** MIGRATED ** - New maintainer consider own comments below & comments of upstream

Original PR: @eh-steve gorilla#676

Summary of Changes

Added r.GetVarNames() function to list all vars a route might need in order to call r.URL()

Imagine an application that makes HTTP requests based on variables coming from an inbound messages (e.g. from a kafka topic). It could take a set of YAML configurations like:

endpoint:
  host_pattern: "www.{domain}.com"
  path_pattern: "/some/prefix/{group}/{item_id:[0-9]+}"
  method: "POST"
  queries:
    blah: "{some_data1:[0-9]+}"
    something: "{some_data2:[0-9]+}"

and builds up mux routes programatically using these configured endpoints.

Then when a payload comes in from an inbound message like:

{
  "domain": "example",
  "group": "my_group",
  "item_id": 75,
  "some_data1": 123,
  "some_data2": 456,
  "other_irrelevant_data": {
    "massive_and_expensive_to_stringify": "..."
  }
}

It would be able to extract keys from the JSON payload (parsed as a map[string]interface{}) and use them as URL vars. Mux vars values need to be strings, so any non-string values would need to be fmt.Sprint()'ed (e.g. item_id in this example).

If your JSON payload contains any additional fields which aren't needed as route vars (e.g. other_irrelevant_data in this example, which contains the field massive_and_expensive_to_stringify), we would still need to stringify them all (which involves a needless allocation) and pass them into route.URL(pairs...), where they would be ignored anyway.

By exposing GetVarNames(), it allows us to only pass variables which we know are required by the route, and to error early if those variables are not available, or don't adhere to a schema etc.

This is a simplified example compared to how we're actually using it, but it illustrates the point.