[
https://issues.apache.org/jira/browse/BEAM-4441?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16969430#comment-16969430
]
Chad Dombrova commented on BEAM-4441:
-------------------------------------
Quick thought on this: the {{Tuple[int, ...]}} case will be harder to solve
than {{List[int]}} because if we use the standard {{IterableCoder}} (as we did
for list) the value will be round-tripped as a list. So we either accept that
limitation or we need to use non-standard coders.
I think it would also be interesting to support set and frozenset.
> Incorrect coder inference for List and Tuple typehints.
> -------------------------------------------------------
>
> Key: BEAM-4441
> URL: https://issues.apache.org/jira/browse/BEAM-4441
> Project: Beam
> Issue Type: Improvement
> Components: sdk-py-core
> Reporter: Valentyn Tymofieiev
> Assignee: Udi Meiri
> Priority: Minor
>
> We seem to use a FastPrimitivesCoder for List and Tuple typehints with
> homogenous element types, and fail to do the type checking:
> {code:python}
> inputs = (1, "a string")
> coder = typecoders.registry.get_coder(typehints.Tuple[int, str])
> print(type(coder)) # <class 'apache_beam.coders.coders.TupleCoder'>
> encoded = coder.encode(inputs)
> # Fails: TypeError: an integer is required - correct behaviour
> coder = typecoders.registry.get_coder(typehints.Tuple[int, ...]) # A tuple of
> integers.
> print(type(coder)) # <class 'apache_beam.coders.coders.FastPrimitivesCoder'>
> - wrong coder?
> encoded = coder.encode(inputs)
> # No errors - incorrect behavior.
> coder = typecoders.registry.get_coder(typehints.List[int]) # A list of
> integers.
> print(type(coder)) # <class 'apache_beam.coders.coders.FastPrimitivesCoder'>
> - wrong coder?
> encoded = coder.encode(inputs)
> # No errors - incorrect behavior.{code}
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
