Queries (SQL)

SELECT Statement Syntax
1
2
3
4
5
6
7
8
SELECT [DISTINCT | TOP <n>] <column expression list>
FROM [<schema name>.]<table name>
    [<join type> JOIN <join table name> ON <join expression>],...
[WHERE <filtering expression list>]
[GROUP BY <grouping expression list>]
[HAVING <group filtering expression list>]
[ORDER BY <ordering expression list>]
[LIMIT [<offset>, ]<num rows>]

Note

  • The * can be used to specify all columns in the column expression list, while <table name>.* can be used to specify all columns from the given table.

  • Columns can be double-quoted to use reserved words as column names, e.g., "PERCENT"; or to use special characters in column names, e.g., "key:value".

  • TOP <n> returns the first n records (up to 20000 records by default), but is configurable.

  • The grouping expression list may contain column names, aliases, expressions, or positions (e.g., GROUP BY 2 to aggregate on the 2nd column in the SELECT list).

  • The having expression list may contain grouping expressions or any grouping expression aliases defined in the SELECT list.

  • The ordering expression list may contain column names, expressions, or column positions (e.g., GROUP BY 2 to aggregate on the 2nd column in the SELECT list). The default ordering is ASC. The default null ordering is NULLS FIRST when using ascending order and NULLS LAST when using descending order. The general format for each comma-separated ordering expression in the list is:

    ORDER BY Expression Syntax
    1
    
    <column name/alias/expression/position> [ASC | DESC] [NULLS FIRST | NULLS LAST]
    
  • LIMIT applies paging to the result set, starting at the 0-based offset (if specified) and returning num rows records.

SELECT Example
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
SELECT
    e.last_name || ', ' || e.first_name AS "Employee_Name",
    m.last_name || ', ' || m.first_name AS "Manager_Name"
FROM
    example.employee e
    LEFT JOIN example.employee m ON e.manager_id = m.id
WHERE
    e.dept_id IN (1, 2, 3)
ORDER BY
    m.id ASC NULLS FIRST,
    e.hire_date

Tableless Query

A query without a FROM clause can be used to return a single row of data containing a constant or expression.

For example, to select the current day of the week:

Tableless Query Example
1
SELECT DAYNAME(NOW()) AS "Today"

Note

A tableless query will create a result set backed by a replicated table, by default.

Join

The supported join types are:

  • INNER - matching rows between two tables
  • LEFT - matching rows between two tables, and rows in the left-hand table with no matching rows in the right-hand table
  • RIGHT - matching rows between two tables, and rows in the right-hand table with no matching rows in the left-hand table
  • FULL OUTER matching rows between two tables, and rows in both tables with no matching rows in the other
  • CROSS - all rows in one table matched against all rows in the other

There are two execution schemes that are used to process joins, depending on the distribution of the joined tables:

  • Local - highly performant, but native join criteria must be met
  • Distributed - highly flexible, as native join restrictions are lifted, but less performant due to interprocessor communication overhead

Important

Though the data distribution restrictions on native joins do not exist for joins made via SQL, following the join guidelines on sharding will result in much more performant queries.

Kinetica supports both JOIN...ON and WHERE clause syntax for inner joins; all outer join types (LEFT, RIGHT, & FULL OUTER) require JOIN...ON syntax.

For example, to list the name of each employee and the name of the employee's manager, using the WHERE clause to specify the join condition:

JOIN via WHERE Clause Example
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
SELECT
    e.last_name || ', ' || e.first_name AS "Employee_Name",
    m.last_name || ', ' || m.first_name AS "Manager_Name"
FROM
    example.employee e,
    example.employee m
WHERE
    e.manager_id = m.id
ORDER BY
    e.last_name,
    e.first_name

To list the name of each employee and the associated manager, even for employees that don't have a manager, using the JOIN...ON syntax to specify the join condition:

JOIN via FROM Clause Example
1
2
3
4
5
6
7
8
9
SELECT
    e.last_name || ', ' || e.first_name AS "Employee_Name",
    m.last_name || ', ' || m.first_name AS "Manager_Name"
FROM
    example.employee e
    LEFT JOIN example.employee m ON e.manager_id = m.id
ORDER BY
    e.last_name,
    e.first_name

ASOF

Kinetica supports the notion of an inexact match join via the ASOF join function. This feature allows each left-side table record to be matched to a single right-side table record whose join column value is the smallest or largest value within a range relative to the left-side join column value. In the case where multiple right-side table records have the same smallest or largest value for a given left-side table record, only one of the right-side table records will be chosen and returned as part of the join.

ASOF Syntax
1
ASOF(<left_column>, <right_column>, <rel_range_begin>, <rel_range_end>, <MIN|MAX>)

The five parameters are:

  • left_column - name of the column to join on from the left-side table
  • right_column - name of the column to join on from the right-side table
  • rel_range_begin - constant value defining the position, relative to each left-side column value, of the beginning of the range in which to match right-side column values; use a negative constant to begin the range before the left-side column value, or a positive one to begin after it
  • rel_range_end - constant value defining the position, relative to each left-side column value, of the end of the range in which to match right-side column values; use a negative constant to end the range before the left-side column value, or a positive one to end after it
  • MIN|MAX - use MIN to return the right-side matched record with the smallest join column value; use MAX to return the right-side matched record with the greatest join column value

Effectively, each matched right-side column value must be:

  • >= <left-side column value> + rel_range_begin
  • <= <left-side column value> + rel_range_end

Within the set of right-side matches for each left-side record, the one with the MIN or MAX column value will be returned in the join. In the case of a tie for the MIN or MAX column value, only one right-side record will be selected for return in the join for that left-side record.

Examples

The following ASOF call might be used to list, for each flight arrival time, the soonest flight departure time that occurs between half an hour and an hour and a half after the arrival; effectively, the time-matching portion of a connecting flight query:

ASOF Time-Based Expression Example
1
ASOF(inbound.eta, outbound.etd, INTERVAL '30' MINUTE, INTERVAL '90' MINUTE, MIN)

This ASOF call returns right-side locations that are nearest eastward to each left-side location, for locations within 5 degrees of the left-side:

ASOF Distance-Based Expression Example
1
ASOF(b.x, n.x, .00001, 5, MIN)

For example, to match a set of stock trades to the opening prices for those stocks (if an opening price record exists within 24 hours prior to the trade), and to include trades for which there is no opening stock price record:

ASOF JOIN Example
1
2
3
4
5
6
7
8
9
SELECT
    t.id,
    t.dt AS execution_dt,
    q.open_dt AS quote_dt,
    t.price AS execution_price,
    q.open_price
FROM
    example.trades t
    LEFT JOIN example.quotes q ON t.ticker = q.symbol AND ASOF(t.dt, q.open_dt, INTERVAL '-1' DAY, INTERVAL '0' DAY, MAX)

While the ASOF join function can only be used as part of a join, it can effectively be made into a filter condition by sub-selecting the filter criteria in the FROM clause and joining on that criteria.

For instance, to look up the stock price for a given company as of a given date:

ASOF JOIN with Filter Example
1
2
3
4
5
6
7
8
9
SELECT
    /* KI_HINT_NO_LATE_MATERIALIZATION */
    t.ticker,
    t.asof_dt,
    q.open_dt,
    q.open_price
FROM
    (SELECT 'EBAY' AS ticker, DATETIME('2006-12-15 12:34:56') AS asof_dt) t
    LEFT JOIN example.quotes q ON t.ticker = q.symbol AND ASOF(t.asof_dt, q.open_dt, INTERVAL '-1' DAY, INTERVAL '0' DAY, MAX)

Important

The use of the KI_HINT_NO_LATE_MATERIALIZATION is key, here, as the join requires a materialized table, which this hint ensures, to succeed.

Aggregation

The GROUP BY clause can be used to segment data into groups and apply aggregate functions over the values within each group. Aggregation functions applied to data without a GROUP BY clause will be applied over the entire result set.

Note

GROUP BY can operate on columns, column expressions, column aliases, or the position of a member of the SELECT clause (where 1 is the first element).

For example, to find the average cab fare from the taxi data set:

Aggregation without GROUP BY Example
1
2
SELECT ROUND(AVG(total_amount),2) AS "Average_Fare"
FROM demo.nyctaxi

To find the minimum, maximum, & average trip distances, as well as the average passenger count for each vendor per year from the taxi data set (weeding out data with errant trip distances):

Aggregate with GROUP BY Example
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
SELECT
    vendor_id AS Vendor_ID,
    YEAR(pickup_datetime) AS Year,
    MAX(trip_distance) AS Max_Trip,
    MIN(trip_distance) AS Min_Trip,
    ROUND(AVG(trip_distance),2) AS Avg_Trip,
    INT(AVG(passenger_count)) AS Avg_Passenger_Count
FROM demo.nyctaxi
WHERE
    trip_distance > 0 AND
    trip_distance < 100
GROUP BY vendor_id, 2
ORDER BY Vendor_ID, Year

Grouping

The GROUP BY clause can also be used to apply the following grouping functions over the values within each group:

With each of these, the GROUPING() aggregate function can be used to distinguish aggregated null values in the data from null values generated by the ROLLUP, CUBE, or GROUPING SETS grouping function.

For instance, the following CASE will turn the aggregated null values in the Sector column into an <UNKNOWN SECTOR> group and the null value generated by the grouping function into an <ALL SECTORS> group:

GROUPING Expression Example
1
2
3
4
CASE
    WHEN (GROUPING(Sector) = 1) THEN '<ALL SECTORS>'
    ELSE NVL(Sector, '<UNKNOWN SECTOR>')
END AS Sector_Group,

ROLLUP

The ROLLUP(expr list) function calculates n + 1 aggregates for n number of columns in expr list.

For example, the following query will aggregate the average opening stock price for these groups:

  • Each market sector & stock symbol pair
  • Each market sector
  • All sectors and symbols
ROLLUP Example
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
SELECT
    CASE
        WHEN (GROUPING(Sector) = 1) THEN '<ALL SECTORS>'
        ELSE NVL(Sector, '<UNKNOWN SECTOR>')
    END AS Sector_Group,
    CASE
        WHEN (GROUPING(Symbol) = 1) THEN '<ALL SYMBOLS>'
        ELSE NVL(Symbol, '<UNKNOWN SYMBOL>')
    END AS Symbol_Group,
    AVG("Open") AS AvgOpen
FROM demo.Stocks
GROUP BY ROLLUP(Sector, Symbol)
ORDER BY Sector_Group, Symbol_Group

CUBE

The CUBE(expr list) function calculates 2n aggregates for n number of columns in expr list.

For example, the following query will aggregate the average opening stock price for these groups:

  • Each market sector & stock symbol pair
  • Each market sector
  • Each stock symbol
  • All sectors and symbols
CUBE Example
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
SELECT
    CASE
        WHEN (GROUPING(Sector) = 1) THEN '<ALL SECTORS>'
        ELSE NVL(Sector, '<UNKNOWN SECTOR>')
    END AS Sector_Group,
    CASE
        WHEN (GROUPING(Symbol) = 1) THEN '<ALL SYMBOLS>'
        ELSE NVL(Symbol, '<UNKNOWN SYMBOL>')
    END AS Symbol_Group,
    AVG("Open") AS AvgOpen
FROM demo.Stocks
GROUP BY CUBE(Sector, Symbol)
ORDER BY Sector_Group, Symbol_Group

GROUPING SETS

The GROUPING SETS(expr list) function calculates aggregates for each group of columns in expr list.

For example, the following query will aggregate the average opening stock price for these groups:

  • Each market sector
  • Each stock symbol
  • All sectors and symbols
GROUPING SETS Example
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
SELECT
    CASE
        WHEN (GROUPING(Sector) = 1) THEN '<ALL SECTORS>'
        ELSE NVL(Sector, '<UNKNOWN SECTOR>')
    END AS Sector_Group,
    CASE
        WHEN (GROUPING(Symbol) = 1) THEN '<ALL SYMBOLS>'
        ELSE NVL(Symbol, '<UNKNOWN SYMBOL>')
    END AS Symbol_Group,
    AVG("Open") AS AvgOpen
FROM demo.Stocks
GROUP BY GROUPING SETS((Sector), (Symbol), ())
ORDER BY Sector_Group, Symbol_Group

Window

Window functions are available through the use of the OVER clause, which can partition rows into frames. Different types of functions can be used to aggregate data over a sliding window.

Window Function Syntax
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
SELECT
    <window function> OVER (
        [PARTITION BY <column expression list>]
        [ORDER BY <ordering expression list>]
        [
            <RANGE | ROWS>
            <
                <UNBOUNDED PRECEDING | <number> PRECEDING | CURRENT ROW | <number> FOLLOWING> |
                BETWEEN <UNBOUNDED PRECEDING | <number> PRECEDING | CURRENT ROW | <number> FOLLOWING>
                    AND <UNBOUNDED FOLLOWING | <number> PRECEDING | CURRENT ROW | <number> FOLLOWING>
            >
        ]
    ) [AS <alias>]

Note

  • If the PARTITION BY clause is not specified, the window function will be computed over the entire data set.
  • The ORDER BY clause is not required when the window function is either FIRST_VALUE() or LAST_VALUE(). These two functions are also the only ranking functions that can contain a RANGE or ROWS frame clause.

The FIRST_VALUE(), LAST_VALUE(), LEAD(), and LAG() window functions support ignoring or respecting null values in the window calculation using the IGNORE NULLS or RESPECT NULLS syntax respectively. The general format for applying this additional syntax is:

Window Null Expression Syntax
1
<window function> [<IGNORE | RESPECT> NULLS]

The default ordering of records within each partition is ASC. The default null ordering is NULLS FIRST when using ascending order and NULLS LAST when using descending order. The general format for each comma-separated ordering expression in the list is:

Window Ordering Syntax
1
<column name/alias/expression/position> [ASC | DESC] [NULLS FIRST | NULLS LAST]

Note

Only one column can be specified in the ordering expression list when using RANGE. When using ROWS, the frame is applied after any ordering; so, while several columns may appear in the order expression list, there will be only one ROWS clause following the list.

When a RANGE frame is specified, CURRENT ROW includes all peer rows (rows with the same ordering values). Thus, when the first of a set of peer rows is encountered, all associated peer rows are included in the frame (not just the first one).

In contrast, when a ROWS frame is specified, CURRENT ROW will direct that only the peer rows up to and including the current row are contained within the frame--the following peer rows will not be included.

The default frame type is RANGE BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW.

If a frame clause is specified without a BETWEEN, the clause is applied to the frame start; the frame end will still be the default of CURRENT ROW.

For example, to calculate the rolling sum of total amounts collected by each taxi vendor over the course of a given day, as well as the number of other trips that occurred within 5 minutes of each trip:

Window Rolling Sum Example
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
SELECT
    vendor_id,
    pickup_datetime,
    total_amount,
    passenger_count,
    DECIMAL
    (
        SUM(total_amount) OVER
            (
                PARTITION BY vendor_id
                ORDER BY pickup_datetime
            )
    ) AS growing_sum,
    COUNT(*) OVER
        (
            PARTITION BY vendor_id
            ORDER BY LONG(pickup_datetime)
            RANGE BETWEEN 300000 PRECEDING AND 300000 FOLLOWING
        ) AS trip_demand
FROM demo.nyctaxi
WHERE pickup_datetime >= '2015-01-01' AND pickup_datetime < '2015-01-01 02:00:00'
ORDER BY
    vendor_id,
    pickup_datetime

To calculate a 5-before and 10-after moving average of 4-passenger trip distances per vendor over the course of a given day:

Window Moving Average Example
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
SELECT
    vendor_id,
    pickup_datetime,
    trip_distance,
    AVG(trip_distance) OVER
        (
            PARTITION BY vendor_id
            ORDER BY pickup_datetime
            ROWS BETWEEN 5 PRECEDING AND 10 FOLLOWING
        ) AS local_avg_dist
FROM demo.nyctaxi
WHERE
    passenger_count = 4 AND
    pickup_datetime >= '2015-01-01' AND pickup_datetime < '2015-01-02'
ORDER BY
    vendor_id,
    pickup_datetime

To rank, by vendor, the total amounts collected from 3-passenger trips on a given day:

Window Ranking Example
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
SELECT
    vendor_id,
    pickup_datetime,
    dropoff_datetime,
    total_amount AS fare,
    RANK() OVER (PARTITION BY vendor_id ORDER BY total_amount) AS ranked_fare,
    DECIMAL(PERCENT_RANK() OVER (PARTITION BY vendor_id ORDER BY total_amount)) * 100 AS percent_ranked_fare
FROM demo.nyctaxi
WHERE
    passenger_count = 3 AND
    pickup_datetime >= '2015-01-11' AND pickup_datetime < '2015-01-12'
ORDER BY
    vendor_id,
    pickup_datetime

To compare each trip's total amount to the lowest (ignoring nulls), highest (ignoring nulls), & average total amount for 5-passenger trips for each vendor over the course of a given day:

Window FIRST_VALUE Example
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
SELECT
    vendor_id,
    pickup_datetime,
    tip_amount,
    tip_amount - 
        FIRST_VALUE(tip_amount) IGNORE NULLS OVER
            (PARTITION BY vendor_id ORDER BY tip_amount) AS lowest_amount,
    tip_amount - 
        DECIMAL
        (
            AVG(tip_amount) OVER
                (
                    PARTITION BY vendor_id
                    ORDER BY tip_amount
                    ROWS BETWEEN UNBOUNDED PRECEDING AND UNBOUNDED FOLLOWING
                )
        ) AS average_amount,
    tip_amount - 
        FIRST_VALUE(tip_amount) IGNORE NULLS OVER
            (PARTITION BY vendor_id ORDER BY tip_amount DESC) AS highest_amount
FROM demo.nyctaxi
WHERE
    passenger_count = 5 AND
    pickup_datetime >= '2015-04-17' AND pickup_datetime < '2015-04-18' AND
    tip_amount > 0 AND
    trip_distance > 0
ORDER BY
    vendor_id,
    pickup_datetime

To compare each vendor's average total amount to their average total amount within the interquartile range:

Window N-Tile Example
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
SELECT
    vendor_id,
    DECIMAL(AVG(total_amount)) AS average_total_amount,
    DECIMAL(AVG(IF(quartile IN (2,3), total_amount, null))) AS average_interq_range_total_amount
FROM
(
    SELECT
        vendor_id,
        total_amount,
        NTILE(4) OVER (PARTITION BY vendor_id ORDER BY total_amount) quartile
    FROM
        demo.nyctaxi
)
GROUP BY vendor_id
ORDER BY vendor_id

PIVOT

The PIVOT clause can be used to pivot columns, "rotating" column values into multiple columns (one for each value), creating wider and shorter denormalized tables from longer, more normalized tables.

