Fix dedicated io transform leaving unused shapes #1419
Conversation
| output: GetFooOutput | ||
| } | ||
|
|
||
| structure GetFooOutput {} |
There was a problem hiding this comment.
Can the naming of this output structure get updated so it's clear that it will be removed in the -after.smithy file?
There was a problem hiding this comment.
The point of this test is to show that it we don't get a GetFooOperationOutput shape after the transform. Before this change, the transformer would think that because the GetFooOutput shape is being used in two places, it should create a new shape for the GetFoo output, when in reality the GetFoo input is going to be changed and thus it should have kept the GetFooOutput shape and just added the @output trait
There was a problem hiding this comment.
I added a second test that verifies that the unused shape will be removed, I'll leave this open to make sure that looks ok
...mithy/model/transform/dedicated-input-output/handles-dangling-single-reference-before.smithy
Outdated
Show resolved
Hide resolved
...odel/src/main/java/software/amazon/smithy/model/transform/CreateDedicatedInputAndOutput.java
Outdated
Show resolved
Hide resolved
This commit changes the way `CreateDedicatedInputOutput` transform checks for singular references to a shape being used as input or output in an operation. It now checks to make sure the shape is only referenced by the specified operation. This fixes an edge case where a shape was being used as both the input and output for an operation, and would be left unused in the model after the transformation was applied. Two new test cases were added to verify the behavior when a shape is used as both input and output.
milesziemer
force-pushed
the
fix-create-dedicated-io-transformer
branch
from
faf1d2f to
e25db28
Compare
This commit changes the way
CreateDedicatedInputOutputtransform checks for singular references to a shape being used as input or output in an operation. It now checks to make sure the shape is only referenced by the specified operation. This fixes an edge case where a shape was being used as both the input and output for an operation, and would be left unused in the model after the transformation was applied. Two new test cases were added to verify the behavior when a shape is used as both input and output.Issue #, if available: #1373
By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.