@frojasg thanks for the proposal, I personally don't have much of an opinion on how to judge criteria for adding this func in

I would love to make this contribution if you think is worth adding this func.

To add to the package, you'd need to motivate that this is so common that it would help real code. Do you have any stats on how often this comes up? My guess would have been that many RoundTrippers are parameterized by state and so they couldn't use this. (Also I don't think there are too many RoundTrippers at all, in contrast to http.Handlers.)

Introducing HandlerFunc was certainly a good thing, but each time you do introduce a FooFunc then you split the world into things using Foo and things using FooFunc. This is why we don't have io.ReaderFunc and so on.

Do you have tools to know how often this comes up? When is this uses case often enough to add it to the package?

Looking at Github, I can tell that 5% of the people using http.Client is using http.RoundTripper. Furthermore, there are 351 cases where people are already using the RoundTripperFunc Adaptor.

Do you think this initial numbers could motivate doing further investigation? If so, I would love to get some guidance where to go next. In the other hand, If you still think the cost is too high. We could close this ticket.

My guess would have been that many RoundTrippers are parameterized by state and so they couldn't use this.

I would have the same guess. I looked at the first page of @frojasg's 351 cases. There are three categories of results:

1. A utility package that defines RoundTripperFunc but doesn't actually use it anywhere. One example, and I cannot find an actual use of that package, although I probably fail at github search. This category of results is not useful.

2. A test that defines a stub RoundTripper. One example.

3. Code that defines a wrapper RoundTripper. One example. In each case like this one, the RoundTripperFunc does use state, but the state happens to come from the parent function context.

I'd be interested in knowing how often the third case happens in real code. Perhaps you could compute stats from rsc's go corpus? I did a quick search of Google's internal codebase and found that about 0.3% of custom RoundTrippers are a RoundTripperFunc. All other custom RoundTrippers are structs.

The evidence seems clear that this is not worthwhile.