PIVOT Syntax
1
2
3
4
5
6
<select statement>
PIVOT
(
    <aggregate expression [AS <alias>]>[,...]
    FOR <column> IN (<column list>)
)

For example, given a source table phone_number, which lists each phone number for a customer as a separate record in the table, a pivot operation can be performed, creating a single record per customer with the home, work, & cell phone numbers as separate columns.

With this data:

PIVOT Input
+--------+--------------+----------------+
| name   | phone_type   | phone_number   |
+--------+--------------+----------------+
| Jane   | Home         | 123-456-7890   |
| Jane   | Work         | 111-222-3333   |
| John   | Home         | 123-456-7890   |
| John   | Cell         | 333-222-1111   |
+--------+--------------+----------------+

The following pivot operation can be applied:

PIVOT Example
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
SELECT
    name,
    Home_Phone,
    Work_Phone,
    Cell_Phone
FROM
    example.phone_list
PIVOT
(
    MAX(phone_number) AS Phone
    FOR phone_type IN ('Home', 'Work', 'Cell')
)

The data will be pivoted into a table like this:

PIVOT Output
+--------+----------------+----------------+----------------+
| name   | Home_Phone     | Work_Phone     | Cell_Phone     |
+--------+----------------+----------------+----------------+
| Jane   | 123-456-7890   | 111-222-3333   |                |
| John   | 123-456-7890   |                | 333-222-1111   |
+--------+----------------+----------------+----------------+

UNPIVOT

The UNPIVOT clause can be used to unpivot columns, "rotating" row values into column values, creating longer, more normalized tables from shorter, more denormalized tables.

UNPIVOT Syntax
1
2
3
4
5
<select statement>
UNPIVOT
(
    <value_column> FOR <var_column> IN (<column list>)
)

For example, given a source table customer_contact, which lists the home, work, & cell phone numbers for each customer in the table, an unpivot operation can be performed, creating separate home, work, & cell phone records for each customer.

With this data:

UNPIVOT Input
+--------+----------------+----------------+----------------+
| name   | home_phone     | work_phone     | cell_phone     |
+--------+----------------+----------------+----------------+
| Jane   | 123-456-7890   | 111-222-3333   |                |
| John   | 123-456-7890   |                | 333-222-1111   |
+--------+----------------+----------------+----------------+

The following unpivot operation can be applied:

UNPIVOT Example
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
SELECT *
FROM
(
    SELECT
        name,
        Home_Phone AS Home,
        Work_Phone AS Work,
        Cell_Phone AS Cell
    FROM
        example.customer_contact
)
UNPIVOT (
    phone_number FOR phone_type IN (Home, Work, Cell)
)

The data will be unpivoted into a table like this:

UNPIVOT Output
+--------+----------------+--------------+
| name   | phone_number   | phone_type   |
+--------+----------------+--------------+
| Jane   | 123-456-7890   | Home         |
| Jane   | 111-222-3333   | Work         |
| Jane   |                | Cell         |
| John   | 123-456-7890   | Home         |
| John   |                | Work         |
| John   | 333-222-1111   | Cell         |
+--------+----------------+--------------+

Note

If the original column names can be used as the values of the unpivot key, as is, the pre-selection and renaming of those columns using a subquery in the FROM clause can be eliminated.

For example, unpivoting without aliasing the quarterly grade columns will result in those exact column names being used as the quarter values:

UNPIVOT without Aliases Example
1
2
3
4
5
6
SELECT *
FROM example.student_grade
UNPIVOT
(
    grade FOR quarter IN (q1_grade, q2_grade, q3_grade, q4_grade)
)
UNPIVOT without Aliases Output
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
+--------------+-------------+------------+
|   student_id |       grade | quarter    |
+--------------+-------------+------------+
|            1 |   80.0      | q1_grade   |
|            1 |   90.0      | q2_grade   |
|            1 |   85.0      | q3_grade   |
|            1 |   95.0      | q4_grade   |
|            2 |   82.0      | q1_grade   |
|            2 |             | q2_grade   |
|            2 |   87.0      | q3_grade   |
|            2 |   92.0      | q4_grade   |
|            3 |   73.0      | q1_grade   |
|            3 |   77.0      | q2_grade   |
|            3 |   97.0      | q3_grade   |
|            3 |             | q4_grade   |
+--------------+-------------+------------+

Set Operations

There are three types of supported set operations, each having the option of returning duplicate records in the result set by using the keyword ALL:

  • UNION [ALL] - return all records from both source data sets
  • INTERSECT [ALL] - return only records that exist in both source data sets
  • EXCEPT [ALL] - return all records that exist in the first data set, but not in the second

UNION

The UNION set operator creates a single list of records from the results of two SELECT statements. Use the ALL keyword to keep all records from both sets; omit it to remove duplicate records and form a single list of records unique between the two sets. See Limitations and Cautions for limitations.

UNION Syntax
1
2
3
<select statement>
UNION [ALL]
<select statement>

For example, given a table of lunch menu items and another table of dinner menu items, a UNION can be used to return all unique lunch & dinner menu items together, including items that are the same on both menus, but of a different price:

UNION Example
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
SELECT
    food_name,
    category,
    price
FROM
    example.lunch_menu
UNION
SELECT
    food_name,
    category,
    price
FROM
    example.dinner_menu

Note

Since the example includes price and all columns selected must match between the two sets for an item to be considered a duplicate, a lunch item that is priced differently as a dinner item would also appear in the result set.

A UNION ALL can be used to return all lunch & dinner menu items together, including duplicates:

UNION ALL Example
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
SELECT
    food_name,
    category,
    price
FROM
    example.lunch_menu
UNION ALL
SELECT
    food_name,
    category,
    price
FROM
    example.dinner_menu

INTERSECT

The INTERSECT set operator creates a single list of records that exist in both of the result sets from two SELECT statements. Use the ALL keyword to keep duplicate records that exist in both sets; omit it to remove duplicate records and form a single list of records that exist in both sets. See Limitations for limitations.

INTERSECT Syntax
1
2
3
<select statement>
INTERSECT [ALL]
<select statement>

For example, given a table of lunch menu items and another table of dinner menu items, an INTERSECT can be used to return all lunch menu items (excluding duplicates) that are also dinner items for the same price:

INTERSECT Example
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
SELECT
    food_name,
    category,
    price
FROM
    example.lunch_menu
INTERSECT
SELECT
    food_name,
    category,
    price
FROM
    example.dinner_menu

Note

Since the example includes price and all columns selected must match between the two sets for an item to be included, a lunch item that is priced differently as a dinner item would not appear in the result set.

EXCEPT

The EXCEPT set operator performs set subtraction, creating a single list of records that exist in the first SELECT statement's result set, but not in the second SELECT statement's result set. Use the ALL keyword to keep duplicate records that exist in the first set but not in the second; omit it to remove duplicate records and form a single list of records that exist in the first set but not the second. See Limitations for limitations:

EXPECT Syntax
1
2
3
<select statement>
EXCEPT [ALL]
<select statement>

For example, given a table of lunch menu items and another table of dinner menu items, an EXCEPT can be used to return all lunch menu items (excluding duplicates) that are not also dinner items for the same price:

EXPECT Example
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
SELECT
    food_name,
    category,
    price
FROM
    example.lunch_menu
EXCEPT
SELECT
    food_name,
    category,
    price
FROM
    example.dinner_menu

Note

Since the example includes price and all columns selected must match between the two sets for an item to be eliminated, a lunch item that is priced differently as a dinner item would still appear in the result set.

WITH (Common Table Expressions)

The WITH operation, also known as a Common Table Expression (CTE) creates a set of data that can be assigned table & column aliases and used one or more times in subsequent operations. The aliased set can be used within the SELECT, FROM, or WHERE clauses of a subsequent query or a subsequent CTE within the same WITH operation.

Recursive WITH operations are not supported--the aliased set cannot refer to itself. The column aliases must be unique within the WITH statement--no other column or column alias can be similarly named, for example. Also, when used in a FROM clause and given a table alias, the table alias must be preceded with AS.

A CTE can be made available to a DML or DDL statement by having the WITH statement follow the CREATE TABLE...AS, INSERT, UPDATE, or DELETE statement (not precede it).

Each CTE definition within a WITH statement is structured as follows:

CTE Syntax
1
<cte name> [(<column alias list>)] AS (<select statement>)

Here, <cte name> is the table alias given to the data set returned by the CTE, and <column alias list> is the list of aliases assigned to the columns that the CTE returns. Each column alias is matched to each column returned in the order that each were defined; e.g., for column aliases (A, B, C) used with a CTE which performs a SELECT x, y, z ..., alias A will reference column x, B will reference y, and C will reference z. If no aliases are used, the names of the source columns themselves can be used to reference the data set returned by the CTE.

Each WITH statement can contain one or more CTE definitions, followed by a SELECT statement, as shown here:

WITH Syntax
1
2
WITH <cte definition>,...
<select statement>
WITH Example
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
WITH
    dept2_emp_sal_by_mgr (manager_id, sal) AS
    (
        SELECT manager_id, salary
        FROM example.employee
        WHERE dept_id = 2
    )
SELECT
    manager_id dept2_mgr_id,
    MAX(sal) dept2_highest_emp_sal_per_mgr,
    COUNT(*) as dept2_total_emp_per_mgr
FROM dept2_emp_sal_by_mgr
GROUP BY manager_id

To apply the CTE to an INSERT statement, follow the INSERT clause with the WITH clause:

WITH in INSERT Example
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
INSERT INTO example.dept2_emp_mgr_roster (emp_first_name, emp_last_name, mgr_first_name, mgr_last_name)
WITH
    dept2_emp AS
    (
        SELECT first_name, last_name, manager_id
        FROM example.employee
        WHERE dept_id = 2
    ),
    dept2_mgr AS
    (
        SELECT first_name, last_name, id
        FROM example.employee
        WHERE dept_id = 2
    )
SELECT d2emp.first_name, d2emp.last_name, d2mgr.first_name, d2mgr.last_name
FROM
    dept2_emp as d2emp
    JOIN dept2_mgr as d2mgr ON d2emp.manager_id = d2mgr.id

Iteration

Kinetica supports iteration over each record within a data set for the purpose of creating a result set with 0 to N result records per record in the original set.

This iteration can be variable, based on some value within each record, or fixed, based on a given constant value.

The iteration is performed by joining against the virtual ITER table, as follows:

Iteration Syntax
1
2
3
SELECT *
FROM table, ITER
WHERE ITER.i < <column expression>

The <column expression> can be replaced by a constant for fixed iteration.

For example, to extract all of the individual letters from a column of words, with one record per letter extracted (using variable iteration):

Iteration Over Column Values Example
1
2
3
4
SELECT id, word, i, SUBSTR(word, i + 1, 1) AS letter
FROM example.dictionary
JOIN ITER ON i < LENGTH(word)
ORDER BY id, i;

To duplicate the set of words five times (using fixed iteration):

Iteration Over Table Example
1
2
3
4
SELECT *
FROM example.dictionary, ITER
WHERE i < 5
ORDER BY id, i;

For more detail, examples, and limitations, see Iteration.

Constants

Each data type has an associated literal constant syntax, which can be used, for instance, to insert constant data values into those columns.

Numeric Constants

Integer and floating point data types can be either single-quoted or not.

Numeric Constants Example
1
2
3
4
5
6
7
8
INSERT INTO example.numeric_types (int_type1, int_type2, float_type1, float_type2)
VALUES
(
    1,
    '2',
    3.4,
    '5.6'
)

String-Based Constants

String-based data types should be single-quoted.

String-Based Constants Example
1
2
3
4
5
6
7
8
INSERT INTO example.string_types (varchar_type, charn_type, ipv4_type, wkt_type)
VALUES
(
    'varchar value',
    'charN value',
    '12.34.56.78',
    'POINT(0 0)'
)

Binary Constants

Binary types can be represented in either of the following forms:

  • single-quoted or unquoted base-10
  • single-quoted hexadecimal
Binary Constants Example
1
2
3
4
5
INSERT INTO example.byte_types (bytes_type)
VALUES
    (12345678901234567890),
    ('12345678901234567890'),
    ('0x00AB54A98CEB1F0AD2')

Date/Time Constants

Kinetica accepts unqualified single-quoted date/time values, ANSI SQL, and ODBC escape sequences in the following formats:

Data TypeNativeANSIODBC
Date'YYYY-[M]M-[D]D'DATE 'YYYY-MM-DD'{d 'YYYY-MM-DD'}
Time'[H]H24:MI:SS.mmm'TIME 'HH24:MI:SS.mmm'{t 'HH24:MI:SS.mmm'}
DateTime'YYYY-[M]M-[D]D[T| ][H]H24:MI:SS.mmm[Z]'TIMESTAMP 'YYYY-MM-DD HH24:MI:SS.mmm'{ts 'YYYY-MM-DD HH24:MI:SS.mmm'}
Timestamp'YYYY-[M]M-[D]D [H]H24:MI:SS.mmm'TIMESTAMP 'YYYY-MM-DD HH24:MI:SS.mmm'{ts 'YYYY-MM-DD HH24:MI:SS.mmm'}
Native Date/Time Constants Example
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
INSERT INTO example.date_time_types (date_type, time_type, datetime_type, timestamp_type)
VALUES
(
    '2000-01-02',
    '12:34:56.789',
    '2000-01-02 12:34:56.789',
    '2000-01-02 12:34:56.789'
),
(
    '2000-1-02',
    '1:23:45.678',
    '2000-01-02T12:34:56.789Z',
    '2000-1-02 1:23:45.678'
)
ANSI Date/Time Constants Example
1
2
3
4
5
6
7
8
INSERT INTO example.date_time_types (date_type, time_type, datetime_type, timestamp_type)
VALUES
(
    DATE '2000-01-02',
    TIME '12:34:56.789',
    TIMESTAMP '2000-01-02 12:34:56.789',
    TIMESTAMP '2000-01-02 12:34:56.789'
)
ODBC Date/Time Constants Example
1
2
3
4
5
6
7
8
INSERT INTO example.date_time_types (date_type, time_type, datetime_type, timestamp_type)
VALUES
(
    {d '2000-01-02'},
    {t '12:34:56.789'},
    {ts '2000-01-02 12:34:56.789'},
    {ts '2000-01-02 12:34:56.789'}
)

Expressions

An expression can consist of a literal constant, a column name, or a function applied to a constant or column name. A compound expression is an operation or function applied to one or more expressions.

The following are the supported expression operators:

  • + addition
  • - subtraction
  • * multiplication
  • / division
  • () grouping
  • || string concatenation

Note

Use double quotes to specify column names in a case-sensitive manner.

Conditional Functions

Conditional functions are subject to short-circuiting to aid in error-checking.

