store.ExecuteIndexes vs IndexCreation.CreateIndexes

[RJacobsRN]

Recently whilst changing indexes in our code I came across a wholly unhelpful exception when an error occurs in the creation of an index. Tracking the down the root cause of this leads me to the usage of IndexCreation.CreateIndexes.

The exception that we get, when we have an index with an error is the following:
System.ArgumentException: An item with the same key has already been added. Key: __all_fields

This exception occurs regardless of which index is errored, and happens in the same index every single time.

Changing the usage of IndexCreation.CreateIndex to store.ExectuteIndexes results in the following exception:
IndexCompilationException: Failed to compile index CandidateListIndex .......
Followed by a detailed accounting of what errors are occuring during the index compilation

As you can imagine, the second case is vastly more helpful.
Which leads to my question, what is the difference between IndexCreation.CreateIndexes and store.ExecuteIndexes?
Is there any advantage to using the IndexCreation.CreateIndexes method that we might be missing out on when switching this over?
Or is there some possible way to make sure IndexCreation.CreateIndexes also throws useful exceptions in case of failures

[ppekrol]

I've created an issue, in both cases you should get a meaningful exception. You can track it here: https://issues.hibernatingrhinos.com/issue/RavenDB-24077

The difference is that IndexCreation.CreateIndexes on failure will process indexes one by one, and DocumentStore.ExecuteIndexes will just fail (there is a batching mechanism inside the endpoint it hits - 50 indexes)

[RJacobsRN]

Thanks for looking into it!
I assume that IndexCreation.CreateIndexes performs better when putting a large number of indexes.
I'll switch our code over to ExecuteIndexes for now until the bug fix for the exception, as it shouldn't impact our deploy flow that much.