What's the issue?
When using observable_source_asset and dataset_partitions_test, the input_data_version appears to be broken
What did you expect to happen?
After running the below example and viewing the partition in the dagster ui, the partitions appear to have an input_data_version/code_partitions_test value of e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855 and an input_data_version/dataset_partitions_test value of e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855. But this sha256sum is that of an empty file (e.g. cat /dev/null | sha256sum) meaning that the sha256sum is meaningless. Ideally, I would prefer these to have the values given by dg.DataVersionsByPartition (e.g. 3) for interpretability.
How to reproduce?
import dagster as dg
import time
datasets_test = dg.DynamicPartitionsDefinition(name="datasets_test")
codes_test = dg.DynamicPartitionsDefinition(name="codes_test")
effctivepartitions = dg.MultiPartitionsDefinition({
"datasets_test": datasets_test,
"codes_test": codes_test
})
@dg.observable_source_asset(partitions_def=datasets_test)
def dataset_partitions_test(context):
context.instance.add_dynamic_partitions(datasets_test.name, ["a", "b"])
out = {"a": dg.DataVersion(str(time.time())), "b": dg.DataVersion("2")}
return dg.DataVersionsByPartition(out)
@dg.observable_source_asset(partitions_def=codes_test)
def code_partitions_test(context):
context.instance.add_dynamic_partitions(codes_test.name, ["c", "d"])
out = {"c": dg.DataVersion("3"), "d": dg.DataVersion("4")}
return dg.DataVersionsByPartition(out)
@dg.asset(partitions_def=effctivepartitions, deps=[dataset_partitions_test, code_partitions_test])
def execution_test(context):
return dg.MaterializeResult()
defs = dg.Definitions(
assets=[dataset_partitions_test, code_partitions_test, execution_test],
)
Dagster version
dagster, version 1.13.10
Deployment type
None
Deployment details
K8S, Helm
Additional information
No response
Message from the maintainers
Impacted by this issue? Give it a 👍! We factor engagement into prioritization.
What's the issue?
When using observable_source_asset and dataset_partitions_test, the input_data_version appears to be broken
What did you expect to happen?
After running the below example and viewing the partition in the dagster ui, the partitions appear to have an
input_data_version/code_partitions_testvalue ofe3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855and aninput_data_version/dataset_partitions_testvalue ofe3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855. But this sha256sum is that of an empty file (e.g.cat /dev/null | sha256sum) meaning that the sha256sum is meaningless. Ideally, I would prefer these to have the values given bydg.DataVersionsByPartition(e.g. 3) for interpretability.How to reproduce?
Dagster version
dagster, version 1.13.10
Deployment type
None
Deployment details
K8S, Helm
Additional information
No response
Message from the maintainers
Impacted by this issue? Give it a 👍! We factor engagement into prioritization.