feat: support nullable input type arguments

[simoneb]

Closes #30

Note that this is the fastest solution to the issue I came up with but I'm not 100% sure that this covers all scenarios or that it doesn't break anything.

The change is minimal but has cascading consequences on many assertions in tests because most types and fields are nullable, therefore will see their json-schema type change from 'whatever' to ['whatever', 'null'].

The scenario raised by the OP is covered and works fine.

[mcollina]

lgtm, this is quite a serious bug indeed.

[mcollina]

@jonnydgreen wdyt? what semversiness this should be?

I'm tempted for a patch and a backport to previous lines too, but it's also a big change.

[jonnydgreen]

LGTM - thanks for sorting this!

[jonnydgreen]

I'm tempted for a patch and a backport to previous lines too, but it's also a big change.

Yeah, I agree with the patch and backport approach - especially as it's fixing spec compliance. Thinking the following?

• 2.0.0 => 2.0.1
• 1.0.0 => 1.0.1

[mcollina]

That's the idea.

[jonnydgreen]

Apologies, I missed this message as I was afk all weekend with the intention of doing this first thing Monday - thank you for releasing these fixes earlier! :)

[mcollina]

no worries, we needed these out asap

[paolochiodi]

Just tested this in my code and works as expected.
Thanks @simoneb and rest of the team.