FunctionDescription
DECODE(expr, match_a, value_a, ..., match_N, value_N[, unmatched_value])Evaluates expr: returns the first value whose corresponding match is equal to expr; returns the optional unmatched_value (or null), if no match is found
IF(expr, value_if_true, value_if_false)

Evaluates expr: if true, returns value_if_true; otherwise, if false or null, returns value_if_false

ParameterDescription
expr

any true/false condition

Note

When an integer column is used directly, this function will will interpret non-zero values as true and zero values as false.

value_if_trueany type; must be the same type as value_if_false
value_if_falseany type; must be the same type as value_if_true

CASE

The case statement acts as a scalar function, but has two more complex forms. Note that for each of these CASE statements, the value expressions must all be of the same or convertible data type.

In the first form, each WHEN is followed by a conditional expression whose corresponding THEN expression will have its value returned, if true. Control will continue through each WHEN until a match is found and the corresponding value returned; if no match is found, the value of the ELSE expression will be returned, or null, if no ELSE clause exists.

CASE without Expression Syntax
1
2
3
4
5
6
CASE
    WHEN <cond_expr_a> THEN <value_expr_a>
    ...
    WHEN <cond_expr_n> THEN <value_expr_n>
    ELSE <value_expr>
END

In the second form, the CASE expression is evaluated. A match of that result will be attempted against each WHEN expression until a match is found and the value of the corresponding THEN expression returned; if no match is found, the value of the ELSE expression will be returned, or null, if no ELSE clause exists.

CASE with Expression Syntax
1
2
3
4
5
6
CASE <expr>
    WHEN <match_expr_a> THEN <value_expr_a>
    ...
    WHEN <match_expr_n> THEN <value_expr_n>
    ELSE <value_expr>
END

Note

This second version below has greater optimization than the first.

CASE without Expression Example
1
2
3
4
5
CASE
    WHEN color = 1 THEN 'Red'
    WHEN color >= 2 THEN 'Green'
    ELSE 'Blue'
END
CASE with Expression Example
1
2
3
4
5
CASE MOD(LENGTH(text), 2)
    WHEN 0 THEN 'Even'
    WHEN 1 THEN 'Odd'
    ELSE null
END

Conversion Functions

FunctionDescription

CAST(expr AS [SQL_]<conv_type>)

or

CONVERT(expr, [SQL_]<conv_type>)

Converts expr into conv_type data type

Conversion Types:

NumericStringDate/Time
BIGINTCHAR(N)DATE
DECIMAL(P,S)VARCHAR(N)DATETIME
DOUBLE TIME
FLOAT TIMESTAMP
INTEGER TYPE_DATE
NUMERIC(P,S) TYPE_TIME
REAL TYPE_TIMESTAMP
SMALLINT  
TINYINT  
UNSIGNED BIGINT  

Note

When using the SQL_ prefix, UNSIGNED BIGINT becomes SQL_UNSIGNED_BIGINT

CHAR(expr)Returns the character associated with the ASCII code given in expr
CHAR1(expr)Converts the given expr to VARCHAR(1) type
CHAR2(expr)Converts the given expr to VARCHAR(2) type
CHAR4(expr)Converts the given expr to VARCHAR(4) type
CHAR8(expr)Converts the given expr to VARCHAR(8) type
CHAR16(expr)Converts the given expr to VARCHAR(16) type
CHAR32(expr)Converts the given expr to VARCHAR(32) type
CHAR64(expr)Converts the given expr to VARCHAR(64) type
CHAR128(expr)Converts the given expr to VARCHAR(128) type
CHAR256(expr)Converts the given expr to VARCHAR(256) type
DATE(expr)Converts expr to date (YYYY-MM-DD) format
DATETIME(expr)Converts expr to datetime (YYYY-MM-DD HH24:MI:SS.mmm) format
DECIMAL(expr)Converts the given expr to DECIMAL type
DOUBLE(expr)Converts the given expr to DOUBLE type
FLOAT(expr)Converts the given expr to REAL type
INT(expr)Converts the given expr to INTEGER type
LONG(expr)Converts the given expr to BIGINT type
TIME(expr)Converts expr to time (HH24:MI:SS) format
TIMESTAMP(expr)Converts expr to the number of milliseconds since the epoch
TO_CHAR(expr, format)

Converts the given date/time expr to a string matching the given format. The format can be a string literal or expression. Arbitrary text can be injected into the format string using double-quotes.

The returned string will be truncated at 32 characters.

Valid format codes include:

FormatDescription
YYYY4-digit year
YY2-digit year
MNumber of month in year, without leading zero [1 - 12]
MMNumber of month in year, with leading zero [01 - 12]
MON3-character abbreviation of month in title case; e.g., Jan
MONTHFull name of month; e.g., January
CNumber of day in month, without leading zero [1 - 31]
DDNumber of day in month, with leading zero [01 - 31]
DY3-character abbreviation of day of the week; e.g., Fri
DAYFull name of day of the week; e.g., Friday
HAlias for H12
HHAlias for HH12
H12Hour of day in 12-hour format, without leading zero [0 - 11]
HH12Hour of day in 12-hour format, with leading zero [00 - 11]
H24Hour of day in 24-hour format, without leading zero [0 - 23]
HH24Hour of day in 24-hour format, with leading zero [00 - 23]
MIMinute of hour [00 - 59]
SSSecond of minute [00 - 59]
AMBefore or after noon designator [AM, PM]
PMAlias for AM

Example:

Function CallTO_CHAR(DATETIME(971181296000), '"Last login: "YYYY-MM-DD')
ReturnLast login: 2000-10-10
ULONG(expr)Converts the given expr to UNSIGNED BIGINT type

Date/Time Functions

This section comprises the following functions:

Date/Time Base Functions

FunctionDescription
CURRENT_DATE()Returns the date as YYYY-MM-DD
CURRENT_DATETIME()Returns the date & time as YYYY-MM-DD HH24:MI:SS.mmm
CURRENT_TIME()Returns the time as HH24:MI:SS.mmm
CURRENT_TIMESTAMP()Returns the date & time as YYYY-MM-DD HH24:MI:SS.mmm; to return the date & time as the number of milliseconds since the epoch, pass the result of this function to LONG()
DATEADD(unit, amount, expr)

Adds the positive or negative integral amount of unit date/time intervals to the date/time in expr

The following date/time intervals are supported for unit:

ConstantDescription
'YEAR'Year is modified by interval amount (not affected by leap year, etc.)
'MONTH'Month is modified by interval amount and date adjusted if overflow/underflow occurs; day adjusted to last day of calculated month if not a valid day for that month (e.g. Apr 31st -> Apr 30th)
'DAY'Day is modified by interval amount (time not affected by daylight savings time, etc.); date is adjusted, if overflow/underflow occurs
'HOUR'Hour is modified by interval amount (time not affected by daylight savings time, etc.); date is adjusted, if overflow/underflow occurs
'MINUTE'Minute is modified by interval amount; date/time are adjusted, if overflow/underflow occurs
'SECOND'Second is modified by interval amount; date/time are adjusted, if overflow/underflow occurs
'MILLISECOND'Millisecond is modified by interval amount; date/time are adjusted, if overflow/underflow occurs
'QUARTER'Month is modified by three times the interval amount, irrespective of the number of days in the months between; day adjusting performed the same as the MONTH description, but only on final month (e.g., Jan 31st + 1 quarter will be Apr 30th, not Apr 28th because of February)
'WEEK'Day is modified by 7 times the interval amount (time not affected by daylight savings time, etc.); month & year are adjusted, if overflow/underflow occurs

Note

Any of these unit types can have a SQL_TSI_ prefix prepended to them; e.g., both DAY and SQL_TSI_DAY are valid unit types for specifying a day interval

Examples:

Function CallResult
DATEADD('YEAR', 1, '2000-10-10')2001-10-10
DATEADD('MONTH', 1, '2000-01-31')2000-02-29
DATEADD('DAY', 1, '2000-12-31')2001-01-01
DATEADD('HOUR', 12, '2000-10-10 12:34:56')2000-10-11 00:34:56.000
DATEADD('MINUTE', 1, '2000-10-10 12:34:56')2000-10-10 12:35:56.000
DATEADD('SECOND', 1, '2000-12-31 23:59:59')2001-01-01 00:00:00.000
DATEADD('MILLISECOND', 1, '2000-10-10 12:34:56')2000-10-10 12:34:56.001
DATEADD('QUARTER', 1, '2000-11-30')2001-02-28
DATEADD('WEEK', 53, '2000-01-01')2001-01-06
DATEDIFF(expr_begin, expr_end)Determines the difference between two dates, irrespective of time component, as the number of days when expr_begin is subtracted from expr_end; returns a negative number of days if expr_begin occurs after expr_end
DAY(expr)Alias for DAYOFMONTH(expr)
DAYNAME(expr)Extracts the day of the week from expr and converts it to the corresponding day name [Sunday - Saturday ]
DAYOFMONTH(expr)Extracts the day of the month from expr [1 - 31]
DAYOFWEEK(expr)

Extracts the day of the week from expr [1 - 7]

Expression ValueResult
Date on Sunday1
Date on Monday2
Date on Tuesday3
Date on Wednesday4
Date on Thursday5
Date on Friday6
Date on Saturday7
DAY_OF_WEEK(expr)Alias for DAYOFWEEK(expr)
DAYOFYEAR(expr)Extracts the day of the year from expr [1 - 366]
DAY_OF_YEAR(expr)Alias for DAYOFYEAR(expr)
HOUR(expr)Extracts the hour of the day from expr [0 - 23]

<expr> + INTERVAL '<amount>' <part>

<expr> - INTERVAL '<amount>' <part>

Adds to or subtracts from the date/time expr the integral amount units of type part. This mirrors the behavior of the TIMESTAMPADD function, only with a different format and different date/time part constants. The following date/time constants are supported for part:

ConstantDescription
YEARYear is modified by interval amount (not affected by leap year, etc.)
MONTHMonth is modified by interval amount and date adjusted if overflow/underflow occurs; day adjusted to last day of calculated month if not a valid day for that month (e.g. Apr 31st -> Apr 30th)
DAYDay is modified by interval amount (time not affected by daylight savings time, etc.); date is adjusted, if overflow/underflow occurs
HOURHour is modified by interval amount (time not affected by daylight savings time, etc.); date is adjusted, if overflow/underflow occurs
MINUTEMinute is modified by interval amount; date/time are adjusted, if overflow/underflow occurs
SECONDSecond is modified by interval amount; date/time are adjusted, if overflow/underflow occurs
LAST_DAY(date)Returns the date of the last day of the month in the given date
MINUTE(expr)Extracts the minute of the day from expr [0 - 59]
MONTH(expr)Extracts the month of the year from expr [1 - 12]
MONTHNAME(expr)Extracts the month of the year from expr and converts it to the corresponding month name [January - December]
MSEC(expr)Extracts the millisecond of the second from expr [0 - 999]
NEXT_DAY(date, day_of_week)

Returns the date of the next day of the week, provided as a day name in day_of_week, that occurs after the given date

Some examples, given that 2000-10-10 is a Tuesday:

Function CallResult
NEXT_DAY('2000-10-10', 'Wednesday')2000-10-11
NEXT_DAY('2000-10-10', 'Friday')2000-10-13
NEXT_DAY('2000-10-10', 'Tuesday')2000-10-17
NOW()Alias for CURRENT_DATETIME()
QUARTER(expr)

Extracts the quarter of the year from expr [1 - 4]

Expression ValueResult
Date in January, February, or March1
Date in April, May, or June2
Date in July, August, or September3
Date in October, November, or December4
SECOND(expr)Extracts the seconds of the minute from expr [0 - 59]
SEC(expr)Alias for SECOND(expr)
TIMESTAMPADD(unit, amount, expr)

Adds the positive or negative integral amount of unit date/time intervals to the date/time in expr

The following date/time intervals are supported for unit:

ConstantDescription
YEARYear is modified by interval amount (not affected by leap year, etc.)
MONTHMonth is modified by interval amount and date adjusted if overflow/underflow occurs; day adjusted to last day of calculated month if not a valid day for that month (e.g. Apr 31st -> Apr 30th)
DAYDay is modified by interval amount (time not affected by daylight savings time, etc.); date is adjusted, if overflow/underflow occurs
HOURHour is modified by interval amount (time not affected by daylight savings time, etc.); date is adjusted, if overflow/underflow occurs
MINUTEMinute is modified by interval amount; date/time are adjusted, if overflow/underflow occurs
SECONDSecond is modified by interval amount; date/time are adjusted, if overflow/underflow occurs
FRAC_SECOND

Nanosecond is modified by interval amount; date/time are adjusted, if overflow/underflow occurs

Note

Time is processed to millisecond precision, so any portion of an amount with finer granularity than 1,000,000 nanoseconds will be ignored (e.g., requesting the addition of 1,234,567 nanoseconds will result in 1 millisecond actually being added)

QUARTERMonth is modified by three times the interval amount, irrespective of the number of days in the months between; day adjusting performed the same as the MONTH description, but only on final month (e.g., Jan 31st + 1 quarter will be Apr 30th, not Apr 28th because of February)
WEEKDay is modified by 7 times the interval amount (time not affected by daylight savings time, etc.); month & year are adjusted, if overflow/underflow occurs

Note

Any of these unit types can have a SQL_TSI_ prefix prepended to them; e.g., both DAY and SQL_TSI_DAY are valid unit types for specifying a day interval

Examples:

Function CallResult
TIMESTAMPADD(SQL_TSI_YEAR, 1, '2000-10-10')2001-10-10
TIMESTAMPADD(SQL_TSI_MONTH, 1, '2000-01-31')2000-02-29
TIMESTAMPADD(SQL_TSI_DAY, 1, '2000-12-31')2001-01-01
TIMESTAMPADD(SQL_TSI_HOUR, 12, '2000-10-10 12:34:56')2000-10-11 00:34:56.000
TIMESTAMPADD(SQL_TSI_MINUTE, 1, '2000-10-10 12:34:56')2000-10-10 12:35:56.000
TIMESTAMPADD(SQL_TSI_SECOND, 1, '2000-12-31 23:59:59')2001-01-01 00:00:00.000
TIMESTAMPADD(SQL_TSI_FRAC_SECOND, 1000000, '2000-10-10 12:34:56')2000-10-10 12:34:56.001
TIMESTAMPADD(SQL_TSI_QUARTER, 1, '2000-11-30')2001-02-28
TIMESTAMPADD(SQL_TSI_WEEK, 53, '2000-01-01')2001-01-06
TIMESTAMPDIFF(unit, begin, end)

Calculates the difference between two date/time expressions, returning the result as an integral difference in the units specified; more precisely, how many whole date/time intervals of type unit need to be added to (or subtracted from) begin to equal end (or get as close as possible without going past it) using the unit types and and rules specified in TIMESTAMPADD.

Note

This is not symmetric with TIMESTAMPADD in all cases, as adding 1 MONTH to Mar 31st results in Apr 30th, but the TIMESTAMPDIFF in MONTH units between those two dates is 0.

Examples:

Function CallResult
TIMESTAMPDIFF(MONTH, DATE('2000-10-10'), DATE('2000-12-31'))2
TIMESTAMPDIFF(MONTH, DATE('2000-03-31'), DATE('2000-04-30'))0
TIMESTAMPDIFF(MONTH, DATE('2000-12-31'), DATE('2000-10-10'))-2
TIMESTAMPDIFF(HOUR, 978222896000, DATETIME('2000-10-10 12:34:56'))-1956
WEEK(expr)Extracts the week of the year from expr [1 - 54]; each full week starts on Sunday (a 1 is returned for the week containing Jan 1st)
YEAR(expr)Extracts the year from expr; 4-digit year, A.D.

Date/Time Complex Conversion Functions

FunctionDescription
DATE_TO_EPOCH_MSECS(year, month, day, hour, min, sec, msec)

Converts the full date to milliseconds since the epoch; negative values are accepted

Example:

Function CallDATE_TO_EPOCH_MSECS(2017, 06, -15, 09, 22, 15, 42)
Return1494926535042
Resolves ToTuesday, May 16, 2017 9:22:15.042 AM
DATE_TO_EPOCH_SECS(year, month, day, hour, min, sec)

Converts the full date to seconds since the epoch; negative values are accepted

Example:

Function CallDATE_TO_EPOCH_SECS(2017, 06, -15, 09, 22, 15)
Return1494926535
Resolves ToTuesday, May 16, 2017 9:22:15 AM
MSECS_SINCE_EPOCH(timestamp)

Converts the timestamp to milliseconds since the epoch

Example:

Function CallMSECS_SINCE_EPOCH('2000-10-10 12:34:56.789')
Return971181296789
TIMESTAMP_FROM_DATE_TIME(date, time)

Converts the given date and time to a composite date/time format

Example:

Function CallTIMESTAMP_FROM_DATE_TIME('2017-06-15', '10:37:30')
Return2017-06-15 10:37:30.000
WEEK_TO_EPOCH_MSECS(year, week_number)

Converts the year and week number to milliseconds since the epoch; negative values are accepted

Example:

Function CallWEEK_TO_EPOCH_MSECS(2017,-32)
Return1463270400000
Resolves ToSunday, May 15, 2016 12:00:00 AM
WEEK_TO_EPOCH_SECS(year, week_number)

Converts the year and week number to seconds since the epoch. Negative values are accepted. Each new week begins Sunday at midnight.

