[
https://issues.apache.org/jira/browse/BEAM-12413?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Dmytro Kozhevin updated BEAM-12413:
-----------------------------------
Description:
Returning a slice of the tuple doesn't alter the type inference output. For
example:
{{@beam.typehints.with_output_types(Tuple[int, str])}}
{{class Foo(beam.DoFn):}}
{{ def process(self, a):}}
{{ yield a, str(a)}}
{{pcoll = beam.Pipeline() }}| beam.Create([1]) | beam.ParDo(Foo()) |
beam.Map(lambda x: x[1:])}}
{{print(pcoll.element_type)}}
This prints Tuple[int, str], while the real type is str. At worst, this should
return Any.
was:
Returning a slice of the tuple doesn't alter the type inference output. For
example:
{{@beam.typehints.with_output_types(Tuple[int, str])}}
{{class Foo(beam.DoFn):}}
{{ def process(self, a):}}
{{ yield a, str(a)}}
{{pcoll = }}{{beam.Pipeline()}}{{ | beam.Create([1]) | beam.ParDo(Foo()) |
beam.Map(lambda x: x[1:])}}
{{print(pcoll.element_type)}}
This prints Tuple[int, str], while the real type is str. At worst, this should
return Any.
> Type inference works incorrectly with tuple slices
> --------------------------------------------------
>
> Key: BEAM-12413
> URL: https://issues.apache.org/jira/browse/BEAM-12413
> Project: Beam
> Issue Type: Bug
> Components: sdk-py-core
> Reporter: Dmytro Kozhevin
> Priority: P2
>
> Returning a slice of the tuple doesn't alter the type inference output. For
> example:
>
> {{@beam.typehints.with_output_types(Tuple[int, str])}}
> {{class Foo(beam.DoFn):}}
> {{ def process(self, a):}}
> {{ yield a, str(a)}}
> {{pcoll = beam.Pipeline() }}| beam.Create([1]) | beam.ParDo(Foo()) |
> beam.Map(lambda x: x[1:])}}
> {{print(pcoll.element_type)}}
>
> This prints Tuple[int, str], while the real type is str. At worst, this
> should return Any.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
