PySpark Transform
PySpark transforms execute PySpark code to transform data.
PySparkTransform
PySparkTransform is defined beneath the following ancestor nodes in the YAML structure:
Below are the properties for the PySparkTransform. Each property links to the specific details section further down in this page.
| Property | Default | Type | Required | Description |
|---|---|---|---|---|
| event_time | string | No | Timestamp column in the component output used to represent event time. | |
| microbatch | boolean | No | Whether to process data in microbatches. | |
| batch_size | string | No | The size/time granularity of the microbatch to process. | |
| lookback | 1 | integer | No | The number of time intervals prior to the current interval (and inclusive of current interval) to process in time-series processing mode. |
| begin | string | No | The 'beginning of time' for this component. If provided, time intervals before this time will be skipped in a time-series run. | |
| inputs | array[InputComponent] | No | List of input names to use in the SQL query. | |
| strategy | Any of: PartitionedStrategy IncrementalStrategy string ("view", "table") | No | Transform strategy - incremental, partitioned, or view/table. | |
| pyspark | PythonTransformComponent | No | PySpark transform function to execute for transforming the data. |
Property Details
Component
A component is a fundamental building block of a data flow. Types of components that are supported include: read, transform, task, test, and more.
| Property | Default | Type | Required | Description |
|---|---|---|---|---|
| component | One of: ReadComponent TransformComponent TaskComponent SingularTestComponent CustomPythonReadComponent WriteComponent CompoundComponent AliasedTableComponent ExternalTableComponent | Yes | Configuration options for the component. |
TransformComponent
A component that executes SQL or Python code to transform data.
| Property | Default | Type | Required | Description |
|---|---|---|---|---|
| data_plane | One of: SnowflakeDataPlane BigQueryDataPlane DuckdbDataPlane SynapseDataPlane FabricDataPlane DatabricksDataPlane | No | Data Plane-specific configuration options for a component. | |
| description | string | No | A brief description of what the model does. | |
| metadata | ResourceMetadata | No | Meta information of a resource. In most cases it doesn't affect the system behavior but may be helpful to analyze project resources. | |
| name | string | Yes | The name of the model | |
| flow_name | string | No | The name of the flow that the component belongs to. | |
| skip | boolean | No | A boolean flag indicating whether to skip processing for the component or not. | |
| data_maintenance | DataMaintenance | No | The data maintenance configuration options for the component. | |
| tests | ComponentTestOptions | No | Defines tests to run on the data of this component. | |
| transform | One of: SqlTransform PythonTransform SnowparkTransform PySparkTransform | Yes | The transform component that executes SQL or Python code to transform data. |
PythonTransformComponent
Python transform function to execute for transforming the data.
| Property | Default | Type | Required | Description |
|---|---|---|---|---|
| entrypoint | string | Yes | The entrypoint for the python transform function. | |
| source | string | Yes | The source file for the python transform function. |
IncrementalStrategy
Incremental Processing Strategy.
| Property | Default | Type | Required | Description |
|---|---|---|---|---|
| incremental | Any of: string MergeStrategy SCDType2Strategy | Yes | Incremental processing strategy. | |
| on_schema_change | string ("ignore", "fail", "append_new_columns", "sync_all_columns") | No | Policy to apply when schema changes are detected. Defaults to 'fail' if not provided. |
PartitionedStrategy
Partitioned Ingest Strategy. The user is expected to provide 2 functions, a list function that lists partitions in the source, and a read function that reads a partition from the source.
| Property | Default | Type | Required | Description |
|---|---|---|---|---|
| partitioned | PartitionedOptions | No | Options for partitioning data. | |
| on_schema_change | string ("ignore", "fail", "append_new_columns", "sync_all_columns") | No | Policy to apply when schema changes are detected. Defaults to 'fail' if not provided. |
PartitionedOptions
Options related to partition optimization - in particular, the policy that determines which partitions to ingest.
| Property | Default | Type | Required | Description |
|---|---|---|---|---|
| enable_substitution_by_partition_name | boolean | Yes | Enable substitution by partition name. | |
| output_type | table | string ("table", "view") | No | Output type for partitioned data. Must be either 'table' or 'view'. This strategy applies only to Transforms. |
SCDType2Strategy
The SCD Type 2 strategy allows users to track changes to records over time, by tracking the start and end times for each version of a record. A brief overview of the strategy can be found at https://en.wikipedia.org/wiki/Slowly_changing_dimension#Type_2:_add_new_row.
| Property | Default | Type | Required | Description |
|---|---|---|---|---|
| scd_type_2 | KeyOptions | No | Options for SCD Type 2 strategy. |
InputComponent
Specification for input components, including how partitioning behaviors should be handled. This additional metadata is required when a component is used as an input to other components in a flow.
| Property | Default | Type | Required | Description |
|---|---|---|---|---|
| flow | string | Yes | Name of the parent flow that the input component belongs to. | |
| name | string | Yes | The input component name. | |
| alias | string | No | The alias to use for the input component. | |
| partition_spec | Any of: string ("full_reduction", "map") RepartitionSpec | No | The type of partitioning to apply to the component's input data before processing the component's logic. Input partitioning is applied before the component's logic is executed. | |
| where | string | No | An optional filter condition to apply to the input component's data. | |
| partition_binding | Any of: string PartitionBinding | No | An optional partition binding specification to apply to the component on a per-output-partition basis against other inputs' partitions. |
MergeStrategy
A strategy that involves merging new data with existing data by updating existing records that match the unique key.
| Property | Default | Type | Required | Description |
|---|---|---|---|---|
| merge | KeyOptions | No | Options for merge strategy. |
KeyOptions
Column options needed for merge and SCD Type 2 strategies, such as unique key and deletion column name.
| Property | Default | Type | Required | Description |
|---|---|---|---|---|
| unique_key | string | Yes | Column or comma-separated set of columns used as a unique identifier for records, aiding in the merge process. | |
| deletion_column | string | No | Column name used in the upstream source for soft-deleting records. Used when replicating data from a source that supports soft-deletion. If provided, the merge strategy will be able to detect deletions and mark them as deleted in the destination. If not provided, the merge strategy will not be able to detect deletions. | |
| merge_update_columns | Any of: string array[string] | No | List of columns to include when updating values in merge. These columns are mutually exclusive with respect to the columns in merge_exclude_columns. | |
| merge_exclude_columns | Any of: string array[string] | No | List of columns to exclude when updating values in merge. These columns are mutually exclusive with respect to the columns in merge_update_columns. | |
| incremental_predicates | Any of: string array[string] | No | List of conditions to filter incremental data. |
PartitionBinding
| Property | Default | Type | Required | Description |
|---|---|---|---|---|
| logical_operator | logical_operator | string ("AND", "OR") | No | The logical operator to use to combine the partition binding predicates provided |
| predicates | predicates | array[string] | No | The list of partition binding predicates to apply to the input component's data |
RepartitionSpec
Specification for repartitioning operations on input component's data
| Property | Default | Type | Required | Description |
|---|---|---|---|---|
| repartition | RepartitionOptions | No | Options for repartitioning the input component's data. |
RepartitionOptions
Options for repartitioning the input component's data.
| Property | Default | Type | Required | Description |
|---|---|---|---|---|
| partition_by | string | Yes | The column to partition by. | |
| granularity | string | Yes | The granularity to use for the partitioning. |