Example:

Function CallWEEK_TO_EPOCH_SECS(2017,-32)
Return1463270400
Resolves ToSunday, May 15, 2016 12:00:00 AM

Error-Checking Functions

Errors will only be successfully caught if the expressions are floating-point types (i.e. double and float).

FunctionDescription
IFERROR(expr, val)Alias for IF_ERROR(expr, val)
IFINF(expr, val)Alias for IF_INF(expr, val)
IFINFINITY(expr, val)Alias for IF_INF(expr, val)
IFNAN(expr, val)Alias for IF_NAN(expr, val)
IF_ERROR(expr, val)

Evaluates the given expr and if it resolves to infinity or NaN, return val

Tip

Conceptually, this function is the same as IF_INF(IF_NAN(expr, val), val)

Example:

Function CallResult
IF_ERROR((double(10)/0), 1)1.0
IF_ERROR(log(-1), 1)1.0
IF_INF(expr, val)

Evaluates the given expr and if it resolves to infinity, return val

Example:

Function CallResult
IF_INF((double(10)/0), 999)999.0
IF_INF(log(-1), 999)NaN
IF_INFINITY(expr, val)Alias for IF_INF(expr, val)
IF_NAN(expr, val)

Evaluates the given expr and if it resolves to NaN, return val

Example:

Function CallResult
IF_NAN((double(10)/0), -1)Infinity
IF_NAN(log(-1), -1)-1.0

Geospatial/Geometry Functions

Tip

  • Use ST_ISVALID to determine if a geometry object is valid. The functions below work best with valid geometry objects.
  • Use the REMOVE_NULLABLE function to remove any nullable column types that could result from calculating a derived column (e.g., as in Projections) using one of the functions below.

Enhanced Performance Scalar Functions

The functions below all compare x and y coordinates to geometry objects (or vice versa), thus increasing their performance in queries. Each of these functions have a geometry-to-geometry version listed in the next section.

FunctionDescription
STXY_CONTAINS(geom, x, y)Returns 1 (true) if geom contains the x and y coordinate, e.g. lies in the interior of geom. The coordinate cannot be on the boundary and also be contained because geom does not contain its boundary
STXY_CONTAINSPROPERLY(geom, x, y)Returns 1 (true) if the x and y coordinate intersects the interior of geom but not the boundary (or exterior) because geom does not contain its boundary but does contain itself
STXY_COVEREDBY(x, y, geom)Returns 1 (true) if the x and y coordinate is covered by geom
STXY_COVERS(geom, x, y)Returns 1 (true) if geom covers the x and y coordinate
STXY_DISJOINT(x, y, geom)Returns 1 (true) if the given x and y coordinate and the geometry geom do not spatially intersect.
STXY_DISTANCE(x, y, geom[, solution])

Calculates the minimum distance between the given x and y coordinate and geom using the specified solution type. Solution types available:

  • 0 (default) - Euclidean; returns 2-D Euclidean distance
  • 1 - Haversine; returns minimum sphere distance in meters
  • 2 - Vincenty; returns minimum spheroid distance in meters, more accurate than Haversine but slower performance

Note: If the x and y coordinate and geom intersect (verify using ST_INTERSECTS), the distance will always be 0.

STXY_DWITHIN(x, y, geom, distance[, solution])

Returns 1 (true) if the x and y coordinate is within the specified distance from geom using the specified solution type. Solution types available:

  • 0 (default) - Euclidean; uses degrees to calculate distance
  • 1 - Sphere; uses meters to calculate sphere distance
  • 2 - Spheroid; uses meters to calculate spheroid distance
STXY_ENVDWITHIN(x, y, geom, distance[, solution])

Returns 1 (true) if the x and y coordinate is within the specified distance from the bounding box of geom using the specified solution type. Solution types available:

  • 0 (default) - Euclidean; uses degrees to calculate distance
  • 1 - Sphere; uses meters to calculate distance
STXY_ENVINTERSECTS(x, y, geom)Returns 1 (true) if the bounding box of the given geometry geom intersects the x and y coordinate.
STXY_H3(x, y, resolution)Returns an H3 grid index, similar to a geohash, for the cell containing the x and y coordinate, with the given resolution. The higher the resolution, the more precise the index is. The resolution must be an integer between 0 and 15.
STXY_INTERSECTION(x, y, geom)Returns the shared portion between the x and y coordinate and the given geometry geom, i.e. the point itself.
STXY_INTERSECTS(x, y, geom)Returns 1 (true) if the x and y coordinate and geom intersect in 2-D.
STXY_TOUCHES(x, y, geom)Returns 1 (true) if the x and y coordinate and geometry geom have at least one point in common but their interiors do not intersect. If geom is a GEOMETRYCOLLECTION, a 0 is returned regardless if the point and geometry touch
STXY_WITHIN(x, y, geom)Returns 1 (true) if the x and y coordinate is completely inside the geom geometry i.e., not on the boundary

Scalar Functions

FunctionDescription
DIST(x1, y1, x2, y2)Computes the Euclidean distance (in degrees), i.e. SQRT( (x1-x2)*(x1-x2) + (y1-y2)*(y1-y2) ).
GEODIST(lon1, lat1, lon2, lat2)Computes the geographic great-circle distance (in meters) between two lat/lon points.
GEOHASH_DECODE_LATITUDE(geohash)Decodes a given geohash and returns the latitude value for the given hash string. Supports a maximum geohash character length of 16.
GEOHASH_DECODE_LONGITUDE(geohash)Decodes a given geohash and returns the longitude value for the given hash string. Supports a maximum geohash character length of 16.
GEOHASH_ENCODE(lat, lon, precision)Encodes a given coordinate pair and returns a hash string with a given precision.
ST_ADDPOINT(linestring, point, position)Adds a given point geometry to the given linestring geometry at the specified position, which is a 0-based index.
ST_ALMOSTEQUALS(geom1, geom2, decimal)Returns 1 (true) if given geometries, geom1 and geom2, are almost spatially equal within the given amount of decimal scale. Note that geometries will still be considered equal if the decimal scale for the geometries is within a half order of magnitude of each other, e.g, if decimal is set to 2, then POINT(63.4 123.45) and POINT(63.4 123.454) are equal, but POINT(63.4 123.45) and POINT(63.4 123.459) are not equal. The geometry types must match to be considered equal.
ST_AREA(geom[, solution])

Returns the area of the given geometry geom if it is a POLYGON or MULTIPOLYGON using the specified solution type. Returns 0 if the input geometry type is (MULTI)POINT or (MULTI)LINESTRING. Solution types available:

  • 0 (default) - 2D Euclidean area
  • 1 - curved surface area on a sphere in square meters
  • 2 - curved surface area on a spheroid in square meters
ST_AZIMUTH(geom1, geom2)Returns the azimuth in radians defined by the segment between two POINTs, geom1 and geom2. Returns a null if the input geometry type is MULTIPOINT, (MULTI)LINESTRING, or (MULTI)POLYGON.
ST_BOUNDARY(geom)Returns the closure of the combinatorial boundary of a given geometry geom. Returns an empty geometry if geom is an empty geometry. Returns a null if geom is a GEOMETRYCOLLECTION
ST_BOUNDINGDIAGONAL(geom)Returns the diagonal of the given geometry's (geom) bounding box.
ST_BUFFER(geom, radius[, style[, solution]])

Returns a geometry that represents all points whose distance from the given geometry geom is less than or equal to the given distance radius. The radius units can be specified by the solution type (default is in degrees) and the radius is created in the provided style. The style options are specified as a list of blank-separated key-value pairs, e.g., 'quad_segs=8 endcap=round'. If an empty style list ('') is provided, the default settings will be used. The style parameter must be specified to provide a solution type.

Available style options:

  • quad_segs -- the number of segments used to approximate a quarter circle (default is 8)
  • endcap -- the endcap style of the buffer (default is round); options are round, flat (or butt), and square
  • join -- the join style of the buffer (default is round); options are round, mitre (or miter), and bevel
  • mitre_limit -- the mitre ratio limit expressed as a floating point number (miter_limit is also acceptable)

Available solution types:

  • 0 (default) - 2D Euclidean radius distance in degrees
  • 1 - curved surface radius distance on a sphere in meters
  • 2 - curved surface radius distance on a spheroid in meters

Tip

To create a 5-meter buffer around geom using the default styles: ST_BUFFER(geom, 5, '', 1). To create a 5-foot (converting feet to meters) buffer around geom using the following styles: ST_BUFFER(geom, 5*0.3048,'quad_segs=4 endcap=flat', 1)

ST_CENTROID(geom)Calculates the center of the given geometry geom as a POINT. For (MULTI)POINTs, the center is calculated as the average of the input coordinates. For (MULTI)LINESTRINGs, the center is calculated as the weighted length of each given LINESTRING. For (MULTI)POLYGONs, the center is calculated as the weighted area of each given POLYGON. If geom is an empty geometry, an empty GEOMETRYCOLLECTION is returned
ST_CLIP(geom1, geom2)Returns the geometry shared between given geometries geom1 and geom2
ST_CLOSESTPOINT(geom1, geom2[, solution])

Calculates the 2-D POINT in geom1 that is closest to geom2 using the specified solution type. If geom1 or geom2 is empty, a null is returned. Solution types available:

  • 0 (default) - Euclidean; calculates the closest point using 2-D Euclidean distance
  • 1 - Haversine; calculates the closest point using sphere distance in meters
  • 2 - Vincenty; returns minimum spheroid distance in meters, more accurate than Haversine but slower performance
ST_COLLECT(geom1, geom2)Returns a MULTI* or GEOMETRYCOLLECTION comprising geom1 and geom2. If geom1 and geom2 are the same, singular geometry type, a MULTI* is returned, e.g., if geom1 and geom2 are both POINTs (empty or no), a MULTIPOINT is returned. If geom1 and geom2 are neither the same type nor singular geometries, a GEOMETRYCOLLECTION is returned.
ST_COLLECTIONEXTRACT(collection, type)

Returns only the specified type from the given geometry collection. Type is a number that maps to the following:

  • 1 = POINT
  • 2 = LINESTRING
  • 3 = POLYGON
ST_COLLECTIONHOMOGENIZE(collection)Returns the simplest form of the given collection, e.g., a collection with a single POINT will be returned as POINT(x y), and a collection with multiple individual points will be returned as a MULTIPOINT.
ST_CONCAVEHULL(geom, target_percent[, allow_holes])Returns a potentially concave geometry that encloses all geometries found in the given geom set. Use target_percent (values between 0 and 1) to determine the percent of area of a convex hull the concave hull will attempt to fill; 1 will return the same geometry as an ST_CONVEXHULL operation. Set allow_holes to 1 (true) to allow holes in the resulting geometry; default value is 0 (false). Note that allow_holes is independent of the area of target_percent.
ST_CONTAINS(geom1, geom2)Returns 1 (true) if no points of geom2 lie in the exterior of geom1 and at least one point of geom2 lies in the interior of geom1. Note that geom1 does not contain its boundary but does contain itself.
ST_CONTAINSPROPERLY(geom1, geom2)Returns 1 (true) if geom2 intersects the interior of geom1 but not the boundary (or exterior). Note that geom1 does not contain its boundary but does contain itself.
ST_CONVEXHULL(geom)Returns the minimum convex geometry that encloses all geometries in the given geom set.
ST_COORDDIM(geom)Returns the coordinate dimension of the given geom, e.g., a geometry with x, y, and z coordinates would return 3.
ST_COVEREDBY(geom1, geom2)Returns 1 (true) if no point in geom1 is outside geom2.
ST_COVERS(geom1, geom2)Returns 1 (true) if no point in geom2 is outside geom1.
ST_CROSSES(geom1, geom2)Returns 1 (true) if the given geometries, geom1 and geom2, spatially cross, meaning some but not all interior points in common. If geom1 and/or geom2 are a GEOMETRYCOLLECTION, a 0 is returned regardless if the two geometries cross
ST_DIFFERENCE(geom1, geom2)Returns a geometry that represents the part of geom1 that does not intersect with geom2.
ST_DIMENSION(geom)Returns the dimension of the given geometry geom, which is less than or equal to the coordinate dimension. If geom is a single geometry, a 0 is for POINT, a 1 is for LINESTRING, and a 2 is for POLYGON. If geom is a collection, it will return the largest dimension from the collection. If geom is empty, 0 is returned.
ST_DISJOINT(geom1, geom2)Returns 1 (true) if the given geometries, geom1 and geom2, do not spatially intersect.
ST_DISTANCE(geom1, geom2[, solution])

Calculates the minimum distance between the given geometries, geom1 and geom2, using the specified solution type. Solution types available:

  • 0 (default) - Euclidean; returns 2-D Euclidean distance
  • 1 - Haversine; returns minimum sphere distance in meters
  • 2 - Vincenty; returns minimum spheroid distance in meters, more accurate than Haversine but slower performance

Note: If geom1 and geom2 intersect (verify using ST_INTERSECTS), the distance will always be 0.

ST_DISTANCEPOINTS(x1, y1, x2, y2[, solution])

Calculates the minimum distance between the given points, x1, y1 and x2, y2, using the specified solution type. Solution types available:

  • 0 (default) - Euclidean; returns 2-D Euclidean distance
  • 1 - Haversine; returns minimum sphere distance in meters
  • 2 - Vincenty; returns minimum spheroid distance in meters, more accurate than Haversine but slower performance
ST_DFULLYWITHIN(geom1, geom2, distance[, solution])

Returns 1 (true) if the maximum distance between geometries geom1 and geom2 is less than or equal to the specified distance of each other using the specified solution type. If geom1 or geom2 is null, 0 (false) is returned. Solution types available:

  • 0 (default) - Euclidean; uses degrees to calculate distance
  • 1 - Sphere; uses meters to calculate distance
  • 2 - Spheroid; uses meters to calculate distance, more accurate than sphere but slower performance
ST_DWITHIN(geom1, geom2, distance[, solution])

Returns 1 (true) if the minimum distance between geometries geom1 and geom2 is within the specified distance of each other using the specified solution type. Solution types available:

  • 0 (default) - Euclidean; uses degrees to calculate distance
  • 1 - Sphere; uses meters to calculate distance
  • 2 - Spheroid; uses meters to calculate distance, more accurate than sphere but slower performance
ST_ELLIPSE(centerx, centery, height, width)

Returns an ellipse using the following values:

  • centerx -- the x coordinate or longitude used to center the ellipse
  • centery -- the y coordinate or latitude used to center the ellipse
  • height -- the height of the ellipse (in degrees)
  • width -- the width of the ellipse (in degrees)
ST_ENDPOINT(geom)Returns the last point of the given geom as a POINT if it's a LINESTRING. If geom is not a a LINESTRING, null is returned.
ST_ENVDWITHIN(geom1, geom2, distance[, solution])

Returns 1 (true) if geom1 is within the specified distance of the bounding box of geom2 using the specified solution type. Solution types available:

  • 0 (default) - Euclidean; uses degrees to calculate distance
  • 1 - Sphere; uses meters to calculate distance
ST_ENVELOPE(geom)Returns the bounding box of a given geometry geom.
ST_ENVINTERSECTS(geom1, geom2)Returns 1 (true) if the bounding box of the given geometries, geom1 and geom2, intersect.
ST_EQUALS(geom1, geom2)Returns 1 (true) if the given geometries, geom1 and geom2, are spatially equal. Note that order does not matter.
ST_EQUALSEXACT(geom1, geom2, tolerance)Returns 1 (true) if the given geometries, geom1 and geom2, are almost spatially equal within some given tolerance. If the values within the given geometries are within the tolerance value of each other, they're considered equal, e.g., if tolerance is 2, POINT(1 1) and POINT(1 3) are considered equal, but POINT(1 1) and POINT(1 3.1) are not. Note that the geometry types have to match for them to be considered equal.
ST_ERASE(geom1, geom2)Returns the result of erasing a portion of geom1 equal to the size of geom2.
ST_EXPAND(geom, units)Returns the bounding box expanded in all directions by the given units of the given geom. The expansion can also be defined for separate directions by providing separate parameters for each direction, e.g., ST_EXPAND(geom, unitsx, unitsy, unitsz, unitsm).
ST_EXPANDBYRATE(geom, rate)Returns the bounding box expanded by a given rate (a ratio of width and height) for the given geometry geom. The rate must be between 0 and 1.
ST_EXTERIORRING(geom)Returns a LINESTRING representing the exterior ring of the given POLYGON geom
ST_FORCE2D(geom)Returns the 2-dimensional version (e.g., X and Y coordinates) of geom, the provided geometry or set of geometries (e.g., via GEOMETRYCOLLECTION or WKT column name).
ST_FORCE3D(geom[, z])

Returns the 3-dimensional version (e.g., X, Y, and Z coordinates) of geom, a provided geometry or set of geometries (e.g., via GEOMETRYCOLLECTION or WKT column name), using z as the geometry's new z-value. The provided z-values can also be derived from a numeric column. If no z is provided, a 0 will be applied.

