Deli putem

Serialization and deserialization formats in data processor pipelines

  • Članak

Important

Azure IoT Operations Preview – enabled by Azure Arc is currently in PREVIEW. You shouldn't use this preview software in production environments.

You will need to deploy a new Azure IoT Operations installation when a generally available release is made available, you won't be able to upgrade a preview installation.

See the Supplemental Terms of Use for Microsoft Azure Previews for legal terms that apply to Azure features that are in beta, preview, or otherwise not yet released into general availability.

The data processor is a data agnostic platform. The data processor can ingest, process, and write out data in any format.

However, to use jq path expressions in some pipeline stages, the data must be in a structured format within a pipeline. You may need to deserialize your data to get it into a suitable structured format.

Some pipeline destinations or call outs from stages may require the data to be in a specific format. You may need to serialize your data to a suitable format for the destination.

Deserialize messages

The data processor natively supports deserialization of various formats at both the data source stage and the call out stages where the pipeline reads external data:

  • The source stage can deserialize incoming data.
  • The call out stages can deserialize the API response.

You may not need to deserialize incoming data if:

  • You're not using the stages that require deserialized data.
  • You're processing metadata only.
  • The incoming data is already in a format that's consistent with the stages being used.

Following table lists the formats for which deserialization is supported and the corresponding stages.

Format Data source Call out
Raw Supported HTTP
JSON Supported HTTP
Protobuf Supported All (HTTP and gRPC)
CSV Supported HTTP
MessagePack Supported HTTP
CBOR Supported HTTP

Tip

Select Raw when you don't require deserialization. The Raw option passes the data through in it's current format.

Serialize messages

The data processor natively supports serialization to various formats at both the destination and call out stages where the pipeline writes external data:

  • The destination stage can serialize outgoing data to suitable format.
  • Call out stages can serialize the data sent in an API request.
Format Call out Output stage
Raw HTTP All except Microsoft Fabric
JSON HTTP All except Microsoft Fabric
Parquet Not supported Microsoft Fabric
Protobuf All All except Microsoft Fabric
CSV HTTP All except Microsoft Fabric
MessagePack HTTP All except Microsoft Fabric
CBOR HTTP All except Microsoft Fabric

Tip

Select Raw when no serialization is required. The Raw option passes the data through in its current format.

Raw/JSON/MessagePack/CBOR data formats

Raw is the option to use when you don't need to deserialize or serialize data. Raw is the default in most stages where deserialization or serialization isn't enforced.

The serialization or deserialization configuration is common for the Raw, JSON, MessagePack, and CBOR formats. For these formats, use the following configuration options.

Use the following configuration options to deserialize data:

Field Type Description Required? Default Example
type string enum The format for deserialization No - JSON
path Path The path to the portion of the data processor message where the deserialized data is written to. (see following note) .payload .payload.response

Note

You don't need to specify path when you deserialize data in the source stage. The deserialized data is automatically placed in the .payload section of the message.

Use the following configuration options to serialize data:

Field Type Description Required? Default Example
type string enum The format for serialization Yes - JSON
path Path The path to the portion of the data processor message that should be serialized. (see following note) .payload .payload.response

Note

You don't need to specify path when you serialize batched data. The default path is ., which represents the entire message. For unbatched data, you must specify path.

The following example shows the configuration for serializing or deserializing unbatched JSON data:

{
    "format": {
        "type": "json",
        "path": ".payload"
    }
}

The following example shows the configuration for deserializing JSON data in the source stage or serializing batched JSON data:

{
    "format": {
        "type": "json"
    }
}

Protocol Buffers data format

Use the following configuration options to deserialize Protocol Buffers (protobuf) data:

Field Type Description Required? Default Example
type string enum The format for deserialization Yes - protobuf
descriptor string The base64 encoded descriptor for the protobuf definition file(s). Yes - Zm9v..
package string The name of the package in the descriptor where the type is defined. Yes - package1..
message string The name of the message type that's used to format the data. Yes - message1..
path Path The path to the portion of the data processor message where the deserialized data should be written. (see following note) .payload .payload.gRPCResponse

Note

You don't need to specify path when you deserialize data in the source stage. The deserialized data is automatically placed in the .payload section of the message.

