(original) (raw)

On 21 March 2017 at 17:09, Matthias Kramm <kramm@google.com> wrote:

The one thing that isn't clear to me is how type checkers will distinguish between
1.) Protocol methods in A that need to implemented in B so that B is considered a structural subclass of A.
2.) Extra methods you get for free when you explicitly inherit from A.

To provide a more concrete example: Since Mapping implements \_\_eq\_\_, do I also have to implement \_\_eq\_\_ if I want my class to be (structurally) compatible with Mapping?

An implicit subtype should implement all methods, so that yes, in this case \_\_eq\_\_ should be implemented for Mapping.

There was an idea to make some methods "non-protocol" (i.e. not necessary to implement), but it was rejected,
since this complicates things. Briefly, consider this function:

def fun(m: Mapping):

m.keys()

The question is should this be an error? I think most people would expect this to be valid.

The same applies to most other methods in Mapping, people expect that

they are provided my Mapping. Therefore, to be on the safe side, we need
to require these methods to be implemented. If you look at definitions in collections.abc,

there are very few methods that could be considered "non-protocol". Therefore, it was decided

to not introduce "non-protocol" methods.

There is only one downside for this: it will require some boilerplate for implicit subtypes of Mapping etc.

But, this applies to few "built-in" protocols (like Mapping and Sequence) and people already subclass them.

Also, such style will be discouraged for user defined protocols. It will be recommended to create compact

protocols and combine them. (This was discussed, but it looks like we forgot to add an explicit statement about this.)

I will add a section on non-protocol methods to rejected/postponed ideas.

--

Ivan