checksum_record Transform

The checksum_record transform creates a checksum based on values in a record and saves it in the factorytx_checksum column. This field will be saved outside of the data.fieldvalues of a record and moved outside of a record’s data structure upon being received by a MDP environment. It is recommended to place this transform last in the list, so that the record has all of its data as well as an explicit timestamp column.

Example:

If we want to create a checksum from the asset and timestamp columns of a record, our configuration will look something like this:

{
    "transform_name": "Checksum",
    "transform_type": "checksum_record",
    "filter_stream": ["*"],
    "record_keys": ["asset", "timestamp"]
}

Configuration:

Required and optional properties that can be configured for a hash_record transform.

• transform_name: Unique name for the transform.

• transform_type: Type of transform to apply. Should be checksum_record.

• filter_stream: List of data streams to transform. Each stream can either be * (all) or asset:stream.

• record_keys: List of keys in a record for generating a checksum of the record. Supported keys are: asset, stream_type, timestamp, and fieldvalues (values in a record). Please refer to Caveats section for more implementation details.

Caveats:

• record_keys order: The order of the specified keys in the record_keys config setting will affect the checksum. For example, checksums from [‘asset’, ‘stream_type’] will not equal checksums from [‘stream_type’, ‘asset’].

• timestamp hash: The value in the timestamp column will be converted into a string for hashing. The data type of the value will affect the hash value. For example, a datetime object (datetime(2019, 9, 30, 12, 12, 12)) will not yield the same hash as its equivalents in ISO-8061 string format ('2019-09-30T12:12:12') and epoch time format (1569845532.0). If there is no timestamp value or if it is NaT, the timestamp will not be included in the checksum.

• fieldvalues hash: Values in a record will be sorted in alphanumerical order by their keys and then serialized into a JSON string, using this conversion table.