Use the following configuration options to serialize protobuf data:

Field Type Description Required? Default Example
type string enum The format for serialization Yes - protobuf
descriptor string The base64 encoded descriptor for the protobuf definition file(s). Yes - Zm9v..
package string The name of the package in the descriptor where the type is defined. Yes - package1..
message string The name of the message type that's used to format the data. Yes - message1..
path Path The path to the portion of the data processor message where data to be serialized is read from. (see following note) - .payload.gRPCRequest

Note

You don't need to specify path when you serialize batched data. The default path is ., which represents the entire message.

The following example shows the configuration for serializing or deserializing unbatched protobuf data:

{
    "format": {
        "type": "protobuf",
        "descriptor": "Zm9v..",
        "package": "package1",
        "message": "message1",
        "path": ".payload"
    }
}

The following example shows the configuration for deserializing protobuf data in the source stage or serializing batched protobuf data:

{
    "format": {
        "type": "protobuf",
        "descriptor": "Zm9v...", // The full descriptor
        "package": "package1",
        "message": "message1"
    }
}

CSV data format

Use the following configuration options to deserialize CSV data:

Field Type Description Required? Default Example
type string enum The format for deserialization Yes - CSV
header boolean This field indicates whether the input data has a CSV header row. Yes - true
columns array The schema definition of the CSV to read. Yes - (see following table)
path Path The path to the portion of the data processor message where the deserialized data should be written. (see following note) - .payload

Note

You don't need to specify path when you deserialize data in the source stage. The deserialized data is automatically placed in the .payload section of the message.

Each element in the columns array is an object with the following schema:

Field Type Description Required? Default Example
name string The name of the column as it appears in the CSV header. Yes - temperature
type string enum The data processor data type held in the column that's used to determine how to parse the data. No string integer
path Path The location within each record of the data where the value of the column should be read from. No .{{name}} .temperature

Use the following configuration options to serialize CSV data:

Field Type Description Required? Default Example
type string enum The format for serialization Yes - CSV
header boolean This field indicates whether to include the header line with column names in the serialized CSV. Yes - true
columns array The schema definition of the CSV to write. Yes - (see following table)
path Path The path to the portion of the data processor message where data to be serialized is written. (see following note) - .payload

Note

You don't need to specify path when you serialize batched data. The default path is ., which represents the entire message.

Field Type Description Required? Default Example
name string The name of the column as it would appear in a CSV header. Yes - temperature
path Path The location within each record of the data where the value of the column should be written to. No .{{name}} .temperature

The following example shows the configuration for serializing unbatched CSV data:

{
    "format": {
        "type": "csv",
        "header": true,
        "columns": [
            {
                "name": "assetId",
                "path": ".assetId"
            },
            {
                "name": "timestamp",
                "path": ".eventTime"
            },
            {
                "name": "temperature",
                // Path is optional, defaults to the name
            }
        ],
        "path": ".payload"
    }
}

The following example shows the configuration for serializing batched CSV data. Omit the top-level path for batched data:

{
    "format": {
        "type": "csv",
        "header": true,
        "columns": [
            {
                "name": "assetId",
                "path": ".assetId"
            },
            {
                "name": "timestamp",
                "path": ".eventTime"
            },
            {
                "name": "temperature",
                // Path is optional, defaults to .temperature
            }
        ]
    }
}

The following example shows the configuration for deserializing unbatched CSV data:

{
    "format": {
        "type": "csv",
        "header": false,
        "columns": [
            {
                "name": "assetId",
                "type": "string",
                "path": ".assetId"
            },
            {
                "name": "timestamp",
                // Type is optional, defaults to string
                "path": ".eventTime"
            },
            {
                "name": "temperature",
                "type": "float"
                // Path is optional, defaults to .temperature
            }
        ],
        "path": ".payload"
    }
}

The following example shows the configuration for deserializing batched CSV data in the source stage:

{
    "format": {
        "type": "csv",
        "header": false,
        "columns": [
            {
                "name": "assetId",
                "type": "string",
                "path": ".assetId"
            },
            {
                "name": "timestamp",
                // Type is optional, defaults to string
                "path": ".eventTime"
            },
            {
                "name": "temperature",
                "type": "float",
                // Path is optional, defaults to .temperature
            }
        ]
    }
}

Related content