Note

If a WKT column is provided for geom and a numeric column is provided for z, the z values will be matched to the provided geometries by row in the source table. If a singular geometry is provided for geom and a column is provided for z, three-dimensional versions of the provided geometry will be returned for each z value found in the provided z column. If columns are provided for both geom and z and nulls are present in either column, the row containing null values will be skipped in the results.

ST_GENERATEPOINTS(geom, num)Creates a MULTIPOINT containing a number num of randomly generated points within the boundary of geom.
ST_GEOHASH(geom, precision)

Returns a hash string representation of the given geometry geom with specified precision (the length of the geohash string). The longer the precision, the more precise the hash is. By default, precision is set to 20; the max for precision is 32. Returns null if geom is an empty geometry.

Note

The value returned will not be a geohash of the exact geometry but a geohash of the centroid of the given geometry

ST_GEOMETRYN(geom, index)Returns the index geometry back from the given geom geometry. The index starts from 1 to the number of geometry in geom.
ST_GEOMETRYTYPE(geom)Returns the type of geometry from the given geom.
ST_GEOMETRYTYPEID(geom)

Returns the type ID of from geom. Type and ID mappings:

  • POINT = 0
  • LINESTRING = 1
  • POLYGON = 3
  • MULTIPOINT = 4
  • MULTILINESTRING = 5
  • MULTIPOLYGON = 6
  • GEOMETRYCOLLECTION = 7
ST_GEOMFROMGEOHASH(geohash, precision)Returns a POLYGON boundary box using the given geohash with a precision set by the integer precision. If precision is specified, the function will use as many characters in the hash equal to precision to create the geometry. If no precision is specified, the full length of the geohash is used.
ST_GEOMFROMH3(h3_index)Returns a POLYGON boundary box of the hex grid identified by the given h3_index.
ST_GEOMFROMTEXT(wkt)Returns a geometry from the given Well-Known text representation wkt. Note that this function is only compatible with constants.
ST_H3(wkt, resolution)Returns an H3 grid index, similar to a geohash, for the cell containing the centroid of the wkt, with the given resolution. The higher the resolution, the more precise the index is. The resolution must be an integer between 0 and 15.
ST_HEXGRID(xmin, ymin, xmax, ymax, cell_side[, limit])

Creates a MULTIPOLYGON containing a grid of hexagons between given minimum and maximum points of a bounding box. The minimum point cannot be greater than or equal to the maximum point. The size (in meters) of the individual hexagons' sides is determined by cell_side. The cell_side cannot be greater than the width or height of the bounding box. The maximum number of cells that can be produced is determined by limit, a positive integer. Supported values for limit:

  • -1 - No limit to the number of cells generated (effectively limited by system memory)
  • 0 (default) - 100 million cells
  • <n> - Custom limit of n cells

If the custom limit request specifies more cells (based on the bounding box and the cell_side) than the system limit, a null is returned.

ST_INTERIORRINGN(geom, n)Returns the n-th interior LINESTRING ring of the POLYGON geom. If geom is not a POLYGON or the given n is out of range, a null is returned. The index begins at 1
ST_INTERSECTION(geom1, geom2)Returns the shared portion between given geometries geom1 and geom2
ST_INTERSECTS(geom1, geom2)Returns 1 (true) if the given geometries, geom1 and geom2, intersect in 2-D
ST_ISCLOSED(geom)Returns 1 (true) if the given geometry's (geom) start and end points coincide
ST_ISCOLLECTION(geom)Returns 1 (true) if geom is a collection, e.g., GEOMETRYCOLLECTION, MULTIPOINT, MULTILINESTRING, etc.
ST_ISEMPTY(geom)Returns 1 (true) if geom is empty
ST_ISRING(geom)Returns 1 (true) if LINESTRING geom is both closed (per ST_ISCLOSED) and "simple" (per ST_ISSIMPLE). Returns 0 if geom is not a LINESTRING
ST_ISSIMPLE(geom)Returns 1 (true) if geom has no anomalous geometric points, e.g., self-intersection or self-tangency
ST_ISVALID(geom)Returns 1 (true) if geom (typically a [MULTI]POLYGON) is well formed. A POLYGON is valid if its rings do not cross and its boundary intersects only at POINTs (not along a line). The POLYGON must also not have dangling LINESTRINGs. A MULTIPOLYGON is valid if all of its elements are also valid and the interior rings of those elements do not intersect. Each element's boundaries may touch but only at POINTs (not along a line)
ST_LENGTH(geom[, solution])

Returns the length of the geometry if it is a LINESTRING or MULTILINESTRING. Returns 0 if another type of geometry, e.g., POINT, MULTIPOINT, etc. GEOMETRYCOLLECTIONs are also supported but the aforementioned type limitation still applies; the collection will be recursively searched for LINESTRINGs and MULTILINESTRINGs and the summation of all supported geometry types is returned (unsupported types are ignored). Solution types available:

  • 0 (default) - 2D Euclidean length
  • 1 - length on a sphere in meters
  • 2 - length on a spheroid in meters
ST_LINEFROMMULTIPOINT(geom)Creates a LINESTRING from geom if it is a MULTIPOINT. Returns null if geom is not a MULTIPOINT
ST_LINEINTERPOLATEPOINT(geom, fraction)Returns a POINT that represents the specified fraction of the LINESTRING geom. If geom is either empty or not a LINESTRING, null is returned
ST_LINELOCATEPOINT(linestring, point)Returns the location of the closest point in the given linestring to the given point as a value between 0 and 1. The return value is a fraction of the total linestring length.
ST_LINEMERGE(geom)Returns a LINESTRING or MULTILINESTRING from a given geom. If geom is a MULTILINESTRING comprising LINESTRINGs with shared endpoints, a contiguous LINESTRING is returned. If geom is a LINESTRING or a MULTILINESTRING comprising LINESTRINGS without shared endpoints, geom is returned If geom is an empty (MULTI)LINESTRING or a (MULTI)POINT or (MULTI)POLYGON, an empty GEOMETRYCOLLECTION is returned.
ST_LINESUBSTRING(geom, start_fraction, end_fraction)Returns the fraction of a given geom LINESTRING where start_fraction and end_fraction are between 0 and 1. For example, given LINESTRING(1 1, 2 2, 3 3) a start_fraction of 0 and an end_fraction of 0.25 would yield the first quarter of the given LINESTRING, or LINESTRING(1 1, 1.5 1.5). Returns null if start_fraction is greater than end_fraction. Returns null if input geometry is (MULTI)POINT, MULTILINESTRING, or (MULTI)POLYGON. Returns null if start_fraction and/or end_fraction are less than 0 or more than 1.
ST_LONGESTLINE(geom1, geom2[, solution])

Returns the LINESTRING that represents the longest line of points between the two geometries. If multiple longest lines are found, only the first line found is returned. If geom1 or geom2 is empty, null is returned. Solution types available:

  • 0 (default) - Euclidean; uses degrees to calculate the longest line
  • 1 - Sphere; uses meters to calculate the longest line
  • 2 - Spheroid; uses meters to calculate the longest line, more accurate than sphere but slower performance
ST_MAKEENVELOPE(xmin, ymin, xmax, ymax)Creates a rectangular POLYGON from the given min and max parameters
ST_MAKELINE(geom[, geom2])

Creates a LINESTRING from geom if it is a MULTIPOINT. If geom is a POINT, there must be at least one other POINT to construct a LINESTRING. If geom is a LINESTRING, it must have at least two points. Returns null if geom is not a POINT, MULTIPOINT, or LINESTRING

Note

This function can be rather costly in terms of performance

ST_MAKEPOINT(x, y)

Creates a POINT at the given coordinate

Note

This function can be rather costly in terms of performance

ST_MAKEPOLYGON(geom)

Creates a POLYGON from geom. Inputs must be closed LINESTRINGs

Note

This function can be rather costly in terms of performance

ST_MAKETRIANGLE2D(x1, y1, x2, y2, x3, y3)Creates a closed 2-D POLYGON with three vertices
ST_MAKETRIANGLE3D(x1, y1, z1, x2, y2, z2, x3, y3, z3)Creates a closed 3-D POLYGON with three vertices
ST_MAXDISTANCE(geom1, geom2[, solution])

Returns the maximum distance between the given geom1 and geom2 geometries using the specifed solution type. If geom1 or geom2 is empty, null is returned. Solution types available:

  • 0 (default) - returns maximum 2-D Euclidean distance
  • 1 - Sphere; returns maximum distance in meters
  • 2 - Spheroid; returns maximum distance in meters, more accurate than sphere but slower performance
ST_MAXX(geom)Returns the maximum x coordinate of a bounding box for the given geom geometry. This function works for 2-D and 3-D geometries.
ST_MAXY(geom)Returns the maximum y coordinate of a bounding box for the given geom geometry. This function works for 2-D and 3-D geometries.
ST_MAXZ(geom)Returns the maximum z coordinate of a bounding box for the given geom geometry. This function works for 2-D and 3-D geometries.
ST_MINX(geom)Returns the minimum x coordinate of a bounding box for the given geom geometry. This function works for 2-D and 3-D geometries.
ST_MINY(geom)Returns the minimum y coordinate of a bounding box for the given geom geometry. This function works for 2-D and 3-D geometries.
ST_MINZ(geom)Returns the minimum z coordinate of a bounding box for the given geom geometry. This function works for 2-D and 3-D geometries.
ST_MULTI(geom)Returns geom as a MULTI- geometry, e.g., a POINT would return a MULTIPOINT.
ST_MULTIPLERINGBUFFERS(geom, distance, outside)

Creates multiple buffers at specified distance around the given geom geometry. Multiple distances are specified as comma-separated values in an array, e.g., [10,20,30]. Valid values for outside are:

  • FULL -- indicates that buffers will overlap or cover the given geom geometry. This is the default.
  • OUTSIDE_ONLY -- indicates that buffers will be rings around the given geom geometry.
ST_NEAR(geom1, geom2)Returns the portion of geom2 that is closest to geom1. If geom2 is a singular geometry object (e.g., POINT, LINESTRING, POLYGON), geom2 will be returned. If geom2 a multi-geometry, e.g., MULTIPOINT, MULTILINESTRING, etc., the nearest singular geometry in geom2 will be returned.
ST_NORMALIZE(geom)Returns geom in its normalized (canonical) form, which may rearrange the points in lexicographical order.
ST_NPOINTS(geom)Returns the number of points (vertices) in geom.
ST_NUMGEOMETRIES(geom)If geom is a collection or MULTI- geometry, returns the number of geometries. If geom is a single geometry, returns 1.
ST_NUMINTERIORRINGS(geom)Returns the number of interior rings if geom is a POLYGON. Returns null if geom is anything else.
ST_NUMPOINTS(geom)Returns the number of points in the geom LINESTRING. Returns null if geom is not a LINESTRING.
ST_OVERLAPS(geom1, geom2)Returns 1 (true) if given geometries geom1 and geom2 share space. If geom1 and/or geom2 are a GEOMETRYCOLLECTION, a 0 is returned regardless if the two geometries overlap
ST_PARTITION(geom, threshold)Returns a MULTIPOLYGON representing the given geom partitioned into a number of POLYGONs with a maximum number of vertices equal to the given threshold. Minimum value for threshold is 10; default value is 10000. If geom is not a POLYGON or MULTIPOLYGON, geom is returned. If the number of vertices in geom is less than the threshold, geom is returned.
ST_PERIMETER(geom[, solution])

Returns the perimeter of the geometry if it is a POLYGON or MULTIPOLYGON. Returns 0 if another type of geometry, e.g., POINT, MULTIPOINT, LINESTRING, or MULTILINESTRING. GEOMETRYCOLLECTIONs are also supported but the aforementioned type limitation still applies; the collection will be recursively searched for POLYGONs and MULTIPOLYGONs and the summation of all supported geometry types is returned (unsupported types are ignored). Solution types available:

  • 0 (default) - 2D Euclidean length
  • 1 - length on a sphere in meters
  • 2 - length on a spheroid in meters
ST_POINT(x, y)Returns a POINT with the given x and y coordinates.
ST_POINTFROMGEOHASH(geohash, precision)

Returns a POINT using the given geohash with a precision set by the integer precision. If precision is specified, the function will use as many characters in the hash equal to precision to create the geometry. If no precision is specified, the full length of the geohash is used.

Note

The POINT returned represents the center of the bounding box of the geohash

ST_POINTGRID(xmin, ymin, xmax, ymax, cell_side[, limit])

Creates a MULTIPOLYGON containing a square-shaped grid of points between given minimum and maximum points of a bounding box. The minimum point cannot be greater than or equal to the maximum point. The distance between the points (in meters) is determined by cell_side. The cell_side cannot be greater than the width or height of the bounding box. The maximum number of cells that can be produced is determined by limit, a positive integer. Supported values for limit:

  • -1 - No limit to the number of cells generated (effectively limited by system memory)
  • 0 (default) - 100 million cells
  • <n> - Custom limit of n cells

If the custom limit request specifies more cells (based on the bounding box and the cell_side) than the system limit, a null is returned.

ST_POINTN(geom, n)Returns the n-th point in LINESTRING geom. Negative values are valid, but note that they are counted backwards from the end of geom. A null is returned if geom is not a LINESTRING.
ST_POINTS(geom)Returns a MULTIPOINT containing all of the coordinates of geom.
ST_PROJECT(geom, distance, azimuth)Returns a POINT projected from a start point geom along a geodesic calculated using distance and azimuth. If geom is not a POINT, null is returned.
ST_REMOVEPOINT(geom, offset)Remove a point from LINESTRING geom using offset to skip over POINTs in the LINESTRING. The offset is 0-based.
ST_REMOVEREPEATEDPOINTS(geom, tolerance)Removes points from geom if the point's vertices are greater than or equal to the tolerance of the previous point in the geometry's list. If geom is not a MULTIPOINT, MULTILINESTRING, or a MULTIPOLYGON, no points will be removed.
ST_REVERSE(geom)Return the geometry with its coordinate order reversed.
ST_SCALE(geom, x, y)Scales geom by multiplying its respective vertices by the given x and y values. This function also supports scaling geom using another geometry object, e.g., ST_SCALE('POINT(3 4)', 'POINT(5 6)') would return POINT(15 24). If specifying x and y for scale, note that the default value is 0, e.g., ST_SCALE('POINT(1 3)', 4) would return POINT(4 0).
ST_SEGMENTIZE(geom, max_segment_length[, solution])

Returns the given geom but segmentized n number of times depending on how the max_segment_length distance (in units based on the solution type) divides up the original geometry. The new geom is guaranteed to have segments that are smaller than the given max_segment_length. Note that POINTs are not able to be segmentized. Collection geometries (GEOMETRYCOLLECTION, MULTILINESTRING, MULTIPOINT, etc.) can be segmentized, but only the individual parts will be segmentized, not the collection as a whole. Solution types available:

  • 0 - Euclidean; uses degrees to calculate distance
  • 1 (default) - Sphere; uses meters to calculate distance
ST_SETPOINT(geom1, position, geom2)Replace a point of LINESTRING geom1 with POINT geom2 at position (base 0). Negative values are valid, but note that they are counted backwards from the end of geom.
ST_SHAREDPATH(geom1, geom2)Returns a collection containing paths shared by geom1 and geom2.
ST_SHORTESTLINE(geom1, geom2)Returns the 2-D LINESTRING that represents the shortest line of points between the two geometries. If multiple shortest lines are found, only the first line found is returned. If geom1 or geom2 is empty, null is returned
ST_SIMPLIFY(geom, tolerance)

Returns a simplified version of the given geom using an algorithm to reduce the number of points comprising a given geometry while attempting to best retain the original shape. The given tolerance determines how much to simplify the geometry. The higher the tolerance, the more simplified the returned geometry. Some holes might be removed and some invalid polygons (e.g., self-intersecting, etc.) might be present in the returned geometry. Only (MULTI)LINESTRINGs and (MULTI)POLYGONs can be simplified, including those found within GEOMETRYCOLLECTIONs; any other geometry objects will be returned unsimplified.

Note

The tolerance should be provided in the same units as the data. As a rule of thumb, a tolerance of 0.00001 would correspond to about one meter.

