A type is analogous to a traditional database definition for a table. Before data can be stored in Kinetica, a type must be specified for that data. For every type, Kinetica assigns a unique GUID. Kinetica will use the same GUID for all types with identical characteristics.
Every type in Kinetica consists of the following:
- Type Label (text to identify this type)
- Type Definition (list of columns)
- Properties (optional)
For details on nullability and working with nulls, see Nulls.
Type Label
A type label serves as a tagging mechanism for the type. The type label can be any text string specified by the client. The type label serves two purposes. First, it identifies tables with similar data. Second, it helps determine a type’s uniqueness.
Type Definition
A type definition (sometimes referred to as type schema) consists of a set of column names, their respective base data types, and their nullability.
Column Types
Each column in the database will have an effective type that is the combination of a base type and an optional set of column properties, listed below.
A given column in a type definition must be of one of the following base types. If no data type column properties are specified to modify the base type, the column will take on the corresponding default effective type.
For example, a column with an int base type and no column properties that modify the data type will have an effective type of int, which is a 32-bit signed integer. A column with an int base type and a data type column property of int16 will have an effective type of int16, which is a 16-bit signed integer.
| Effective Type | Base Type | Bytes in Memory | Minimum Value | Maximum Value |
|---|---|---|---|---|
| int (default) | int | 4 | -2147483648 | 2147483647 |
| int16 | 2 | -32768 | 32767 | |
| int8 | 1 | -128 | 127 | |
| boolean | 1 | 0 | 1 | |
| long (default) | long | 8 | -9223372036854775808 | 9223372036854775807 |
| timestamp | 8 | -30610224000000 (1/1/1000 00:00:00.000) | 29379542399999 (12/31/2900 23:59:59.999) | |
| float | float | 4 | -3.40282 * 1038 | 3.40282 * 1038 |
| double | double | 8 | -1.79769313486231 * 10308 | 1.79769313486231 * 10308 |
| string (default) | string | 8 | (empty string) | (100,000,000,000 bytes) |
| char256 | 256 | (empty string) | (256 bytes) | |
| char128 | 128 | (empty string) | (128 bytes) | |
| char64 | 64 | (empty string) | (64 bytes) | |
| char32 | 32 | (empty string) | (32 bytes) | |
| char16 | 16 | (empty string) | (16 bytes) | |
| char8 | 8 | (empty string) | (8 bytes) | |
| char4 | 4 | (empty string) | (4 bytes) | |
| char2 | 2 | (empty string) | (2 bytes) | |
| char1 | 1 | (empty string) | (1 byte) | |
| ipv4 | 4 | 0.0.0.0 | 255.255.255.255 | |
| decimal | 8 | -922337203685477.5808 | 922337203685477.5807 | |
| ulong | 8 | 0 | 18446744073709551615 | |
| date | 4 | 1000-01-01 | 2900-12-31 | |
| 9999-01-01 | 9999-12-31 | |||
| time | 4 | 00:00:00.000 | 23:59:59.999 | |
| datetime | 8 | 1000-01-01 00:00:00.000 | 2900-12-31 23:59:59.999 | |
| 9999-01-01 00:00:00.000 | 9999-12-31 23:59:59.999 | |||
| wkt | varies | (empty string) | (100,000,000,000 bytes) | |
| uuid | 16 | (empty string) | (36 bytes) | |
| bytes (default) | bytes | N/A | (empty array) | (100,000,000,000 bytes) |
| wkt | varies | (empty array) | (100,000,000,000 bytes) |
Note
- Character types use UTF-8 encoding and are defined using byte-designated sizes; the use of wide (multi-byte) characters will decrease the total number of characters that can be stored in a given fixed-width string type.
- The decimal type should be used instead of float or double when exact values of up to 4 decimal places need to be represented. This is often used with currency.
- Any date/time within the year 9999 can be used to denote an unknown or out-of-range date. No dates/times in between 2900 and 9999 will be recognized.
- Adding nullability to a column requires an additional byte per value; e.g., a nullable integer requires 5 bytes in memory instead of 4.
- A valid UUID can be either 30 characters (without hyphens) or 36 characters (with hyphens), e.g., 1bd0b4cc0cbc11eb97be02420a000046 or 1bd0b4cc-0cbc-11eb-97be-02420a000046
Column Properties
Kinetica has an additional layer of semantic regarding column data type. At the time of creation, a column can be given one or more of the supported properties, which give the column special meaning or handling. The properties can refine the data type, direct special handling, or define the keyed nature of the column. These modifiers can impact the number of records that can be stored in memory, the performance of queries, and the types of operations that can be performed on the data.
