[build] Allow tip failures

* Document router.Match

The return values are getting confusing. Hopefully this helps.

* Simplify some language.

* Remove double the

#306 changed UseEncodedPath to use native go encoded path handling so cautions in it's docs are no longer applicable.

* Test method-based subrouters for multiple matching paths

* Pass TestMethodsSubrouter

* Change http.Method* constants to string literals
- Make compatible with Go v1.5

* Make TestMethodsSubrouter stateless and concurrent

* Remove t.Run and break up tests for concurrency

* Use backticks to remove quote escaping

* Remove global method handlers and HTTP method constants

* [docs] Note StrictSlash re-direct behaviour #308

* StrictSlash enabled routes return a 301 to the client
* As per the HTTP standards, non-idempotent methods, such as POST or PUT, will be followed with a GET by the client
* Users should use middleware if they wish to change this behaviour to return a HTTP 308.

* Update description of StrictSlash

* Add example usage for Route.HeadersRegexp

* Improve example_route_test.go style

* Add a function to set url params for test

* [docs] add justification for SetURLVars and description of alternative approach to setting url vars.
* rename SetURLParams to SetURLVars as this is more descriptive.
* rename testing to testing_helpers as this is more descriptive.

* [docs] add stipulation to SetURLVars that it should only be used for testing purposes

The existing options matchPrefix, matchHost, and matchQueries are
mutually exclusive so there's no point in having a separate boolean
argument for each one. It's clearer if there's a single type variable.

strictSlash and useEncodedPath were also being passed as naked bools so
I've wrapped these in a struct called routeRegexpOptions for more clarity
at the call site.

* mux.Router now has a `Use` method that allows you to add middleware to request processing.

Addresses #294 (comment)

Fixes #323.

Also removed the duplicate "listing routes" example.

They actually return an error instead of an empty list. `GetMethods` happened to not return an error, but it should for consistency, so I added that as well.

* [docs] Clarify SetURLVars

Clarify in documentation that SetURLVars does not modify the given
*htttp.Request, provide an example of usage.

* Short and sweet function doc, example test.

* Create authentication middleware example.

For #339

* Fix example test filename.

* Modify http status code to variable

* Modify doc

* Modify http status code to variable

* Modify doc

* Modify README

Enables neater syntax when chaining several middleware functions.

Prior to this change, the example documentation
found in the README.md has an errant code which
won't work in the table-driven code example.

This change modifies the variable name from `t` to `tc`
so it does not conflict with the `t *testing.T` struct
definition.

* Adds a range clause to the `for` statement
* Modifies `for` statement scope to use `tc.shouldPass`, and `tc.routeVariable`

Doc: https://github.com/gorilla/mux#testing-handlers

CORSMethodMiddleware sets the Access-Control-Allow-Methods response header
on a request, by matching routes based only on paths. It also handles
OPTIONS requests, by settings Access-Control-Allow-Methods, and then
returning without calling the next HTTP handler.

The example in the README does not pass the request through a mux therefore the request variables from the path are never populated. Update the sample to create a minimum viable router to use.

Fixes #373

Fixes #394

* Drop support for Go < 1.7: remove gorilla/context
* Remove Go < 1.7 from Travis CI config
* Remove unneeded _native from context files

Only run vet on the latest Go version.

* Pull out common shared `routeConf` so that config is pushed on to child
routers and routes.
* Removes obsolete usages of `parentRoute`
* Add tests defining compositional behavior
* Exercise `copyRouteConf` for posterity

* remove redundant code that remove support gorilla/context

* backward compatible for remove redundant code

Because domain.com is an actual business, example.com should be used for example purposes.

Previously, when searching for a match, matchErr would be erroneously set, and prevent middleware from running (no match == no middleware runs).

This fix clears matchErr before traversing the next subrouter in a multi-subrouter router.

Previously, getHost only returned the host. As it now returns the
port as well, any .Host matches on a route will need to be updated
to also support matching on the port for cases where the port is
non default, eg: 80 for http or 443 for https.

MatchErr is set by the router to ErrNotFound if no route matches. If
no route of a Subrouter matches the error can by safely ignored. This
implementation only ignores these errors and does not ignore other
errors like ErrMethodMismatch.

From the docs: Changing the header map after a call to WriteHeader (or
Write) has no effect unless the modified headers are
trailers.

In lieu of checking the template pattern on every Match request, a bool is added to the routeRegexp, and set
if the routeRegexp is a host AND there is no ":" in the template. I dislike extending the type, but I'd dislike
doing a string match on every single Match, even more.