ST_SIMPLIFYPRESERVETOPOLOGY(geom, tolerance)

Returns a simplified version of the given geom using an algorithm to reduce the number of points comprising a given geometry while attempting to best retain the original shape. The given tolerance determines how much to simplify the geometry. The higher the tolerance, the more simplified the returned geometry. No holes will be removed and no invalid polygons (e.g., self-intersecting, etc.) will be present in the returned geometry. Only (MULTI)LINESTRINGs and (MULTI)POLYGONs can be simplified, including those found within GEOMETRYCOLLECTIONs; any other geometry objects will be returned unsimplified.

Note

The tolerance should be provided in the same units as the data. As a rule of thumb, a tolerance of 0.00001 would correspond to about one meter.

ST_SNAP(geom1, geom2, tolerance)Snaps geom1 to geom2 within the given tolerance. If the tolerance causes geom1 to not snap, the geometries will be returned unchanged.
ST_SPLIT(geom1, geom2)Returns a collection of geometries resulting from the split between geom1 and geom2 geometries.
ST_SQUAREGRID(xmin, ymin, xmax, ymax, cell_side[, limit])

Creates a MULTIPOLYGON containing a grid of squares between given minimum and maximum points of a bounding box. The minimum point cannot be greater than or equal to the maximum point. The size (in meters) of the individual squares' sides is determined by cell_side. The cell_side cannot be greater than the width or height of the bounding box. The maximum number of cells that can be produced is determined by limit, a positive integer. Supported values for limit:

  • -1 - No limit to the number of cells generated (effectively limited by system memory)
  • 0 (default) - 100 million cells
  • <n> - Custom limit of n cells

If the custom limit request specifies more cells (based on the bounding box and the cell_side) than the system limit, a null is returned.

ST_STARTPOINT(geom)Returns the first point of LINESTRING geom as a POINT. Returns null if geom is not a LINESTRING.
ST_SYMDIFFERENCE(geom1, geom2)Returns a geometry that represents the portions of geom1 and geom2 geometries that do not intersect.
ST_TOUCHES(geom1, geom2)Returns 1 (true) if the given geometries, geom1 and geom2, have at least one point in common but their interiors do not intersect. If geom1 and/or geom2 are a GEOMETRYCOLLECTION, a 0 is returned regardless if the two geometries touch
ST_TRANSLATE(geom, deltax, deltay[, deltaz])Translate geom by given offsets deltax and deltay. A z-coordinate offset can be applied using deltaz.
ST_TRIANGLEGRID(xmin, ymin, xmax, ymax, cell_side[, limit])

Creates a MULTIPOLYGON containing a grid of triangles between given minimum and maximum points of a bounding box. The minimum point cannot be greater than or equal to the maximum point. The size (in meters) of the individual triangles' sides is determined by cell_side. The cell_side cannot be greater than the width or height of the bounding box. The maximum number of cells that can be produced is determined by limit, a positive integer. Supported values for limit:

  • -1 - No limit to the number of cells generated (effectively limited by system memory)
  • 0 (default) - 100 million cells
  • <n> - Custom limit of n cells

If the custom limit request specifies more cells (based on the bounding box and the cell_side) than the system limit, a null is returned.

ST_UNION(geom1, geom2)Returns a geometry that represents the point set union of the two given geometries, geom1 and geom2.
ST_UNIONCOLLECTION(geom)Returns a geometry that represents the point set union of a single given geometry geom.
ST_UPDATE(geom1, geom2)Returns a geometry that is geom1 geometry updated by geom2 geometry
ST_VORONOIPOLYGONS(geom, tolerance)Returns a GEOMETRYCOLLECTION containing Voronoi polygons (regions consisting of points closer to a vertex in geom than any other vertices in geom) calculated from the vertices in geom and the given tolerance. The tolerance determines the distance at which points will be considered the same. An empty GEOMETRYCOLLECTION is returned if geom is an empty geometry, a single POINT, or a LINESTRING or POLYGON composed of equivalent vertices (e.g., POLYGON((0 0, 0 0, 0 0, 0 0)), LINESTRING(0 0, 0 0)).
ST_WITHIN(geom1, geom2)Returns 1 (true) if the geom1 geometry is inside the geom2 geometry. Note that as long as at least one point is inside of geom2, geom1 is considered within geom2 even if the rest of the geom1 lies along the boundary of geom2
ST_WKTTOWKB(geom)

Returns the binary form (WKB) of a geom (WKT)

Note

This function can only be used in queries against a single table.

ST_X(geom)Returns the X coordinate of the POINT geom; if the coordinate is not available, null is returned. geom must be a POINT.
ST_XMAX(geom)Alias for ST_MAXX()
ST_XMIN(geom)Alias for ST_MINX()
ST_Y(geom)Returns the Y coordinate of the POINT geom; if the coordinate is not available, null is returned. geom must be a POINT.
ST_YMAX(geom)Alias for ST_MAXY()
ST_YMIN(geom)Alias for ST_MINY()
ST_ZMAX(geom)Alias for ST_MAXZ()
ST_ZMIN(geom)Alias for ST_MINZ()

Aggregation Functions

FunctionDescription
ST_AGGREGATE_COLLECT(geom)Alias for ST_COLLECT_AGGREGATE()
ST_AGGREGATE_INTERSECTION(geom)Alias for ST_INTERSECTION_AGGREGATE()
ST_COLLECT_AGGREGATE(geom)Returns a GEOMETRYCOLLECTION comprising all geometries found in the geom set. Any MULTI* geometries will be divided into separate singular geometries, e.g., MULTIPOINT((0 0), (1 1)) would be divided into POINT(0 0) and POINT(1 1) in the results; the same is true for elements of a GEOMETRYCOLLECTION found in geom, where a GEOMETRYCOLLECTION within the provided geom set will also be parsed, effectively flattening it and adding the individual geometries to the resulting GEOMETRYCOLLECTION. Any empty geometries in geom are ignored even if they are part of a GEOMETRYCOLLECTION. Any duplicate WKTs will be retained.
ST_DISSOLVE(geom)Dissolves all geometries within a given set into a single geometry. Note that the resulting single geometry can still be a group of noncontiguous geometries but represented as a single group, e.g., a GEOMETRYCOLLECTION. Best performance when used in conjunction with adjacent geometries
ST_DISSOLVEOVERLAPPING(geom)Dissolves all geometries within a given set into a single geometry. Note that the resulting single geometry can still be a group of noncontiguous geometries but represented as a single group, e.g., a GEOMETRYCOLLECTION. Best performance when used in conjunction with overlapping geometries
ST_INTERSECTION_AGGREGATE(geom)Returns a POLYGON or MULTIPOLYGON comprising the shared portion between all geometries found in the geom set. Returns an empty GEOMETRYCOLLECTION if there is no shared portion between all geometries. Functionally equivalent to ST_INTERSECTION(ST_INTERSECTION(ST_INTERSECTION(geom1, geom2), geom3), ... geomN).
ST_LINESTRINGFROMORDEREDPOINTS(x, y, t)Returns a LINESTRING that represents a "track" of the given points (x, y) ordered by the given sort column t (e.g., a timestamp or sequence number). If any of the values in the specified columns are null, the null "point" will be left out of the resulting LINESTRING. If there's only one non-null "point" in the source table, a POINT is returned. If there are no non-null "points" in the source table, a null is returned
ST_LINESTRINGFROMORDEREDPOINTS3D(x, y, z, t)Returns a LINESTRING that represents a "track" of the given 3D points (x, y, z) ordered by the given sort column t (e.g., a timestamp or sequence number). If any of the values in the specified columns are null, the null "point" will be left out of the resulting LINESTRING. If there's only one non-null "point" in the source table, a POINT is returned. If there are no non-null "points" in the source table, a null is returned
ST_POLYGONIZE(geom)Returns a GEOMETRYCOLLECTION containing POLYGONs comprising the provided (MULTI)LINESTRING(s). (MULTI)POINT and (MULTI)POLYGON geometries are ignored when calculating the resulting GEOMETRYCOLLECTION. If a valid POLYGON cannot be constructed from the provided (MULTI)LINESTRING(s), an empty GEOMETRYCOLLECTION will be returned.

Math Functions

FunctionDescription
ABS(expr)Calculates the absolute value of expr
ACOS(expr)Returns the inverse cosine (arccosine) of expr as a double
ACOSF(expr)Returns the inverse cosine (arccosine) of expr as a float
ACOSH(expr)Returns the inverse hyperbolic cosine of expr as a double
ACOSHF(expr)Returns the inverse hyperbolic cosine of expr as a float
ASIN(expr)Returns the inverse sine (arcsine) of expr as a double
ASINF(expr)Returns the inverse sine (arcsine) of expr as a float
ASINH(expr)Returns the inverse hyperbolic sine of expr as a double
ASINHF(expr)Returns the inverse hyperbolic sine of expr as a float
ATAN(expr)Returns the inverse tangent (arctangent) of expr as a double
ATANF(expr)Returns the inverse tangent (arctangent) of expr as a float
ATANH(expr)Returns the inverse hyperbolic tangent of expr as a double
ATANHF(expr)Returns the inverse hyperbolic tangent of expr as a float
ATAN2(x, y)Returns the inverse tangent (arctangent) using two arguments as a double
ATAN2F(x, y)Returns the inverse tangent (arctangent) using two arguments as a float
ATN2(x, y)Alias for ATAN2
ATN2F(x, y)Alias for ATAN2F
CBRT(expr)Returns the cube root of expr as a double
CBRTF(expr)Returns the cube root of expr as a float
CEIL(expr)Alias for CEILING
CEILING(expr)Rounds expr up to the next highest integer
COS(expr)Returns the cosine of expr as a double
COSF(expr)Returns the cosine of expr as a float
COSH(expr)Returns the hyperbolic cosine of expr as a double
COSHF(expr)Returns the hyperbolic cosine of expr as a float
COT(expr)Returns the cotangent of expr as a double
COTF(expr)Returns the cotangent of expr as a float
DEGREES(expr)Returns the conversion of expr (in radians) to degrees as a double
DEGREESF(expr)Returns the conversion of expr (in radians) to degrees as a float
DIVZ(a, b, c)Returns the quotient a / b unless b == 0, in which case it returns c
EXP(expr)Returns e to the power of expr as a double
EXPF(expr)Returns e to the power of expr as a float
FLOOR(expr)Rounds expr down to the next lowest integer
GREATER(expr_a, expr_b)Returns whichever of expr_a and expr_b has the larger value, based on typed comparison
HYPOT(x, y)Returns the hypotenuse of x and y as a double
HYPOTF(x, y)Returns the hypotenuse of x and y as a float
ISNAN(expr)Returns 1 (true) if expr is not a number by IEEE standard; otherwise, returns 0 (false)
IS_NAN(expr)Alias for ISNAN
ISINFINITY(expr)Returns 1 (true) if expr is infinity by IEEE standard; otherwise, returns 0 (false)
IS_INFINITY(expr)Alias for ISINFINITY
LDEXP(x, exp)Returns the value of x * 2exp as a double
LDEXPF(x, exp)Returns the value of x * 2exp as a float
LESSER(expr_a, expr_b)Returns whichever of expr_a and expr_b has the smaller value, based on typed comparison
LN(expr)Returns the natural logarithm of expr as a double
LNF(expr)Returns the natural logarithm of expr as a float
LOG(expr)Alias for LN
LOG10(expr)Returns the base-10 logarithm of expr as a double
LOG10F(expr)Returns the base-10 logarithm of expr as a float
LOG1P(expr)Returns the natural logarithm of one plus expr as a double
LOG1PF(expr)Returns the natural logarithm of one plus expr as a float
LOG2(expr)Returns the binary (base-2) logarithm of expr as a double
LOG2F(expr)Returns the binary (base-2) logarithm of expr as a float
LOGF(expr)Alias for LNF
MAX_CONSECUTIVE_BITS(expr)Calculates the length of the longest series of consecutive 1 bits in the integer expr
MOD(dividend, divisor)Calculates the remainder after integer division of dividend by divisor
PI()Returns the value of pi
POW(base, exponent)Alias for POWER
POWF(base, exponent)Alias for POWERF
POWER(base, exponent)Returns base raised to the power of exponent as a double
POWERF(base, exponent)Returns base raised to the power of exponent as a float
RADIANS(expr)Returns the conversion of expr (in degrees) to radians as a double
RADIANSF(expr)Returns the conversion of expr (in degrees) to radians as a float
RAND()Returns a random floating-point value.
REGR_VALX(y, x)Returns NULL if the first argument is NULL; otherwise, returns the second argument.
REGR_VALY(y, x)Returns NULL if the second argument is NULL; otherwise, returns the first argument.
ROUND(expr, scale)

Rounds expr to the nearest decimal number with scale decimal places when scale is a positive number; rounds to the nearest number such that the result has -(scale) zeros to the left of the decimal point when scale is negative; use scale of 0 to round to the nearest integer. Examples:

Function CallResult
ROUND(12345.678, 2)12345.68
ROUND(12345.678, 0)12346
ROUND(12345.678, -2)12300
SIGN(expr)

Determines whether a number is positive, negative, or zero; returns one of the following three values:

Expression ValueResult
positive1
zero0
negative-1
SIN(expr)Returns the sine of expr as a double
SINF(expr)Returns the sine of expr as a float
SINH(expr)Returns the hyperbolic sine of expr as a double
SINHF(expr)Returns the hyperbolic sine of expr as a float
SQRT(expr)Returns the square root of expr as a double
SQRTF(expr)Returns the square root of expr as a float
TAN(expr)Returns the tangent of expr as a double
TANF(expr)Returns the tangent of expr as a float
TANH(expr)Returns the hyperbolic tangent of expr as a double
TANHF(expr)Returns the hyperbolic tangent of expr as a float
TRUNCATE(expr, scale)

Rounds expr down to the nearest decimal number with scale decimal places, following the same rules as ROUND. Examples:

Function CallResult
TRUNCATE(12345.678, 2)12345.67
TRUNCATE(12345.678, 0)12345
TRUNCATE(12345.678, -2)12300

Null Functions

Some of the following null functions require parameters to be of convertible data types. Note that limited-width (charN) & unlimited-width (non-charN) string types are not convertible.

FunctionDescription
COALESCE(expr_a, ..., expr_N)Returns the value of the first expression that is not null starting with expr_a and ending with expr_N. If all are null, then null is returned. All expressions should be of the same or convertible data type.
IFNULL(expr_a, expr_b)Returns expr_a if it is not null; otherwise, returns expr_b. Both should be of the same or convertible data type. See Short-Circuiting for error-checking details.
ISNULL(expr)Returns 1 if expr is null; otherwise, returns 0
IS_NULL(expr)Synonymous with ISNULL(expr)
NULLIF(expr_a, expr_b)Returns null if expr_a equals expr_b; otherwise, returns the value of expr_a; both expressions should be of the same or convertible data type.
NVL(expr_a, expr_b)Alias for IFNULL
NVL2(expr, value_if_not_null, value_if_null)Evaluates expr: if not null, returns value_if_not_null; if null, returns value_if_null. Both value_if_not_null & value_if_null should be of the same data type as expr or implicitly convertible; see Short-Circuiting for error-checking details.
REMOVE_NULLABLE(expr)Alias for ZEROIFNULL
ZEROIFNULL(expr)Replaces null values with appropriate values based on the column type (e.g., 0 if numeric column, an empty string if charN column, etc.). Also removes the nullable column property if used to calculate a derived column.

String Functions

There are three different types of string functions:

CharN Functions

The following functions will only work with fixed-width string columns (VARCHAR(1) - VARCHAR(256)).

FunctionDescription
ASCII(expr)Returns the ASCII code for the first character in expr
CHAR(expr)The character represented by the standard ASCII code expr in the range [ 0 - 127 ]
CONCAT(expr_a, expr_b)

Performs a string concatenation of expr_a & expr_b; use nested CONCAT calls to concatenate more than two strings

Note

The resulting field size of any CONCAT will be a charN field big enough to hold the concatenated fields, e.g., concatenating a char32 column and a char64 column will result in a char128 column. Columns of type char256 cannot be used with CONCAT.

CONCAT_TRUNCATE(expr_a, expr_b)

Returns the concatenation of expr_a and expr_b, truncated at the maximum size of expr_a. For columns, this size is explicit; for string constants, this will be the smallest charN type that can hold the expr_a string.

Examples:

Function CallResult
CONCAT_TRUNCATE('ABC123','!')ABC123!
CONCAT_TRUNCATE('ABC123','DEFG')

ABC123DE

(char8 is the minimum size required to hold the ABC123 value, so the result is truncated at 8 characters)

CONCAT_TRUNCATE('ABCD1234','DEFG')

ABCD1234

(char8 is the minimum size required to hold the ABCD1234 value, so no additional characters can be concatenated)

CONTAINS(match_expr, ref_expr)Returns 1 if ref_expr contains match_expr by string-literal comparison; otherwise, returns 0
DIFFERENCE(expr_a, expr_b)Returns a value between 0 and 4 that represents the difference between the sounds of expr_a and expr_b based on the SOUNDEX() value of the strings--a value of 4 is the best possible sound match
EDIT_DISTANCE(expr_a, expr_b)Returns the Levenshtein edit distance between expr_a and expr_b; the lower the the value, the more similar the two strings are
ENDS_WITH(match_expr, ref_expr)Returns 1 if ref_expr ends with match_expr by string-literal comparison; otherwise, returns 0
INITCAP(expr)Returns expr with the first letter of each word in uppercase
IPV4_PART(expr, part_num)

Returns the octet of the IP address given in expr at the position specified by part_num. Valid part_num values are constants from 1 to 4.

