Beta)

The query hint MATERIALIZED or NOT MATERIALIZED controls whether common table expressions (CTEs) produce an internal results table that is cached in engine RAM (MATERIALIZED) or calculated each time the sub-query runs. NOT MATERIALIZED is the default. MATERIALIZED must be specified explicitly.

Materialized results can be accessed more quickly in some circumstances. By using the proper materialization hint, you can control when a CTE gets materialized and improve query performance. We recommend the MATERIALIZED hint to improve query performance in the following circumstances:

• The CTE is reused at the main query level more than once.

• The CTE is computationally expensive, producing a relatively small number of rows.

• The CTE calculation is independent of the main query, and no external optimizations from the main table are needed for it to be fast.

• The materialized CTE fits into the nodes’ ram.

Syntax

WITH <subquery_table_name> AS [MATERIALIZED|NOT MATERIALIZED] <subquery>

Example

The following example retrieves all players who have subscribed to receive the game newsletter, having the results of the WITH query in the temporary table nl_subscribers.

The results of the main query then list the nickname and email for those customers, sorted by nickname.

WITH nl_subscribers AS (
	SELECT
		*
	FROM
		players
	WHERE
		issubscribedtonewsletter=TRUE
)
SELECT
	nickname,
	email
FROM
	players
ORDER BY
	nickname

FROM

Use the FROM clause to list the tables and any relevant join information and functions necessary for running the query.

Syntax

FROM <from_item> [, ...n]

Example

In the following example, the query retrieves all entries from the players table for which the agecategory value is “56+”.

SELECT
	*
FROM
	players
WHERE
	agecategory='56+'

JOIN

A JOIN operation combines rows from two data sources, such as tables or views, and creates a new table of combined rows that can be used in a query.

JOIN operations can be used with an ON clause for conditional logic or a USING clause to specify columns to match.

JOIN with ON clause syntax

FROM <join_table1> [ INNER | LEFT | RIGHT | FULL ] JOIN <join_table2> ON <join_condition>

JOIN with USING clause syntax

FROM <join_table1> [ INNER | LEFT | RIGHT | FULL ] JOIN <join_table2> USING (column_list)

JOIN Types

The type of JOIN operation specifies which rows are included between two specified tables. If unspecified, JOIN defaults to INNER JOIN.

JOIN types include:

Examples

The JOIN examples below use two tables, num_test and num_test2. These tables are created and populated with data as follows.

CREATE DIMENSION TABLE num_test (
    firstname TEXT,
    score INTEGER);

INSERT INTO num_test VALUES
    ('Carol', 11),
    ('Albert', 50),
    ('Sammy', 90),
    ('Peter', 50),
    ('Deborah', 90),
    ('Frank', 87),
    ('Thomas', 85),
    ('Humphrey', 56);

CREATE DIMENSION TABLE num_test2 (
    firstname TEXT,
    score INTEGER);

INSERT INTO num_test2 VALUES
    ('Sammy', 90),
    ('Hector', 56),
    ('Tom', 85),
    ('Peter', 50),
    ('Carl', 100),
    ('Frank', 87),
    ('Deborah', 90),
    ('Albert', 50);

The tables and their data are shown below.

INNER JOIN example

The INNER JOIN example below includes only the rows where the firstname and score values match.

SELECT
    *
FROM
    num_test
INNER JOIN
    num_test2
    USING (
        firstname,
        score
	);

This query is equivalent to:

SELECT
    *
FROM
    num_test
INNER JOIN
    num_test2
        ON num_test.firstname = num_test2.firstname
        AND num_test.score = num_test2.score;

Returns

LEFT OUTER JOIN example

The LEFT OUTER JOIN example below includes all firstname values from the num_test table. Any rows with no matching value in the num_test2 table return NULL.

SELECT
    num_test.firstname,
    num_test2.firstname
FROM num_test
LEFT OUTER JOIN
    num_test2
    USING (firstname);

Returns

RIGHT OUTER JOIN example

The RIGHT OUTER JOIN example below includes all firstname values from num_test2. Any rows with no matching values in the num_test table return NULL.

SELECT
    num_test.firstname,
    num_test2.firstname
FROM
    num_test
RIGHT OUTER JOIN
    num_test2
    USING (firstname);

Returns

FULL OUTER JOIN example

The FULL OUTER JOIN example below includes all values from num_test and num_test2. Any rows with no matching values return NULL.

SELECT
    num_test.firstname,
    num_test2.firstname
FROM
    num_test
FULL OUTER JOIN
    num_test2
    USING (firstname);

Returns

CROSS JOIN example

A CROSS JOIN produces a table with every combination of row values in the specified columns.

This example uses two tables, crossjoin_test and crossjoin_test2, each with a single letter column. The tables contain the following data.

The CROSS JOIN example below produces a table of every possible pairing of these rows.

