Data types in Amazon Athena

When you run CREATE TABLE , you specify column names and the data type that each column can contain. Athena supports the data types listed below. For information about the data type mappings that the JDBC driver supports between Athena, JDBC, and Java, see Data types in the JDBC Driver Installation and Configuration Guide . For information about the data type mappings that the ODBC driver supports between Athena and SQL, see Data types in the ODBC Driver Installation and Configuration Guide .

tinyint – A 8-bit signed integer in two's complement format, with a minimum value of -2 ⁷ and a maximum value of 2 ⁷ -1.

smallint – A 16-bit signed integer in two's complement format, with a minimum value of -2 ¹⁵ and a maximum value of 2 ¹⁵ -1.

int and integer – Athena uses different expressions for integer depending on the type of query.

integer – In DML queries like SELECT * FROM , use the integer data type. integer is represented as a 32-bit signed value in two's complement format, with a minimum value of -2 ³¹ and a maximum value of 2 ³¹ -1.

To ensure compatibility with business analytics applications, the JDBC driver returns the integer type.

bigint – A 64-bit signed integer in two's complement format, with a minimum value of -2 ⁶³ and a maximum value of 2 ⁶³ -1.

double – A 64-bit signed double-precision floating point number. The range is 4.94065645841246544e-324d to 1.79769313486231570e+308d, positive or negative. double follows the IEEE Standard for Floating-Point Arithmetic (IEEE 754).

float – A 32-bit signed single-precision floating point number. The range is 1.40129846432481707e-45 to 3.40282346638528860e+38, positive or negative. float follows the IEEE Standard for Floating-Point Arithmetic (IEEE 754). Equivalent to the real in Presto. In Athena, use float in DDL statements like CREATE TABLE and real in SQL functions like SELECT CAST . The AWS Glue crawler returns values in float , and Athena translates real and float types internally (see the June 5, 2018 release notes).


               decimal

–


              
               precision

is the total number of digits.


              
               scale

(optional) is the number of digits in fractional part with a default of 0. For example, use these type definitions:


              decimal(11,5)


              decimal(15)

. The maximum value for


              precision

is 38, and the maximum value for


              scale

is 38.

To specify decimal values as literals, such as when selecting rows with a specific decimal value in a query DDL expression, specify the decimal type definition, and list the decimal value as a literal (in single quotes) in your query, as in this example: decimal_value = decimal '0.12' .

char – Fixed length character data, with a specified length between 1 and 255, such as char(10) . For more information, see CHAR Hive data type .

Note

To use the substr function to return a substring of specified length from a char data type, you must first cast the char value as a varchar , as in the following example.


substr(cast(col1 as varchar), 1, 4)

varchar – Variable length character data, with a specified length between 1 and 65535, such as varchar(10) . For more information, see VARCHAR Hive data type .

string – A string literal enclosed in single or double quotes. For more information, see STRING Hive data type .

Note

Non-string data types cannot be cast to string in Athena; cast them to varchar instead.

binary – Used for data in Parquet.

date – A date in ISO format, such as YYYY - MM - DD . For example, date '2008-09-15' . An exception is the OpenCSVSerDe, which uses the number of days elapsed since January 1, 1970. For more information, see OpenCSVSerDe for processing CSV .

timestamp – Date and time instant in a java.sql.Timestamp compatible format up to a maximum resolution of milliseconds, such as yyyy - MM - dd HH : mm : ss [.f ... ] . For example, timestamp '2008-09-15 03:04:05.324' . An exception is the OpenCSVSerDe , which uses timestamp data in the UNIX numeric format (for example, 1579059880000 ).

For more information about working with timestamps, see Working with timestamp data later in this document.

array < data_type > – An array of the given component type.

Example


CREATE TABLE table array_table (c1 array<integer>) LOCATION '...';
INSERT INTO array_table values(ARRAY[1,2,3]);

map < primitive_type , data_type > – A map between the given component types.

Example


CREATE TABLE map_table(c1 map<string, integer>) LOCATION '...';
INSERT INTO map_table values(MAP(ARRAY['foo', 'bar'], ARRAY[1, 2]));

struct < col_name : data_type [comment col_comment ] , ... > – A collection of elements of different component types.

Example


CREATE TABLE struct_table(c1 struct<name:varchar(10), age:integer>) LOCATION '...';
INSERT INTO struct_table SELECT CAST(ROW('Bob', 38) AS ROW(name VARCHAR(10), age INTEGER));

Working with timestamp data

This section describes some considerations for working with timestamp data in Athena.

Note

The treatment of timestamps has changed somewhat between Athena engine version 2 and Athena engine version 3. For information about timestamp-related errors that can occur in Athena engine version 3 and suggested solutions, see Timestamp changes in the Athena engine version 3 reference.

Format for writing timestamp data to Amazon S3 objects

The format in which timestamp data should be written into Amazon S3 objects depends on both the column data type and the SerDe library that you use.

If you have a table column of type DATE , Athena expects the corresponding column or property of the data to be a string in the ISO format YYYY-MM-DD , or a built-in date type like those for Parquet or

If you have a table column of type TIME , Athena expects the corresponding column or property of the data to be a string in the ISO format HH:MM:DD , or a built-in time type like those for Parquet or

If you have a table column of type TIMESTAMP , Athena expects the corresponding column or property of the data to be a string in the format YYYY-MM-DD HH:MM:SS.SSS (note the space between the date and time), or a built-in time type like those for Parquet, ORC, or Ion.

Note

OpenCSVSerDe timestamps are an exception and must be encoded as millisecond resolution UNIX epochs.

Ensuring that time-partitioned data matches the timestamp field in a record

The producer of the data must make sure partition values align with the data within the partition. For example, if your data has a timestamp property and you use Kinesis Data Firehose to load the data into Amazon S3, you must use dynamic partitioning because the default partitioning of Kinesis Data Firehose is wall-clock-based.

Use string as the data type for partition keys

For performance reasons, it is preferable to use STRING as the data type for partition keys. Even though Athena recognizes partition values in the format YYYY-MM-DD as dates when you use the DATE type, this can lead to poor performance. For this reason, we recommend that you use the STRING data type for partition keys instead.

How to write queries for timestamp fields that are also time-partitioned

How you write queries for timestamp fields that are time-partitioned depends on the type of table that you want to query.

Hive tables

With the Hive tables most commonly used in Athena, the query engine has no knowledge of relationships between columns and partition keys. For this reason, you must always add predicates in your queries for both the column and the partition

For example, suppose you have an event_time column and an event_date partition key and want to query events between 23:00 and 03:00. In this case, you must include predicates in your query for both the column and the partition key, as in the following example.


WHERE event_time BETWEEN start_time AND end_time 
  AND event_date BETWEEN start_time_date AND end_time_date

Iceberg tables

With Iceberg tables, you can use computed partition values, which simplifies your queries. For example, suppose your Iceberg table was created with a PARTITIONED BY clause like the following:


PARTITIONED BY (event_date month(event_time))

In this case, the query engine automatically prunes partitions based on the values of the event_time predicates. Because of this, your query only needs to specify a predicate for event_time , as in the following example.