Examples:

Function CallResult
IPV4_PART('12.34.56.78', 1)12
IPV4_PART('12.34.56.78', 4)78
IS_IPV4(expr)Returns 1 if expr is an IPV4 address; returns 0 otherwise
LCASE(expr)Converts expr to lowercase
LEFT(expr, num_chars)Returns the leftmost num_chars characters from expr
LENGTH(expr)Returns the number of characters in expr
LOCATE(match_expr, ref_expr, [start_pos])Returns the starting position of the first match of match_expr in ref_expr, starting from position 1 or start_pos (if specified)
LOWER(expr)Alias for LCASE
LPAD(base_expr, length, pad_expr)

Left pads the given base_expr string with the pad_expr string to the given length of characters. If base_expr is longer than length, the return value is shortened to length characters. If length is larger than 256, it will be truncated to 256.

Examples:

Function CallResult
LPAD('test', 9, 'pad')padpatest
LPAD('test', 3, 'pad')tes
LTRIM(expr)Removes whitespace from the left side of expr
POSITION(match_expr, ref_expr, [start_pos])Alias for LOCATE
REPLACE(ref_expr, match_expr, repl_expr)Replaces every occurrence of match_expr in ref_expr with repl_expr
REVERSE(expr)

Returns expr with the order of characters reversed.

Examples:

Function CallResult
REVERSE('Reverse')esreveR
REVERSE('Was it a bat I saw?')?was I tab a ti saW
RIGHT(expr, num_chars)Returns the rightmost num_chars characters from expr
RPAD(base_expr, length, pad_expr)

Right pads the given base_expr string with the pad_expr string to the given length of characters. If base_expr is longer than length, the return value is shortened to length characters. If length is larger than 256, it will be truncated to 256.

Examples:

Function CallResult
RPAD('test', 9, 'pad')testpadpa
RPAD('test', 3, 'pad')tes
RTRIM(expr)Removes whitespace from the right side of expr
SOUNDEX(expr)

Returns a soundex value from expr. Only the first word in the string will be considered in the calculation.

Note

This is the algorithm used by most programming languages.

SPACE(n)Returns a string consisting of n space characters. The value of n can only be within the range of 0-256.
SPLIT(expr, delim, group_num)

Splits expr into groups delimited by the delim character and returns the group_num split group. If group_num is positive, groups will be counted from the beginning of expr; if negative, groups will be counted from the end of expr going backwards. Two consecutive delimiters will result in an empty string being added to the list of selectable groups. If no instances of delim exist in expr, the entire string is available at group 1 (and -1). Group 0 returns nothing.

Examples:

Function CallResult
SPLIT('apple', 'p', 1)a
SPLIT('apple', 'p', 2)<empty string>
SPLIT('apple', 'p', -1)le
STARTS_WITH(match_expr, ref_expr)Returns 1 if ref_expr starts with match_expr by string-literal comparison; otherwise, returns 0
STRCMP(expr_a, expr_b)

Compares expr_a to expr_b in a lexicographical sort

SituationResult
expr_a and expr_b are the same0
expr_a comes before expr_b, lexicographically-1
expr_a comes after expr_b, lexicographically1
SUBSTR(expr, start_pos, num_chars)Alias for SUBSTRING
SUBSTRING(expr, start_pos, num_chars)Returns num_chars characters from the expr, starting at the 1-based start_pos
TRIM(expr)Removes whitespace from both sides of expr
UCASE(expr)Converts expr to uppercase
UPPER(expr)Alias for UCASE

LIKE

Simple pattern-matching capability is provided through the use of the LIKE clause, for fixed-width string columns (VARCHAR(1) - VARCHAR(256)).

LIKE Syntax
1
<ref_expr> [NOT] LIKE <match_expr>

This clause matches records where reference expression ref_expr does (or does NOT) match the string value of match_expr. The match is a string literal one, with the following exceptions:

  • % matches any string of 0 or more characters
  • _ matches any single character

For example, to search for employees whose last name starts with C, ends with a, and has exactly five letters:

LIKE Single Character Match Example
1
2
3
SELECT *
FROM example.employee
WHERE last_name LIKE 'C___a'

To search for employees whose first name does not start with Brook:

LIKE Multiple Character Match Example
1
2
3
SELECT *
FROM example.employee
WHERE first_name NOT LIKE 'Brook%'

FILTER_BY_STRING

The FILTER_BY_STRING table function allows for several types of pattern matching for fixed-width string columns (VARCHAR(1) - VARCHAR(256)), as well as full text search capability for unrestricted-width string columns (VARCHAR). Its features are based on the /filter/bystring endpoint.

The basic form of the FILTER_BY_STRING function when called in a SELECT statement follows.

FILTER_BY_STRING Table Function Syntax
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
SELECT *
FROM TABLE
(
    FILTER_BY_STRING
    (
        TABLE_NAME => INPUT_TABLE(<table name | select statement>),
        [COLUMN_NAMES => '<column list>',]
        MODE => '<filter type>',
        EXPRESSION => '<filter expression>',
        [OPTIONS => KV_PAIRS('<option name>' = '<option value>'[,...])]
    )
)

The basic form of the FILTER_BY_STRING function when called in an EXECUTE FUNCTION statement follows. When called this way, the results are saved in the specified view.

FILTER_BY_STRING Execute Function Syntax
1
2
3
4
5
6
7
8
9
EXECUTE FUNCTION FILTER_BY_STRING
(
    TABLE_NAME => INPUT_TABLE(<table name | select statement>),
    VIEW_NAME => '[<schema name>.]<view name>',
    [COLUMN_NAMES => '<column list>',]
    MODE => '<filter type>',
    EXPRESSION => '<filter expression>',
    [OPTIONS => KV_PAIRS('<option name>' = '<option value>'[,...])]
)
ParametersDescription
TABLE_NAME

The data set to filter, which should be passed via query to the INPUT_TABLE function.

To perform a string filter on a column in the customer table, pass the name of the table to INPUT_TABLE:

INPUT_TABLE(customer)

To perform a string filter on a column in the result of a query, pass the query to INPUT_TABLE:

INPUT_TABLE(SELECT * FROM customer_west UNION SELECT * FROM customer_east)
VIEW_NAME

Name of the view in which the results are to be stored, in [schema_name.]table_name format, using standard name resolution rules and meeting table naming criteria

Important

This parameter is only applicable when using EXECUTE FUNCTION syntax.

COLUMN_NAMES

Comma-separated list of columns to which the filter will be applied.

Important

Omit this parameter when using the search filter MODE.

MODE

Type of string filter to apply:

ModeDescription
containsPartial substring match (not accelerated). If the column is a string type (non-charN) and the number of records is too large, it will return 0.
equalsExact whole-string match (accelerated).
regexFull regular expression search (not accelerated). If the column is a string type (non-charN) and the number of records is too large, it will return 0.
search

Full text search with wildcards and boolean operators, for unrestricted-width string columns (VARCHAR) that have the TEXT_SEARCH column property applied. See Full Text Search for syntax & grammar detail.

Important

Omit the COLUMN_NAMES parameter when using this MODE. All text-searchable string columns will be searched.

starts_withMatch strings that start with the given expression (not accelerated). If the column is a string type (non-charN) and the number of records is too large, it will return 0.
EXPRESSIONString literal or pattern, depending on MODE selected, to use in filtering string columns
OPTIONS

Optional list of string filtering options, specified as a set of key/value pairs passed as a comma-delimited list of <key> = '<value>' assignments to the KV_PAIRS function; e.g.:

KV_PAIRS(case_sensitive = 'false')
OptionDescription
case_sensitive

If true, the filter will be case-sensitive; if false, case-insensitive.

Note

Not applicable when MODE is search.

To see the message text & time for events in the event_log table containing the word ERROR:

FILTER_BY_STRING Example
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
SELECT *
FROM TABLE
(
    FILTER_BY_STRING
    (
        TABLE_NAME => INPUT_TABLE(SELECT event_time, message FROM example.event_log),
        COLUMN_NAMES => 'message',
        MODE => 'contains',
        EXPRESSION => 'ERROR'
    )
)

User/Security Functions

FunctionDescription
CURRENT_USER()Alias for USER
HASH(column[, seed])Returns a non-negative integer representing an obfuscated version of column, using the given seed; default seed is 0

IS_MEMBER(role[, user])

or

IS_ROLEMEMBER(role[, user])

Returns whether the current user (or the given user, if specified) has been assigned the given role, either directly or indirectly:

SituationResult
Current/given user has been granted role1
Current/given user has not been granted role0
Role role does not existnull
MASK(expr, start, length[, char])

Masks length characters of expr, beginning at the position identified by start, with * characters (or the character specified in char):

Function CallResult
MASK('Characters', 4, 5)Cha*****rs
MASK('Characters', 5, 2, '#')Char##ters
OBFUSCATE(column[, seed])Alias for HASH
SHA256(expr)Returns the hex digits of the SHA-256 hash of the given value expr as a char64 string.
SYSTEM_USER()Alias for USER
USER()Returns the username of the current user

Aggregation Functions

FunctionDescription
ATTR(expr)If MIN(expr) = MAX(expr), returns expr; otherwise *
ARG_MIN(agg_expr, ret_expr)The value of ret_expr where agg_expr is the minimum value (e.g. ARG_MIN(cost, product_id) returns the product ID of the lowest cost product)
ARG_MAX(agg_expr, ret_expr)The value of ret_expr where agg_expr is the maximum value (e.g. ARG_MAX(cost, product_id) returns the product ID of the highest cost product)
AVG(expr)Calculates the average value of expr
CORR(expr1, expr2)Calculates the correlation coefficient of expr1 and expr2
CORRELATION(expr1, expr2)Alias for CORR
CORRCOEF(expr1, expr2)Alias for CORR
COUNT(*)Returns the number of records in a table
COUNT(expr)Returns the number of non-null data values in expr
COUNT(DISTINCT expr)Returns the number of distinct non-null data values in expr
COV(expr1, expr2)Alias for COVAR_POP
COVAR(expr1, expr2)Alias for COVAR_POP
COVARIANCE(expr1, expr2)Alias for COVAR_POP
COVAR_POP(expr1, expr2)Calculates the population covariance of expr1 and expr2
COVAR_SAMP(expr1, expr2)Calculates the sample covariance of expr1 and expr2
GROUPING(expr)

Used primarily with ROLLUP, CUBE, and GROUPING SETS, to distinguish the source of null values in an aggregated result set, returns whether expr is part of the aggregation set used to calculate the values in a given result set row. Returns 0 if expr is part of the row's aggregation set, 1 if expr is not (meaning that aggregation took place across all expr values).

For example, in a ROLLUP(A) operation, there will be two potential rows with null in the result set for column A. One row will contain null values of A aggregated together, and the other will contain null, but be an aggregation over the entire table, irrespective of A values. In this case, GROUPING(A) will return 0 for the null values of A aggregated together (as well as all other grouped A values) and 1 for the row resulting from aggregating across all A values.

KURT(expr)Alias for KURTOSIS_POP
KURTOSIS(expr)Alias for KURTOSIS_POP
KURTOSIS_POP(expr)Calculate the population kurtosis of expr
KURTOSIS_SAMP(expr)Calculate the sample kurtosis of expr
KURT_POP(expr)Alias for KURTOSIS_POP
KURT_SAMP(expr)Alias for KURTOSIS_SAMP
MAX(expr)Finds the maximum value of expr
MEAN(expr)Alias for AVG
MIN(expr)Finds the minimum value of expr
PRODUCT(expr)Multiplies all the values of expr
REGR_AVGX(y, x)Average of the independent variable (SUM(x)/N)
REGR_AVGY(y, x)Average of the dependent variable (SUM(y)/N)
REGR_COUNT(y, x)Number of input rows in which both expressions are nonnull
REGR_INTERCEPT(y, x)Y-intercept of the least-squares-fit linear equation determined by the (X, Y) pairs
REGR_R2(y, x)Square of the correlation coefficient
REGR_SLOPE(y, x)Slope of the least-squares-fit linear equation determined by the (X, Y) pairs
REGR_SXX(y, x)"Sum of squares" of the independent variable (SUM(x^2) - SUM(x)^2/N)
REGR_SXY(y, x)"Sum of Products" of independent variable times dependent variable (SUM(x * y) - SUM(x) * SUM(y)/N)
REGR_SYY(y, x)"Sum of squares" of the dependent variable (SUM(y^2) - SUM(y)^2/N)
SKEW(expr)Alias for SKEWNESS_POP
SKEWNESS(expr)Alias for SKEWNESS_POP
SKEWNESS_POP(expr)Calculate the population skew of expr
SKEWNESS_SAMP(expr)Calculate the sample skew of expr
SKEW_POP(expr)Alias for SKEWNESS_POP
SKEW_SAMP(expr)Alias for SKEWNESS_SAMP
STDDEV(expr)Alias for STDDEV_POP
STDDEV_POP(expr)Calculates the population standard deviation of the values of expr
STDDEV_SAMP(expr)Calculates the sample standard deviation of the values of expr
SUM(expr)Sums all the values of expr
VAR(expr)Alias for VAR_POP
VAR_POP(expr)Calculates the population variance of the values of expr
VAR_SAMP(expr)Calculates the sample variance of the values of expr

Grouping Functions

FunctionDescription
ROLLUP(expr)Calculates n + 1 aggregates for n number of columns in expr
CUBE(expr)Calculates 2n aggregates for n number of columns in expr
GROUPING SETS(expr)Calculates aggregates for any given aggregates in expr, including ROLLUP() and CUBE()

Distribution Functions

Distribution functions are column expressions that affect the sharded/replicated nature of the result set of a given query. It may be necessary to force a result set to be distributed in a certain way for a subsequent operation on that result set to be performant.

Important

Employing these functions will prevent any automatic resharding of data to allow the query to succeed. Use only when a better query plan (with respect to data distribution) is known than any the system can devise.

FunctionDescription
KI_REPLICATE()Force a scalar result set to be replicated (query with no GROUP BY)
KI_REPLICATE_GROUP_BY(0)Force an aggregated result set to be replicated (query with GROUP BY)
KI_MATCH_COLUMN(0)Aligns the column count of queries that are part of a UNION, INTERSECT or EXCEPT with a query whose column list has been amended with either KI_REPLICATE_GROUP_BY or KI_SHARD_KEY
KI_SHARD_KEY(<column list>)

Force the result set to be sharded on the given columns. This will override any implicitly-derived or explicitly-defined replication status the table would have had.

Note

The column(s) listed in column list must also appear in the SELECT list; KI_SHARD_KEY merely identifies which of the selected columns should be used as the shard key.

Sharding Example

For example, a query for all employees and their total employees managed, including employees who don't manage anyone, could employ a UNION like this:

Sharding Example
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
SELECT manager_id, COUNT(*)
FROM example.employee
GROUP BY manager_id
UNION
SELECT id, 0
FROM example.employee
WHERE id NOT IN
  (
    SELECT manager_id
    FROM example.employee
    WHERE manager_id IS NOT NULL
  )

In this example, the employee table is sharded on id. Since the first part of the UNION aggregates on manager_id, the result will be replicated. The second part of the UNION does no aggregation and includes the shard key in the SELECT list; the result of this will be sharded.

Prior to Kinetica v7.0, the limitation of UNION operations requiring that both parts of a UNION have to be distributed the same way, would make the query fail, with the following message:

GPUdb Error: either all input tables must be replicated or all input tables
must be non-replicated

In order to work around this limitation, a distribution function could be used. This is what the SQL engine of Kinetica, versions 7.0 or later, do automatically.

One option is to shard the first part of the UNION to match the second part:

Re-shard Example
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
SELECT
   manager_id,
   COUNT(*),
   KI_SHARD_KEY(manager_id)
FROM example.employee
GROUP BY manager_id
UNION
SELECT
  id,
  0,
  KI_MATCH_COLUMN(0)
FROM example.employee
WHERE id NOT IN
  (
    SELECT manager_id
    FROM example.employee
    WHERE manager_id IS NOT NULL
  )

Here, the distribution function KI_SHARD_KEY is used to make the selected manager_id column the new shard key for the first part of the UNION. Now, the shard key for the first part of the UNION (manager_id) aligns with the shard key for the second part (id), and the query succeeds. Note the use of KI_MATCH_COLUMN, which aligns the selected column lists on each side of the UNION. Without this matching distribution function, the UNION would appear to be merging three columns from the first part of the query into two columns in the second part and would fail.

