Hi I ve been playing a bit more with my POC for data class d

Question

Hi I ve been playing a bit more with my POC for data class diffs making it recursive for nested data classes got it looking like ```data class diff for com sksamuel kotest matchers Baz booFoo data class diff for com sksamuel kotest matchers BooFoo a expected lt 9L gt but was lt 8L gt foo data class diff for com sksamuel kotest matchers Foo a expected lt 123L gt but was lt 321L gt b expected lt a gt but was lt b gt c expected lt a gt but was lt c gt i expected lt 66 gt but was lt 67 gt things expected lt cheese 22 gt but was lt egg 1 gt ``` however it s lots of string manipulation it uses `shouldBe` to choose the correct `Eq` instance Then it gets the string from the the `Throwable message` and if it s a nested data class it does a little more manipulation My only thoughts to make this a little cleaner would be to change the `Eq` interface and the instances such that instead of returning a `Throwable ` it returns a type which contains the differences That way we don t have string manipulate as we d have access to the diff data However this might complicate things for the purpose of one feature Be good to get the thoughts of people a bit more familiar with the codebase I m sure there s a much simper way I just don t can t see it see no evil If you want to see my extremely <https github com kotest kotest compare master jaymoid 826 data class diff expand=1 diff 7cd6841f5e692a251650d89f54963354|hacky POC code> for reference

sam · Answer

I ll review today

dr.dreigh · Answer

Cheers Sam it s totally just a rough POC at the moment very hacky to see how far I could get and if I can do anything better without changing the Eq interface Your eyes would be greatly appreciated Changing EQ seems a bit of a big change though for a nicer diff so didn t want to just go ahead ripping things apart

sam · Answer

Sorry for delay in reviewing

sam · Answer

I agree with you Eq should return a better object

sam · Answer

I think what we need is a subclass of ```AssertionError``` which contains the error the expected and the actual values

sam · Answer

Then Eq could return that

sam · Answer

This is similar to tricks that are done elsewhere for example in the console runner ```when error is org opentest4j AssertionFailedError gt if error isActualDefined amp amp error isExpectedDefined type comparisonFailure actual error actual stringRepresentation expected error expected stringRepresentation ```

sam · Answer

`AssertionFailedError` is the opentest4k s version of an error that contains actual and expected

dr.dreigh · Answer

Cool thanks I ll have a go at that Should give me more structure to work with

dr.dreigh · Answer

Hi think I have hit a bit of a stumbling block with this I have no way of knowing if a data class has overridden equals In which case a detailed diff of all the fields declared in the constructor would be very irrelevant

sam · Answer

I think it s fair to assume a data class doesn t have an overriden equals

sam · Answer

If it does then that s circumventing normal expectations

sam · Answer

The whole point of a data class is the equals hashcode and component methods

sam · Answer

We can also use a flag to disable this behavior if needed

dr.dreigh · Answer

ok cool I ll crack on Yeah I must admit we don t override equals in our codebase and I had to even second guess if it was possible to override equals in a data class but I can t see any way via reflection to determine where the equals method is defined I m still a little unsure of the best way to do the recursive bit nicely but I m sure I m come up with several nasty ways and we can choose the least worst

sam · Answer

Ok