Documentation ¶
Overview ¶
Package regression contains pipeline regression tests.
Index ¶
- Constants
- func DirectParDo() *beam.Pipeline
- func DirectParDoAfterGBK() *beam.Pipeline
- func EmitParDo() *beam.Pipeline
- func EmitParDoAfterGBK() *beam.Pipeline
- func LPErrorPipeline(s beam.Scope) beam.PCollection
- func LPErrorReshufflePipeline(s beam.Scope) beam.PCollection
- func MixedOutputParDo() *beam.Pipeline
- func MultiEmitParDo() *beam.Pipeline
Constants ¶
const ( // MetricNamespace is the namespace for regression test metrics. MetricNamespace = string("regression") // FruitCounterName is the name of the fruit counter metric. FruitCounterName = string("fruitCount") )
Variables ¶
This section is empty.
Functions ¶
func DirectParDoAfterGBK ¶
DirectParDoAfterGBK generates a pipeline with a direct-form ParDo after a GBK. See: BEAM-3978 and BEAM-4175.
func EmitParDoAfterGBK ¶
EmitParDoAfterGBK generates a pipeline with a emit-form ParDo after a GBK. See: BEAM-3978 and BEAM-4175.
func LPErrorPipeline ¶
func LPErrorPipeline(s beam.Scope) beam.PCollection
LPErrorPipeline constructs a pipeline that has a GBK followed by a CoGBK using the same input, with schema encoded structs as elements. This ends up having the stage after the CoGBK fail since the decoder post-cogbk is missing a Length Prefix coder that was applied to the GBK input, but not the CoGBK output. Root is likely in that there's no Beam standard CoGBK format for inject and expand. JIRA: BEAM-12438
func LPErrorReshufflePipeline ¶
func LPErrorReshufflePipeline(s beam.Scope) beam.PCollection
LPErrorReshufflePipeline checks a Row type with reshuffle transforms. It's intentionally just a prefix with validation done in the specific test cases, as the success/failure is dependent on subsequent pipeline use of data.
This pipeline will output a pcollection containing 3 fruit.
func MixedOutputParDo ¶
MixedOutputParDo tests mixed direct + emit form output DoFns.
func MultiEmitParDo ¶
MultiEmitParDo tests double emit form output DoFns.
Types ¶
This section is empty.