Note

The manager_id column must exist in the SELECT list in order for the KI_SHARD_KEY function to designate it as the shard key.

Predicates

Predicate are generally used within a SQL WHERE clause to query records. They compare the values of two or more expressions; whenever a record meets the criteria defined in a predicate clause it will be marked as eligible to be part of the query result set. If it meets all predicate clauses defined within a query, it will be returned in the result set.

A single predicate clause may use a simple predicate operator to compare the values of two expressions or a more complex predicate clause form. A compound predicate clause uses a compound predicate operator to link together multiple predicate clauses to further refine a result set.

Unlimited-width (non-charN) strings can only be used within equality-based predicates, e.g. =, IN, etc.

Predicate Operators

  • = equality
  • != or <> inequality
  • < less than
  • <= less than or equal to
  • > greater than
  • >= greater than or equal to

Predicate Clauses

In the following list of predicate clauses, ref_expr is the reference expression to apply the predicate to; note that EXISTS has no reference expression.

Predicate ClauseDescription
<expr_a> <pred_op> <expr_b>Matches records where expr_a relates to expr_b according to predicate operator pred_op.
<ref_expr> <pred_op> ALL (<select statement>)Matches records where the reference expression ref_expr relates to all of the results of select statement according to the predicate operator pred_op
<ref_expr> <pred_op> ANY (<select statement>)Matches records where the reference expression ref_expr relates to any of the results of select statement according to the predicate operator pred_op
<ref_expr> [NOT] BETWEEN <begin_expr> AND <end_expr>Matches records where the reference expression ref_expr is (or is NOT) between the values of begin_expr and end_expr
<ref_expr> [NOT] IN (<match_list>)Matches records where the reference expression ref_expr is (or is NOT) in the match_list list of match values. The list can either be a comma-separated list of terms/expressions or the result of a SELECT statement.
<ref_expr> IS [NOT] NULLMatches records where the reference expression ref_expr is (or is NOT) null.
[NOT] EXISTS (<select statement>)Matches records where select statement returns 1 or more records.

Compound Predicate Operators

Predicate OperatorDescription
<pred_a> AND <pred_b>Matches records where both pred_a & pred_b are true
<pred_a> OR <pred_b>Matches records where either pred_a or pred_b is true
NOT <pred_b>Matches records where pred is false

Subqueries

Non-Correlated Subqueries

These are subqueries that are self-contained, in that they can be executed independently of the surrounding query.

Non-Correlated Subquery in SELECT Example
1
2
3
4
5
6
7
8
9
SELECT
    last_name,
    first_name,
    sal,
    (
        SELECT AVG(sal)
        FROM example.employee
    ) AS avg_emp_sal
FROM example.employee
Non-Correlated Subquery in FROM Example
1
2
3
4
5
6
7
SELECT
    l || ', ' || f AS "Employee_Name"
FROM
    (
        SELECT last_name AS l, first_name AS f
        FROM example.employee
    )
Non-Correlated Subquery in WHERE Example
1
2
3
4
5
6
7
8
SELECT last_name, first_name
FROM example.employee
WHERE dept_id IN
    (
        SELECT id
        FROM example.department
        WHERE code = 'ENGR' OR code = 'MKTG'
    )

Correlated Subqueries

These are subqueries that depend on the values in the surrounding query, and cannot be executed independently of the surrounding query.

Correlated Subquery Example
1
2
3
4
5
6
7
8
SELECT *
FROM demo.nyctaxi o
WHERE fare_amount =
    (
        SELECT MAX(fare_amount)
        FROM demo.nyctaxi i
        WHERE o.passenger_count = i.passenger_count
    )

Important

Correlated subqueries have the following limitations:

  • They cannot reference grouping columns in the parent query
  • They cannot reference tables beyond the immediate outer query; i.e., a table cannot be referenced in a correlated subquery that is two or more levels of nesting deeper than it is
  • They cannot contain disjunctive conditions
  • They cannot be part of an OUTER JOIN ON clause condition

Hints

Hint strings (KI_HINT) can be added as comments within queries, and affect just the query in which they appear. They will override the corresponding client & server settings (when such settings exist). For example:

Hint Example
1
2
3
4
5
CREATE TABLE example.taxi_trip_daily_totals AS
/* KI_HINT_GROUP_BY_PK, KI_HINT_INDEX(transaction_date) */
SELECT vendor_id, DATE(dropoff_datetime) AS transaction_date, COUNT(*) AS total_trips
FROM demo.nyctaxi
GROUP BY vendor_id, transaction_date
HintDescription
KI_HINT_BATCH_SIZE(n)

Use an ingest batch size of n records. Default: 10,000.

Only applicable when issuing INSERT statements.

KI_HINT_CHUNK_SIZE(n)Use chunk sizes of n records per chunk within result sets. Suffixes of K & M can be used to represent thousands or millions of records; e.g., 20K, 50M.
KI_HINT_COMPARABLE_EXPLAIN

Simplify EXPLAIN output removing unique identifiers from table names.

Note

Potentially unsafe for use in a multiuser environment, as the table names produced will not be unique and may collide with the table names of other users who execute EXPLAIN with this hint.

KI_HINT_DICT_PROJECTIONRetain the dictionary encoding attributes of source columns in a projection result set.
KI_HINT_DISTRIBUTED_OPERATIONSReshard data when doing so would be the only way to process one or more operations in this query.
KI_HINT_DONT_COMBINEDon’t combine joins and unions for this query.
KI_HINT_GROUP_BY_FORCE_REPLICATEDMake all result tables within a single query replicated; useful when meeting the input table requirements of JOIN, UNION, etc. in a query containing aggregated subqueries which generate differently-sharded result tables.
KI_HINT_GROUP_BY_PK

Create primary keys for all GROUP BY result sets on the grouped-upon columns/expressions within a given query; often used to create a primary key on the result of a CREATE TABLE...AS that ends in a GROUP BY, and can also make materialized views containing grouping operations more performant.

Note

If any of the grouped-on expressions are nullable, no primary key will be applied.

KI_HINT_HAS_HEADERAssume the first line of the source CSV file has a Kinetica header row. Only used with CSV ingestion.
KI_HINT_INDEX(<column list>)Create an index on each of the comma-separated columns in the given list; often used with CREATE TABLE...AS to create an index on a persisted result set.
KI_HINT_JOBID_PREFIX(x)Tag corresponding database job names(s) with x; e.g., KI_HINT_JOBID_PREFIX(tag) will result in job names like ODBC_tag_01234567-89ab-cdef-0123-456789abcdef.
KI_HINT_JOIN_TABLE_CHUNK_SIZE(n)Use chunk sizes of n records per chunk within joins. Suffixes of K & M can be used to represent thousands or millions of records; e.g., 20K, 50M.
KI_HINT_KEEP_TEMP_TABLESDon’t erase temp tables created by this query.
KI_HINT_NO_COST_BASED_OPTIMIZATIONDon't use the cost-based optimizer when calculating the query plan.
KI_HINT_NO_DICT_PROJECTIONDon't retain the dictionary encoding attributes of source columns in a projection result set.
KI_HINT_NO_DISTRIBUTED_OPERATIONSDon't reshard data when doing so would be the only way to process one or more operations in this query.
KI_HINT_NO_HEADERAssume the source CSV file has no Kinetica header row. Only used with CSV ingestion.
KI_HINT_NO_JOIN_COUNTOptimize joins by not calculating intermediate set counts.
KI_HINT_NO_LATE_MATERIALIZATIONForce the materialization of intermediary result sets.
KI_HINT_NO_PARALLEL_EXECUTIONExecute all components of this query in series.
KI_HINT_NO_PLAN_CACHEDon't cache the query plan calculated for this query.
KI_HINT_NO_RULE_BASED_OPTIMIZATIONDon't use the rule-based optimizer when calculating the query plan.
KI_HINT_NO_VALIDATE_CHANGEDon't fail an ALTER TABLE command when changing column types/sizes and the column data is too long/large. Truncate the data instead and allow the modification to succeed.
KI_HINT_PROJECT_MATERIALIZED_VIEWForce the materialization of a materialized view. Some materialized views containing JOIN clauses will be backed by a native join view. This is done to improve the performance of materialized view refreshes and reduce memory usage at the cost of reduced query performance. This hint will induce the reverse of this trade-off -- increased query performance at the cost of reduced refresh performance and increased memory usage.
KI_HINT_REPL_SYNC

Instruct the target database to treat this statement as one that should be run synchronously across all HA clusters in its ring. See High Availability Operation Handling for details.

Note

The target database must be configured for high-availability for this to have an effect. See High Availability Configuration & Management for details.

KI_HINT_REPL_ASYNC

Instruct the target database to treat this statement as one that should be run asynchronously across all HA clusters in its ring. See High Availability Operation Handling for details.

Note

The target database must be configured for high-availability for this to have an effect. See High Availability Configuration & Management for details.

KI_HINT_REQUEST_TIMEOUT(m)Use a timeout of m minutes when processing this command.
KI_HINT_TRUNCATE_STRINGSTruncate all strings being inserted into restricted-width (charN) columns to their max width. Used with any INSERT INTO statement, including CSV ingestion.
KI_HINT_UPDATE_ON_EXISTING_PKChange the record collision policy for inserting into a table with a primary key to an upsert scheme; any existing table record with a primary key that matches a record being inserted will be replaced by that new record. Without this hint, the record being inserted will be discarded. If the specified table does not have a primary key, then this hint is ignored.

EXPLAIN

Outputs the execution plan of a given SQL statement.

Tip

For the visual explain plan utility in GAdmin, see the Explain feature under SQL Tool.

EXPLAIN Syntax
1
EXPLAIN [LOGICAL|PHYSICAL|ANALYZE|VERBOSE|VERBOSE ANALYZE] [FORMAT <JSON|TABLE>] <statement>

If LOGICAL is specified, the algebraic execution tree of the statement is output. Otherwise, the physical execution plan will be output.

Each supporting API endpoint call that is made in servicing the request is listed as an execution plan step in the output, along with any input or output tables associated with the call and the prior plan execution steps on which a given execution step depends.

The following options can be specified:

  • LOGICAL - outputs the algebraic execution tree of the statement
  • PHYSICAL - (default) outputs the physical execution plan with the following endpoint-level details per step:
    • ID - execution step number
    • ENDPOINT - name of native API endpoint called
    • INPUT_TABLES - input tables used by the endpoint (if any)
    • OUTPUT_TABLE - output table created by the endpoint (if any)
    • DEPENDENCIES - list of prior execution steps upon which this step depends
  • ANALYZE - same as PHYSICAL, including additional run-time details:
    • RUN_TIME - execution time of each endpoint call
    • RESULT_ROWS - number of records produced in the endpoint call
  • VERBOSE - same as PHYSICAL, including endpoint parameter details:
    • COLUMNS - columns passed to the endpoint call
    • EXPRESSIONS - expressions passed to the endpoint call
    • OPTIONS - option keys & values passed to the endpoint call
    • LAST_USE_TABLES - list of tables that will not be used by any following execution step
    • ADDITIONAL_INFO - other parameters passed to the endpoint call
  • VERBOSE ANALYZE - same as VERBOSE & ANALYZE together, including the execution plan for any joins contained within the query
  • FORMAT JSON - outputs the result in JSON format
  • FORMAT TABLE - (default) outputs the result in tabular format

Important

Specifying ANALYZE will cause the statement to be executed in order to collect run-time statistics on the endpoint calls made.

For example, to output the execution plan for a query that aggregates the number of taxi rides between boroughs (using KI_HINT_COMPARABLE_EXPLAIN hint to simplify output):

EXPLAIN Example
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
EXPLAIN /* KI_HINT_COMPARABLE_EXPLAIN */
SELECT
    n_begin.ntaname AS boro_begin,
    boro_end,
    COUNT(*) AS total_trips
FROM
(
    SELECT pickup_latitude, pickup_longitude, n_end.ntaname AS boro_end
    FROM demo.nyctaxi t
    JOIN example.nyct2010 n_end ON STXY_INTERSECTS(dropoff_longitude, dropoff_latitude, geom)
)
JOIN example.nyct2010 n_begin ON STXY_INTERSECTS(pickup_longitude, pickup_latitude, geom)
GROUP BY 1, 2

The execution plan is listed in table format, as follows:

EXPLAIN Output
1
2
3
4
5
6
7
8
+------+-------------------------+--------------------------------------------------------------------------+----------------------------+----------------+
| ID   | ENDPOINT                | INPUT_TABLES                                                             | OUTPUT_TABLE               | DEPENDENCIES   |
+------+-------------------------+--------------------------------------------------------------------------+----------------------------+----------------+
| 0    | /create/jointable       | demo.nyctaxi AS TableAlias_0_,example.nyct2010 AS TableAlias_1_          | sys_sql_temp.Join_3        | -1             |
| 1    | /create/jointable       | sys_sql_temp.Join_3 AS TableAlias_0_,example.nyct2010 AS TableAlias_1_   | sys_sql_temp.Join_6        | 0              |
| 2    | /aggregate/groupby      | sys_sql_temp.Join_6                                                      | sys_sql_temp.Aggregate_8   | 1              |
| 3    | /get/records/bycolumn   | sys_sql_temp.Aggregate_8                                                 |                            | 2              |
+------+-------------------------+--------------------------------------------------------------------------+----------------------------+----------------+

If there is an error processing a query, the error can be returned in the JSON-formatted execution plan:

EXPLAIN with Detail Example
1
2
3
4
EXPLAIN VERBOSE ANALYZE FORMAT JSON
SELECT * /* KI_HINT_NO_DISTRIBUTED_OPERATIONS */
FROM example.explain_table t1
JOIN example.explain_table t2 ON t1.shard_column = t2.not_shard_column

The execution plan is listed in JSON format with the query error, as follows:

EXPLAIN with Detail Output
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
"PLAN": [
    {
        "ADDITIONAL_INFO": "Not all of the non-replicated tables are equated by shard keys (TM/FPc:1976); error in Job pre-process; error",
        "COLUMNS": "TableAlias_0_.shard_column AS shard_column,TableAlias_0_.not_shard_column AS not_shard_column,TableAlias_1_.shard_column AS shard_column0,TableAlias_1_.not_shard_column AS not_shard_column0",
        "DEPENDENCIES": "-1",
        "ENDPOINT": "/create/jointable",
        "EXPRESSIONS": "inner join TableAlias_0_, TableAlias_1_  on   (TableAlias_0_.shard_column = TableAlias_1_.not_shard_column )  ",
        "ID": "0",
        "JSON_REQUEST": "{
            "join_table_name": "sys_sql_temp.2840_Join_1_2ba7eaf4-8fc2-11ea-b0a4-080027c462bf",
            "table_names": [
                "example.explain_table AS TableAlias_0_",
                "example.explain_table AS TableAlias_1_"
            ],
            "column_names": [
                "TableAlias_0_.shard_column AS shard_column",
                "TableAlias_0_.not_shard_column AS not_shard_column",
                "TableAlias_1_.shard_column AS shard_column0",
                "TableAlias_1_.not_shard_column AS not_shard_column0"
            ],
            "expressions": [
                "inner join TableAlias_0_, TableAlias_1_  on   (TableAlias_0_.shard_column = TableAlias_1_.not_shard_column )  "
            ],
            "options": {
                "create_explain": "true",
                "show_filters": "true",
                "ttl": "-1"
            }
        }",
        "LAST_USE_TABLES": "",
        "OPTIONS": "{create_explain,true} {show_filters,true} {ttl,-1}",
        "RESULT_DISTRIBUTION": "shard_column;shard_column0; / ;",
        "RESULT_ROWS": "0",
        "RUN_TIME": "0"
    },
    {
        "ADDITIONAL_INFO": "",
        "COLUMNS": "shard_column, not_shard_column, shard_column0, not_shard_column0",
        "DEPENDENCIES": "0",
        "ENDPOINT": "/get/records/bycolumn",
        "EXPRESSIONS": "",
        "ID": "1",
        "JSON_REQUEST": "{
            "table_name": "sys_sql_temp.Join_1_fde49585-4a71-45a9-8b0f-6d3b35d414b1",
            "column_names": [
                "shard_column",
                "not_shard_column",
                "shard_column0",
                "not_shard_column0"
            ],
            "offset": 0,
            "limit": -9999,
            "encoding": "binary",
            "options": {}
        }",
        "LAST_USE_TABLES": "sys_sql_temp.Join_1_fde49585-4a71-45a9-8b0f-6d3b35d414b1",
        "OPTIONS": "",
        "RESULT_DISTRIBUTION": "",
        "RESULT_ROWS": "0",
        "RUN_TIME": "0"
    }
]