Ensure that ObjectExecutorFactory resolves the handler that was used to generate metadata

[jakubfijalkowski]

Previously, the ObjectExecutorFactory resolved IXHandler<A> directly from the container. The problem with that approach is that we can't be sure that the TypeAssembly that we get in CQRSObjectsRegistrationSource.AddCQRSObjects represents all the assemblies that the user scanned for handlers when adding them to DI. For example, it is still possible to have this:

Contracts.csproj
   - Command : ICommand

HandlersA.csproj
  - CH1: ICommandHandler<Command>

HandlersB.csproj
  - CH2: ICommandHandler<Command>

Then in App have:

serviceCollection
    .AddCQRS(TypesAssembly.Of<Command>(), TypesAssembly.Of<CH1>())
    .AddCQRSObjects(TypesAssembly.Of<Command>(), TypesAssembly.Of<CH2>());

which would result, in the end, in (effectively, the code will be slightly different):

serviceCollection.AddTransient<ICommandHandler<Command>, CH1>();
serviceCollection.AddTransient<ICommandHandler<Command>, CH2>();

But the CQRSObjectsRegistrationSource.objects will have only the Command, CH1 pair. Thus, the pipeline will think that it executes CH1, but CH2 will be resolved in ObjectExecutorFactory.ExecuteCommandAsync.

I admit that this is rather unlikely situation, but when it occurs, it might be very hard to detect. I also don't think this PR makes detection easier, but at least it will resolve CH1 when it thinks about CH1. This might be more important for local execution though.

[github-actions]

Test Results

 35 files  ±0  120 suites  ±0   1m 5s ⏱️ -2s
662 tests +3  649 ✅ +3  13 💤 ±0  0 ❌ ±0 
678 runs  +3  658 ✅ +3  20 💤 ±0  0 ❌ ±0 

Results for commit 03011ef. ± Comparison against base commit 2b28316.

♻️ This comment has been updated with latest results.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (2b28316) 64.02% compared to head (628e4c3) 64.02%.

❗ Current head 628e4c3 differs from pull request most recent head 03011ef. Consider uploading reports for the commit 03011ef to get more accurate results

Additional details and impacted files

@@           Coverage Diff           @@
##             v8.0     #639   +/-   ##
=======================================
  Coverage   64.02%   64.02%           
=======================================
  Files         245      245           
  Lines        5823     5824    +1     
  Branches      376      376           
=======================================
+ Hits         3728     3729    +1     
  Misses       1986     1986           
  Partials      109      109           

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

[marcin-dardzinski]

serviceCollection
    .AddCQRS(TypesAssembly.Of<Command>(), TypesAssembly.Of<CH1>())
    .AddCQRSObjects(TypesAssembly.Of<Command>(), TypesAssembly.Of<CH2>());

Interesting corner case, shouldn't CQRSObjectsRegistrationSource throw on such cases. If I register command with two handlers there's clearly sth wrong.

[jakubfijalkowski]

Interesting corner case, shouldn't CQRSObjectsRegistrationSource throw on such cases.

It probably should, but it won't solve the problem - the user will still be able to overcome that with two calls to the AddCQRS or manual AddTransient<ICommandHandler<Command>, CH2>().