* [ci] Create CircleCI config
* Fix typos in container versions
* Add CircleCI badge

* Use subtests for middleware tests
* Don't use subtests for MiddlewareAdd

* More sensical CORSMethodMiddleware

* Only sets Access-Control-Allow-Methods on valid preflight requests
* Does not return after setting the Access-Control-Allow-Methods header
* Does not append OPTIONS header to Access-Control-Allow-Methods
regardless of whether there is an OPTIONS method matcher
* Adds tests for the listed behavior

* Add example for CORSMethodMiddleware

* Do not check for preflight and add documentation to the README

* Use http.MethodOptions instead of "OPTIONS"

* Add link to CORSMethodMiddleware section to readme

* Add test for unmatching route methods

* Rename CORS Method Middleware to Handling CORS Requests in README

* Link CORSMethodMiddleware in README to godoc

* Break CORSMethodMiddleware doc into bullets for readability

* Add comment about specifying OPTIONS to example in README for CORSMethodMiddleware

* Document cURL command used for testing CORS Method Middleware

* Update comment in example to "Handle the request"

* Add explicit comment about OPTIONS matchers to CORSMethodMiddleware doc

* Update circleci config to only check gofmt diff on latest go version

* Break up gofmt and go vet checks into separate steps.

* Use canonical circleci config

* Update config.yml

* Update config.yml

No need to convert here.

Use a single append call instead of a ranged for loop.

* Add documentation for using mux to serve a SPA

* r -> router to prevent shadowing

* Expand SPA acronym

* BrowserRouter link

* Add more comments to explain how the spaHandler.ServeHTTP method works

#510)

…age (#514)

* Guess the scheme if r.URL.Scheme is unset
It's not expected that the request's URL is fully populated when used on
the server-side (it's more of a client-side field), so we shouldn't
expect it to be present.

In practice, it's only rarely set at all on the server, making mux's
`Schemes` matcher tricky to use as it is.

This commit adds a test which would have failed before demonstrating the
problem, as well as a fix which I think makes `.Schemes` match what
users expect.

* [doc] Add more detail to Schemes and URL godocs

* Add route url test for schemes

* Make httpserver test use more specific scheme matchers

* Update test to have different responses per route

* Remove context helpers in context.go
* Update request context funcs to take concrete types
* Move TestNativeContextMiddleware to mux_test.go
* Clarify KeepContext Go 1.7+ comment

Mux doesn't build on Go < 1.7 so the comment doesn't really need to
clarify anymore.

* Adds a test case for the repro given in issue #534
* Fixes the logic in CORSMethodMiddleware to handle matching routes
better

Fix the CORSMethodMiddleware bug with subrouters

A production server is seeing a significant amount of allocations in (*routeRegexp).getURLQuery

Since it is only interested in a single value and only the first value we create a specialized function for that.

Comparing a few parameter parsing scenarios:

```
Benchmark_findQueryKey/0-8 	 7184014	       168 ns/op	       0 B/op	       0 allocs/op
Benchmark_findQueryKey/1-8 	 5307873	       227 ns/op	      48 B/op	       3 allocs/op
Benchmark_findQueryKey/2-8 	 1560836	       770 ns/op	     483 B/op	      10 allocs/op
Benchmark_findQueryKey/3-8 	 1296200	       931 ns/op	     559 B/op	      11 allocs/op
Benchmark_findQueryKey/4-8 	  666502	      1769 ns/op	       3 B/op	       1 allocs/op

Benchmark_findQueryKeyGoLib/0-8 	 1740973	       690 ns/op	     864 B/op	       8 allocs/op
Benchmark_findQueryKeyGoLib/1-8 	 3029618	       393 ns/op	     432 B/op	       4 allocs/op
Benchmark_findQueryKeyGoLib/2-8 	  461427	      2511 ns/op	    1542 B/op	      24 allocs/op
Benchmark_findQueryKeyGoLib/3-8 	  324252	      3804 ns/op	    1984 B/op	      28 allocs/op
Benchmark_findQueryKeyGoLib/4-8 	   69348	     14928 ns/op	   12716 B/op	     130 allocs/op
```

Comment for CurrentRoute claimed that setting the KeepContext option would propagate
the Route even after the request. The KeepContext option is deprecated and has no effect.

Continuing from PR #447 we have to add extra
check to ignore the port as well

add tests to cover this case

* build: CircleCI 2.1 + build matrix

* build: drop Go 1.9 & Go 1.10

* build: remove erroneous version

* Update README.md

* build: gofmt on latest

* regexp: use iota instead of hardcoded values

…ile (#693)

* initial commit

* fix `README.md` - authenticationMiddleware initialization