[v3] Skip flattening if properties would clash

[thomas11]

Skip the flattening of nested properties indicated by x-ms-client-flatten if it would lead to overwriting a property, creating incorrect schema and SDKs. This case happens when inner and outer property have the same name. For a report on all occurrences see #3013.

This change is breaking and could therefore only be applied to v3 of the provider.

Resolves #3195

[github-actions]

Does the PR have any schema changes?

Looking good! No breaking changes found.
No new resources/functions.

[codecov]

Codecov Report

Attention: Patch coverage is 44.59459% with 41 lines in your changes missing coverage. Please review.

Project coverage is 56.87%. Comparing base (b5b0a21) to head (0bc1091).

Files with missing lines	Patch %	Lines
provider/pkg/gen/properties.go	44.59%	38 Missing and 3 partials ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #3801      +/-   ##
==========================================
- Coverage   57.00%   56.87%   -0.13%     
==========================================
  Files          78       78              
  Lines       12035    12078      +43     
==========================================
+ Hits         6860     6869       +9     
- Misses       4675     4709      +34     
  Partials      500      500              

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[mikhailshilkov]

Is there a way to see the effects in 3.0?

[thomas11]

Is there a way to see the effects in 3.0?

There should be, by combining this change and the v3 branch and running schemagen, but at the moment that fails for unrelated reasons.

[t0yv0]

codecov actually indicates a few lines that might be interesting to check out, but I live up to you to decide what's worth it.

[danielrbradley]

I'm also struggling to parse the diff. I also wonder if a first refactor commit to pull out a function could make the actual change more clear and remove the need for the extra level of nesting.

[thomas11]

I'm also struggling to parse the diff. I also wonder if a first refactor commit to pull out a function could make the actual change more clear and remove the need for the extra level of nesting.

I don't think this is a good investment of time. The code is quite convoluted and half of it (v2) will be removed soon anyways. It's more important here that there is a single code block to remove post v3 for simple cleanup, as Mikhail also pointed out above.

[thomas11]

Superseded by #3839