SELECT
    crossjoin_test.letter,
    crossjoin_test2.letter
FROM
    crossjoin_test
CROSS JOIN
    crossjoin_test2;

Returns

UNNEST

An UNNEST operator performs join between the table in the left side, and the array in the right side. The output table repeats rows of the table from the left for every element of the array. If the array is empty, the corresponding row from the table is eliminated.

Syntax

FROM <from_item> UNNEST(<array_column> [[ AS ] <alias_name>][,<array_column>...])

Example

The example is based on the following table:

CREATE FACT TABLE table_with_arrays
(
    product TEXT,
    cost ARRAY(INTEGER)
) PRIMARY INDEX product;

Assume the table was populated and contains the following values:

The following query with UNNEST:

SELECT
	product,
	cost
FROM
	table_with_arrays UNNEST(cost);

Returns the following result:

WHERE

Use the WHERE clause to define conditions for the query in order to filter the query results. When included, the WHERE clause always follows the FROM clause as part of a command such as SELECT.

Syntax

WHERE <condition>

Example

In the following example, the query retrieves all entries from the customers table for which the region value is “EMEA”.

SELECT
	*
FROM
	customers
WHERE
	region = 'EMEA'

The following query retrieves users who registered after August 30, 2020 from the users’ table:

SELECT
	user_id,
	city,
	country
FROM
	users
WHERE
	registration_date >= TO_DATE('2020-08-30');

The following query retrieves users who registered after August 30 2020 and made a purchase:

SELECT
	user_id,
	city,
SELECT
	user_id,
	city,
	country
FROM
	users
WHERE
	registration_date >= TO_DATE('2020-08-30')
	AND user_id IN (
		SELECT
			user_id
		FROM
			purchases
	)

GROUP BY

The GROUP BY clause groups together input rows. Multiple input rows which have same values of expressions in the GROUP BY clause become a single row in the output. GROUP BY is typically used in conjunction with aggregate functions (such as SUM, MIN, etc). Query with GROUP BY clause and without aggregate functions is equivalent to SELECT DISTINCT.

Syntax

GROUP BY [ <grouping_element> [, ...n] | ALL ]

Example

In the following example, the results that are retrieved are grouped by the product_name and then by the product_id columns.

SELECT
	product_name,
	country,
	sum(total_sales)
FROM
	purchases
GROUP BY
	product_name,
	country

If the expression in GROUP BY clause is exactly the same as in the SELECT list, then its position can be used instead

SELECT
	product_name,
	country,
	SUM(total_sales)
FROM
	purchases
GROUP BY
	1,
	2

GROUP BY clause must include all expressions in the SELECT list which are not involving aggregate functions. It may include expressions which are not part of SELECT list.

SELECT SUM(total_sales) FROM purchases GROUP BY product_name

However, the following will cause an error, since SELECT list has an expression which is not an aggregate function and it is not listed in GROUP BY clause.

SELECT product_name, country, SUM(total_sales) FROM purchases GROUP BY product_name

GROUP BY ALL

For the common case of GROUP BY clause repeating all the non aggregate function expressions in the SELECT list, it is possible to use GROUP BY ALL syntax. It will automatically group by all non aggregate functions expressions from the SELECT list.

SELECT
	product_name,
	country,
	SUM(total_sales)
FROM
	purchases
GROUP BY ALL

HAVING

The HAVING clause is used in conjunction with the GROUP BY clause, and is computed after computing the GROUP BY clause and aggregate functions. HAVING is used to further eliminate groups that don’t satisfy the <condition> by filtering the GROUP BY results.

Syntax

HAVING <condition> [, ...n]

UNION [ALL]

The UNION operator combines the results of two or more SELECT statements into a single query.

• UNION combines with duplicate elimination.
• UNION ALL combines without duplicate elimination.

When including multiple clauses, the same number of columns must be selected by all participating SELECT statements. Data types of all column parameters must be the same. Multiple clauses are processed left to right; use parentheses to define an explicit order for processing.

Syntax

<select_expr1> UNION [ALL] <select_expr2> [ ...n]

ORDER BY

The ORDER BY clause sorts a result set by one or more output expressions. ORDER BY is evaluated as the last step after any GROUP BY or HAVING clause. ASC and DESC determine whether results are sorted in ascending or descending order. When the clause contains multiple expressions, the result set is sorted according to the first expression. Then the second expression is applied to rows that have matching values from the first expression, and so on.

The default null ordering is NULLS LAST, regardless of ascending or descending sort order.

Syntax

ORDER BY <expression> [ ASC | DESC ] [ NULLS FIRST | NULLS LAST] [, ...]

LIMIT

The LIMIT clause restricts the number of rows that are included in the result set.

Syntax

LIMIT <count>

OFFSET

The OFFSET clause specifies a non-negative number of rows that are skipped before returning results from the query.

Syntax

OFFSET <start>