ifExists=FIND_OR_CREATE_VERSION and firstVersion behaviour

[jonsnowseven]

Hello.

I am trying to understand the behaviour of ifExists=FIND_OR_CREATE_VERSION and firstVersion.

If I create a Schema with the following AVRO content: "{\"type\":\"record\",\"name\":\"ExampleType\",\"fields\":[{\"name\":\"sdfgfsdgsdg\",\"type\":\"string\"},{\"name\":\"field2\",\"type\":\"int\"},{\"name\":\"field3\",\"type\":\"string\",\"default\":\"\"}]}" with a firstVersion "1".

If I create the Schema again with the same content, version "1" is returned.

If I create the Schema again with a different but compatible content, there is an error: VersionAlreadyExistsException: An artifact version '1' for artifact ID ... in group ... already exists.. I was expecting that a new version would be created in this case.

My question is, how does ifExists and firstVersion co-exist?

Thank you.

[EricWittmann]

When using FIND_OR_CREATED_VERSION as the value for ifExists, you are telling the server what to do when an artifact you are trying to create already exists. It's only useful when providing something in firstVersion, since it is not possible to find or create a version without version content provided in the request.

The logic should be this:

• If the artifact does not exist, create the artifact and add the version as version 1
• If the artifact does exist,
  • If a version with the given content already exists, return it
  • If a version with the given context does NOT exist, create a new version and return that

You can also optionally add the canonical=true query parameter if you want the "does version exist" check to be less strict.

I just threw together a unit test to make sure everything was working as I expected (we have tests for this already but I wanted to gather all the variants together in one place to make sure it works and to illustrate it better for you):

https://gist.github.com/EricWittmann/a4b606d622b6a936ea64bbd379490408

It's possible you're doing something slightly different than this? Let me know and I will add appropriate tests.

[jonsnowseven]

I am questioning this point If a version with the given context does NOT exist, create a new version and return that because the version string I provide in the firstVersion parameter is the same as the one that already exists.
Is that the case?

[EricWittmann]

It is supposed to work as you have described. I believe my unit test that I linked above confirms that. Can you provide a reproducer?

[jonsnowseven]

Hey @EricWittmann
I think a spotted a difference: in your unit test, you are not setting the version string in the firstVersion, right? If you set it, what happens?