You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Refactor and adapt the existing testing framework to use the already generated complete OpenAPI specification rather than the manually captured individual snippets.
What problems are you trying to solve?
Currently to add a new test requires manually extracting each individual operation as a separate OpenAPI model. These then can fall out of sync with the originating Smithy model unless they're manually updated, and therefore aren't necessarily testing what is in the Smithy spec. This also adds a lot of friction for adding test cases.
Why should it be built? Any reason not to?
If we reduce friction in the process for adding tests we can better enforce testing of newly added specs.
This is (appears to be) a relatively low-hanging fruit to reduce a major friction point. However, there may be a more fit for purpose pre-existing testing framework/strategy that would be better to switch to rather than improving the existing one.
The text was updated successfully, but these errors were encountered:
Closing this since Smithy is no longer used and the test framework has been removed as a result.
The new coverage workflow will also help us uncover missing endpoints. We can add a new test framework to test the spec against the server if need to. But in my opinion, it should be the other way around: the server should be tested against the spec, and every update to OpenSearch features should start with the spec.
What/Why
What are you proposing?
Refactor and adapt the existing testing framework to use the already generated complete OpenAPI specification rather than the manually captured individual snippets.
What problems are you trying to solve?
Currently to add a new test requires manually extracting each individual operation as a separate OpenAPI model. These then can fall out of sync with the originating Smithy model unless they're manually updated, and therefore aren't necessarily testing what is in the Smithy spec. This also adds a lot of friction for adding test cases.
Why should it be built? Any reason not to?
The text was updated successfully, but these errors were encountered: