[sdk/python] Handle unknown results from methods #7677
Conversation
|
|
||
| resolve_value.set_result(value) | ||
| resolve_is_known.set_result(True) | ||
| resolve_is_known.set_result(is_known) |
There was a problem hiding this comment.
I feel we're coding against a very low-level interface here so it sounds like it was an easy mistake to make.
tests/integration/construct_component_methods_unknown/testcomponent-python/testcomponent.py
Outdated
Show resolved
Hide resolved
|
Curious, the integration tests are fine here but how do you feel about using Mock-based tests instead? I think this could run against a MockMonitor and exercise the changes in rpc.py but maybe I'm mistaken. They're a bit more tedious but run faster. Or would that lose essential coverage? |
Update the existing Python methods test to include a type check in the method result class, to match what would be generated from the SDK, for some extra coverage.
justinvp
force-pushed
the
justin/call_unknown
branch
from
ff07ffe
to
7b9539f
Compare
Good point. We could probably have Mock-based tests for this, but I think I'd still additionally want integration tests to test E2E. Also, at some point soon I hope to be able to add SDK generation for many of these integration tests (generate SDKs for the provider resources based on a schema, to fully exercise the codegen + runtime), so in some ways this is setting the stage for that. |
This fixes handling of unknown results from methods in the Python SDK. Node.js and Go were already handling this appropriately.
This adds tests in each of the languages. The test program passes an unknown value to the provider, and asserts that an
applyis not run for it, and then returns the value, and the program asserts that anapplydoes not run on the result. The Python test program fails before the fix and succeeds after.Fixes #7670