Allow omitting type and infer it when validating

[fizruk]

This fixes #138.

For type inference I use the same algorithm as in https://github.com/json-schema-faker/json-schema-faker#inferred-types:

• if type is present everything works like before;
• if type is absent, it is inferred based on properties used in schema;
• if no type can be inferred — validation fails;
• if more than one type is inferred — validation fails.

Note that this is a breaking change:

• users have to change type_ .~ ... to type ?~ ...
• mempty :: Schema is now with implicit type (was SwaggerString before).

This PR supersedes #163 (that one used a different inference mechanism).

[kahlil29]

Hey @fizruk @phadej
Just got hit by this issue (#138) while testing the parsing of Swagger specs. (I've written a tool that parses Swagger specs and generates some Haskell code)

Any idea if there's something blocking the merging of this PR?
Would really appreciate some insight into this 😄

[fizruk]

@phadej I have rebased, so it should probably be fine for merge now 😊

[kahlil29]

@phadej A gentle reminder about this, whenever you get the time 😸

[kahlil29]

@phadej
Do let us know if there's anything pending in order for this to be merged 😄

[kahlil29]

@phadej ping! 😄
Hope I'm not pestering you too much 😛

[phadej]

Merged. But I don't know when I'll have proper time to make the release (it's including major bump, so one would like to use that opportunity to merge other breaking stuff)

[phadej]

Ok. I accidentally removed the inferring when rebasing #163, and none of the tests failed. I.e. there are no tests.

Reverting this until there are.

[kahlil29]

@phadej Cool, thanks a lot.
I'm currently using it as an extra-dep and pointing to the latest commit on this branch. Guess I'll continue doing that till this is released.

Do tests need to be written? I could probably try and help with that, if necessary.
@fizruk

[phadej]

@kahlil29 validateToJSON is tested only with doctests in Schema.Validation; given how much code is in Internal/Schema/Validation.hs.

Specifically I don't see any test where type is omitted and schema is inferred.