I want to expose the following URLs from my service:
GET /api/foo
GET /api/bar
I also want to structure it as a router nested inside another. The toplevel router will match all requests to /api and serve them with the nested router which will match requests to /foo and /bar. Basically, namespacing.
I could just have a single router and give the /api prefix to both the routes:
router.GET("/api/foo", apiFoo)
router.GET("/api/bar", apiBar)
But I would like the convenience of having a single router for all routes inside the /api prefix so that I can add appropriate middlewares to them all with a single function call.
Here's what I tried:
package main
import (
"log"
"net/http"
"github.com/julienschmidt/httprouter"
)
func apiFoo(w http.ResponseWriter, r *http.Request) {}
func apiBar(w http.ResponseWriter, r *http.Request) {}
func main() {
api := httprouter.New()
api.HandlerFunc(http.MethodGet, "/foo", apiFoo)
api.HandlerFunc(http.MethodGet, "/bar", apiBar)
router := httprouter.New()
router.Handler(http.MethodGet, "/api", api)
log.Fatal(http.ListenAndServe(":8080", router))
}
However, getting 404 not found on going to http://localhost:8080/api/foo or http://localhost:8080/api/bar
I had thought that nested routers would work because routers implement the http.Handler interface. What am I missing?
httprouter does not concatenate the path, so you can't do it that way. Both routers will just inspect the request path and act accordingly.
There is an open pull request for 7 years, that would implement it. You could have a look there and implement similar logic yourself, the PR is based on concatenating the path. Maybe you can write a small helper function for that.
If you are willing to switch the router package, you could look into alternatives such as chi, which support router groups.
Related
What is the difference between chi.Use and chi.With when setting up a middleware with Chi router.
Use must be declared before all routes under the same group, whereas r.With allows you to "inline" middlewares.
As a matter of fact, the function signatures are different. Use returns nothing, With returns a chi.Router.
Let's say you have a route and want to add a middleware only to one of them, you would use r.With:
r.Route("/myroute", func(r chi.Router) {
r.Use(someMiddleware) // can declare it here
r.Get("/bar", handlerBar)
r.Put("/baz", handlerBaz)
// r.Use(someMiddleware) // can NOT declare it here
}
r.Route("/other-route", func(r chi.Router) {
r.Get("/alpha", handlerBar)
r.Put("/beta", handlerBaz)
r.With(someMiddleware).Get("/gamma", handlerQuux)
}
In the first example, someMiddleware is declared for all sub-routes, whereas in the second example r.With allows you to add a middleware only for the /other-route/gamma route.
According to the documentation of chi.Use and chi.With.
Use appends a middleware handler to the Mux middleware stack.
The middleware stack for any Mux will execute before searching for a matching route to a specific handler, which provides opportunity to respond early, change the course of the request execution, or set request-scoped values for the next http.Handler.
With adds inline middlewares for an endpoint handler.
Let see how chi.Use and chi.With example
The use case is pretty straight forward with chi.Use the registered middleware will run before all the routes handler which are register with the Router
r := chi.NewRouter()
r.Use(middleware.Logger)
r.Get("/", func(w http.ResponseWriter, r *http.Request) {
w.Write([]byte("welcome"))
})
http.ListenAndServe(":3000", r)
For eg: here Logger middleware will be called before all the register routes handler.
whereas with chi.With you are returned new route on which the middleware would be ran so if any routes is registered on the returned Router the registered middleware will run. Here the use case is very specific suppose if you want to run a specific middleware for a group of routes or want to perform some operation for specific routes then for the case you can use chi.Use
r.Route("/articles", func(r chi.Router) {
r.With(paginate).Get("/", listArticles) // GET /articles
r.With(paginate).Get("/{month}-{day}-{year}", listArticlesByDate) // GET /articles/01-16-2017
r.Post("/", createArticle) // POST /articles
r.Get("/search", searchArticles) // GET /articles/search
// Regexp url parameters:
r.Get("/{articleSlug:[a-z-]+}", getArticleBySlug) // GET /articles/home-is-toronto
// Subrouters:
r.Route("/{articleID}", func(r chi.Router) {
r.Use(ArticleCtx)
r.Get("/", getArticle) // GET /articles/123
r.Put("/", updateArticle) // PUT /articles/123
r.Delete("/", deleteArticle) // DELETE /articles/123
})
})
In the above example the paginate middleware will only be called for all the articles with /articles/ and /{month}-{day}-{year} day wise route for other routes chi.With won't be called if there any middlware registered with chi.Use over main route then that would be called.