Data Types
The following properties can be used to modify the allowable set of values for the corresponding base type. Only one of the following properties may be applied to a given column, and the column must be of the stated base type. The exception to this is nullability, which can be applied to any column in addition to the data type specifiers listed here. However, nullable is not used at type creation, but will be returned in a /show/table call. Denoting a column's nullability is API-dependent, and, with the exception of the Java API, does not involve the direct use of the nullable column property.
| Property | Base Type | Description |
|---|---|---|
| char1 | string | Text of up to 1 character; optimizes memory, disk, and query performance |
| char2 | string | Text of up to 2 characters; optimizes memory, disk, and query performance |
| char4 | string | Text of up to 4 characters; optimizes memory, disk, and query performance |
| char8 | string | Text of up to 8 characters; optimizes memory, disk, and query performance |
| char16 | string | Text of up to 16 characters; optimizes memory, disk, and query performance |
| char32 | string | Text of up to 32 characters; optimizes memory, disk, and query performance |
| char64 | string | Text of up to 64 characters; optimizes memory, disk, and query performance |
| char128 | string | Text of up to 128 characters; optimizes memory, disk, and query performance |
| char256 | string | Text of up to 256 characters; optimizes memory, disk, and query performance |
| date | string | Interprets a string field as a date of the form YYYY-[M]M-[D]D |
| datetime | string | Interprets a string field as a combination of date and time in the form of YYYY-[M]M-[D]D [H]H24:MI:SS[.mmm] |
| decimal | string | Interprets a string field as a decimal number, with up to 19 digits of precision and 4 digits of scale |
| ulong | string | Interprets a string field as a 64-bit unsigned long |
| boolean | int | Numbers limited to 0 & 1; optimizes memory and query performance |
| int8 | int | Numbers limited to 8-bit signed integers; optimizes memory and query performance |
| int16 | int | Numbers limited to 16-bit signed integers; optimizes memory and query performance |
| ipv4 | string | Dotted decimal IPv4 addresses of the form: A.B.C.D where A, B, C and D are between 0 and 255, inclusive (e.g. 127.0.0.1); optimizes memory, disk, and query performance |
| nullable | <any> | Values can be set to null Note See Nulls for working with null types & values in the API |
| time | string | Interprets a string field as a time of the form [H]H24:MI:SS[.mmm] |
| timestamp | long | Timestamps in milliseconds since the Unix epoch: Jan 1 1970 00:00:00 |
| wkt | string or bytes | Indicates that the column has WKT (or WKB) strings that should be handled as geometry objects. |
| uuid | string | Interprets a string field as a 30 character (without hyphens) or 36 character (with hyphens) universally unique identifier (UUID) |
Data Handling
One or more of the following properties can be assigned to a column to alter the way the data is stored & handled. Valid combinations are detailed in the descriptions.
| Property | Description |
|---|---|
| data | Directs that the column's data should be stored in memory, making it available for use in query expressions. Default property for all numeric and string type columns, unless overridden by store_only. |
| dict | This property will dictionary encode the associated column, reducing memory and disk usage. Queries against the column will also be faster. For Dictionary Encoding details, see Dictionary Encoding. |
| disk_optimized | Prevents variable-width strings from being written to an indexing service, saving disk space at the cost of some functionality. A /filter/bystring applied will only work in the equals mode, /aggregate/unique cannot be applied, and /aggregate/groupby can only be used when the count or count_distinct function is applied to the column--the column itself cannot otherwise appear in the column list. Requires the data property. |
| store_only | Reduces system memory usage by not keeping a copy of the data in memory and only persisting the values. The data is able to be queried by column name, but since not in memory, no expressions (column, filter, aggregation, etc.) can be applied. It is mutually exclusive with the data property. Default property for bytes type columns. |
| text_search | Enables full text search for string columns. Can be set independently of data, disk_optimized, & store_only. For full text search details and limitations, see Full Text Search. |
Data Keys
One or both of these keyed attributes can be assigned to one or more columns. For more information, see the sections on primary keys & shard keys. Foreign keys cannot be assigned as column properties.
| Property | Description |
|---|---|
| primary_key | Makes this column part of (or the entire) primary key |
| shard_key | Makes this column part of (or the entire) shard key |
Data Replacement
The following property can be assigned to a column to replace certain kinds of data.
| Property | Description |
|---|---|
| init_with_now | For date, time, datetime, and timestamp columns, this property will replace empty strings and invalid timestamp values with NOW(). |
| init_with_uuid | For uuid columns, this property will replace empty strings with a universally unique identifier (UUID). |
