« Back to index

Types

Table of Contents

  1. Introduction
  2. Primitives
  3. Any
  4. Duration
  5. Enum
  6. List
  7. Struct
  8. Tuple
  9. Windowed

Introduction

KSML supports a wide range of simple and complex types. Types are very useful.

Primitives

The following native types are supported.

Type Description
? Any type
boolean Boolean values, ie. true or false
bytes Byte array
double Double precision floating point
float Single precision floating point
byte 8-bit integer
short 16-bit integer
int 32-bit integer
long 64-bit long
string String of characters
struct Untyped key-value map, where keys are always of type string

Any

The special type any can be used in places where input is uncertain. Code that deals with input of this type should always perform proper type checking before assuming any specific underlying type.

Enum

Enumerations can be defined as individual types, through:

enum(literal1, literal2, ...)

List

Lists contain elements of the same type. They are defined using:

[elementType]

Examples:

[string]
[long]
[avro:SensorData]
[(long,string)]

Struct

Structs are key-value maps, as usually found in AVRO or JSON messages. They are defined using:


## Tuple

Tuples combine multiple subtypes into one. For example `(1, "text")` is a tuple containing an integer and a string element.
Tuple types always have a fixed number of elements.

Examples:

(long, string) (avro:SensorData, string, long, string) ([string], long)


## Union

Unions are 'either-or' types. They have their own internal structure and can be described
by respective data schema. Unions are defined using:

union(type1, type2, …)


Examples:

union(null, string) union(avro:SensorData, long, string)


## Windowed

Some Kafka Streams operations modify the key type from _K_ to _Windowed\<K>_. Kafka Streams uses the
_Windowed\<K>_ type to group Kafka messages with similar keys together. The result is always a time-bound
window, with a defined start and end time.

KSML can convert the internal _Windowed\<K>_ type into a struct type with five fields:
* `start`: The window start timestamp (type `long`)
* `end`: The window end timestamp (type `long`)
* `startTime`: The window start time (type `string`)
* `endTime`: The window end time (type `string`)
* `key`: The key used to group items together in this window (type is the same as the original key type)

However, in pipelines or topic definitions users may need to refer to this type explicitly. This
is done in the following manner:

notation:windowed(keytype)


For example:

avro:windowed(avro:SensorData) xml:windowed(long) ```