Golang SDK for Data Transforms
API reference documentation for Redpanda data transforms.
Data transforms
Package transform
contains the core functions and types for transforming data within Redpanda.
Data transforms functions
OnRecordWritten
func OnRecordWritten(fn OnRecordWrittenCallback)
The OnRecordWritten
function registers a callback of type OnRecordWrittenCallback
, which is invoked when a record is written to the input topic.
The function should be called in a package’s main
function to register the transform function that will be applied.
Data transforms types
OnRecordWrittenCallback
type OnRecordWrittenCallback func(e WriteEvent, w RecordWriter) error
The OnRecordWrittenCallback
type is a callback to transform records after a write event happens in the input topic. It’s the type of the parameter for the OnRecordWritten
function.
Record
type Record struct {
// Key is an optional field.
Key []byte
// Value is the blob of data that is written to Redpanda.
Value []byte
// Headers are client specified key/value pairs that are
// attached to a record.
Headers []RecordHeader
// Attrs is the attributes of a record.
//
// Output records should leave these unset.
Attrs RecordAttrs
// The timestamp associated with this record.
//
// For output records this can be left unset as it will
// always be the same value as the input record.
Timestamp time.Time
// The offset of this record in the partition.
//
// For output records this field is left unset,
// as it will be set by Redpanda.
Offset int64
}
The Record
type is a record that has been written to Redpanda.
RecordHeader
type RecordHeader struct {
Key []byte
Value []byte
}
The RecordHeader
type is an optional key/value pair that is passed along with records.
WriteEvent
type WriteEvent interface {
// Access the record associated with this event
Record() Record
}
The WriteEvent
type contains information about a record that was written.
RecordWriter
type RecordWriter interface {
// Write writes a record to the output topic.
//
// When writing a record, only the key, value and headers are
// used other information like the timestamp will be overridden
// by the broker.
Write(Record) error
}
RecordWriter
is an interface for writing transformed records to the destination topic.
Schema Registry client
Package sr
is a Schema Registry client for usage within Redpanda data transforms.
Data transforms support reading schemas and writing schemas to Redpanda’s Schema Registry.
After a schema is obtained, a record may be decoded using the appropriate library. Using the TinyGo compiler and runtime, the following libraries for Apache Avro and Protocol Buffers are known to work:
Schema Registry variables
var (
// ErrNotRegistered is returned from Serde when attempting to encode a
// value or decode an ID that has not been registered, or when using
// Decode with a missing new value function.
ErrNotRegistered = errors.New("registration is missing for encode/decode")
// ErrBadHeader is returned from Decode when the input slice is shorter
// than five bytes, or if the first byte is not the magic 0 byte.
ErrBadHeader = errors.New("5 byte header for value is missing or does not have the 0 magic byte")
)
Schema Registry types
ClientOpt
type ClientOpt interface {
// contains filtered or unexported methods
}
ClientOpt
is an option to configure a SchemaRegistryClient
MaxCacheEntries
func MaxCacheEntries(size int) ClientOpt
MaxCacheEntries
configures how many entries to cache within the client.
By default the cache is unbounded. Use 0 to disable the cache.
Reference
type Reference struct {
Name string
Subject string
Version int
}
SchemaReference
is a way for one schema to reference another schema. The details for how referencing is done are type specific; for example, JSON objects that use the key "$ref" can refer to another schema via URL. See Reference a schema.
Schema
type Schema struct {
Schema string
Type SchemaType
References []Reference
}
Schema
is a schema that can be registered within the Schema Registry.
SchemaRegistryClient
type SchemaRegistryClient interface {
// LookupSchemaById looks up a schema via its global ID.
LookupSchemaById(id int) (s *Schema, err error)
// LookupSchemaByVersion looks up a schema via a subject for a specific version.
//
// Use version -1 to get the latest version.
LookupSchemaByVersion(subject string, version int) (s *SubjectSchema, err error)
// CreateSchema creates a schema under the given subject, returning the version and ID.
//
// If an equivalent schema already exists globally, that schema ID will be reused.
// If an equivalent schema already exists within that subject, this will be a noop and the
// existing schema will be returned.
CreateSchema(subject string, schema Schema) (s *SubjectSchema, err error)
}
SchemaRegistryClient
is a client for interacting with Redpanda’s Schema Registry.
The client provides caching out of the box, which can be configured with options.
NewClient
func NewClient(opts ...ClientOpt) (c SchemaRegistryClient)
NewClient
creates a new SchemaRegistryClient
with the specified options applied.
SchemaType
type SchemaType int
SchemaType
is an enum for the different types of schemas that can be stored in the Schema Registry.
const (
TypeAvro SchemaType = iota
TypeProtobuf
TypeJSON
)
Serde
type Serde[T any] struct {
// contains filtered or unexported fields
}
Serde
encodes and decodes values according to the Schema Registry wire format. A Serde
itself does not perform schema auto-discovery and type auto-decoding. To aid in strong typing and validated encoding/decoding, you must register IDs and values.
To use a Serde
for encoding, you must first preregister the schema IDs and values that you will encode. The latest registered ID that supports encoding is used to encode.
To use a Serde
for decoding, you can either preregister the schema IDs and values that you will consume, or you can discover the schema every time you receive an ErrNotRegistered
error from decode.
(*Serde[T]) AppendEncode
func (s *Serde[T]) AppendEncode(b []byte, v T) ([]byte, error)
AppendEncode
appends an encoded value to b
according to the schema registry wire format and returns it. If EncodeFn
was not used, this returns ErrNotRegistered
.
(*Serde[T]) Decode
func (s *Serde[T]) Decode(b []byte, v T) error
Decode
decodes b
into v
. If the DecodeFn
option was not used, this returns ErrNotRegistered
.
Serde does not handle references in schemas. You must register the full decode function for any top-level ID, regardless of how many other schemas are referenced in the top-level ID.
(*Serde[T]) Encode
func (s *Serde[T]) Encode(v T) ([]byte, error)
Encode
encodes a value according to the Schema Registry wire format and returns it. If EncodeFn
was not used, this returns ErrNotRegistered
.
(*Serde[T]) MustAppendEncode
func (s *Serde[T]) MustAppendEncode(b []byte, v T) []byte
MustAppendEncode
returns the value of AppendEncode
, panicking on error. This is a shortcut for when your encode function cannot error.
(*Serde[T]) MustEncode
func (s *Serde[T]) MustEncode(v T) []byte
MustEncode
returns the value of Encode
, panicking on error. This is a shortcut for when your encode function cannot error.
(*Serde[T]) Register
func (s *Serde[T]) Register(id int, opts ...SerdeOpt[T])
Register
registers a schema ID and the value it corresponds to, as well as the encoding or decoding functions. Register functions depending on whether you are only encoding, only decoding, or both.
(*Serde[T]) SetDefaults
func (s *Serde[T]) SetDefaults(opts ...SerdeOpt[T])
SetDefaults
sets default options to apply to every registered type. These options are always applied first, so you can override them as necessary when registering.
This can be useful if you always want to use the same encoding or decoding functions.
SerdeOpt
type SerdeOpt[T any] interface {
// contains filtered or unexported methods
}
SerdeOpt
is an option to configure a Serde
.
DecodeFn
func DecodeFn[T any](fn func([]byte, T) error) SerdeOpt[T]
DecodeFn
allows Serde
to decode into a value.
EncodeFn
func EncodeFn[T any](fn func(T) ([]byte, error)) SerdeOpt[T]
EncodeFn
allows Serde
to encode a value.