source/extensions/filters/http/response_map: support disabling using per route config

[esmet]

Copies the pattern in source/extensions/filters/http/grpc_http1_reverse_bridge/*

[LukeShu]

Copies the pattern in source/extensions/filters/http/grpc_http1_reverse_bridge/*

That would be a good thing to put in the commit message, FWIW.

[LukeShu]

I see those validate. options on a few other filters (but not the grpc_http1_reverse_bridge filter that you mentioned cribbing the approach from), so I guess I trust that they're correct, but I'm flummoxed as to interpreting what they actually mean.

[esmet]

i tested this and it means disabled must be a bool equal to true. false is not valid.

[LukeShu]

Why shouldn't false be valid? And if it's validate.required, then how do you omit it (I guess you just leave out the entire ResponseMapPerRoute?)?

[esmet]

i think the idea is to omit if you want it to be not disabled. it's opinionated for sure.

[LukeShu]

As simple as this block is, I found it hard to read. I guess that's just to be expected when you have type names that are 71 characters log. And I guess that's just to be expected with C++/Envoy :/

Does 1 arg per line make it better?

Router::RouteSpecificFilterConfigConstSharedPtr
ResponseMapFilterFactory::createRouteSpecificFilterConfigTyped(
    const envoy::extensions::filters::http::response_map::v3::ResponseMapPerRoute& proto_config,
    Server::Configuration::ServerFactoryContext&,
    ProtobufMessage::ValidationVisitor&) {
  return std::make_shared<FilterConfigPerRoute>(proto_config);
}

Maybe? But does the long proto_config line run afoul of some automated style checker?

[LukeShu]

Should that be , and disabled =? IDK what style is, I'm just noting that there was an and there before.

[esmet]

i removed the and in favor of comma separating the 3 values

[LukeShu]

lgtm, but I do want to understand the protobuf options.

[kflynn]

Looks OK, I think.