This might just be because of my inexperience, but I've taken over a webserver solution written in Go and I have some issues.
The routing is set up in such a way that each "top" route is mounted to the router with router.Mount() with a handler attached to each mounted route. Example:
router.Mount("/group", (handler.NewGroupHandler(groupSrv, render)).Router())
So, within this mounted route I want to add a middleware to get a specific URL parameter. I've tried around, and followed this guide https://medium.com/#szablowska.patrycja/chi-and-missing-urlparam-in-middleware-9435c48a063b which seems to have worked for other. However, this haven't worked for me, which I now suspect is because of the approach with mounting.
The idea of the middleware is quite simple, just check the URL param against some requirements. I could of course do this manually in every path, but a middleware would be better both for readability and future development. How the middleware is written:
func GroupMiddleware() func(next http.Handler) http.Handler {
return func(next http.Handler) http.Handler {
fn := func(w http.ResponseWriter, r *http.Request) {
groupIDStr := chi.URLParam(r, "group_id")
if checkAgainstGroupID {
// Do something, raise error
} else {
next.ServeHTTP(w, r)
}
}
return http.HandlerFunc(fn)
}
}
And then applied like this:
r.With(subhandler.GroupMiddleware).Route("/{group_id}", func(r chi.Router) {
r.Get("/", h.get)
})
Which obviously doesn't work. Any help would be appreciated because I can't believe that mounting routes makes it impossible to add middleware to subgroups.
Edit: This worked perfectly, it was just me that didn't register that I called an unaffected route. So in case someone else should end up with this same question, know that this actually is a valid approach.
How would one go about adding a method to a struct that is a different file?
This is what I've tried so far but it doesn't seem to be working.
// ./src
package routes
type App struct{
}
func (a *App) initializeRoutes() {
a.Subrouter.HandleFunc("/products", a.getSomething).Methods("GET")
}
// ./src/controller
package routes
func (a *App) getSomething(w http.ResponseWriter, r *http.Request){
...
}
They are in the same package.
They are not in the same package. A Go package has both a name and a path. They're both named routes but they have different paths. The actual packages are routes and controller/routes. The result is subdirectories are different packages.
See Package Names on the Go Blog for more information.
Since they're in different packages, they can only access each other's public members and exported methods. You can't monkey patch someone else's package or interface in Go. This is by design to keep all the functionality of a package in one place, no action-at-a-distance.
You have options. You could put all methods of routes into a single package. If they all belong together, there's no need to split it up into multiple files.
If they don't really belong together, you can write a new struct with routes embedded into it and define new methods on that. Then you can access either the wrapper struct to get your added methods, or its embedded struct to get routes' methods. See this answer for an example.
But really I think you need to think about how your code is arranged. The App probably shouldn't be defined by the routes package, they should be separate. Instead, Go prefers a has-a relationship. App would contain an instance of routes.Route.
You'd rearrange your code tree like so:
app/
app.go
app_test.go
routes/
routes.go
routes_test.go
Note that rather than having it all in src/ it's now contained in its own project directory. app.go would look something like this.
// src/app/app.go
package app
import(
"app/routes"
"fmt"
"net/http"
)
type App struct {
routes routes.Route
}
func (a *App) initializeRoutes() {
a.routes.AddRoute("/products", a.getSomething)
}
func (a *App) getSomething(w http.ResponseWriter, r *http.Request) {
fmt.Println("Something!")
}
Notice how we're delegating responsibility to adding a route to a.routes rather than having App do it itself. This avoids the desire to smash all functionality into one ginormous package. routes.Route would be defined in app/routes/routes.go.
// src/app/routes/routes.go
package routes
import "net/http"
// A type specifying the interface for route handlers.
type RouteHandler func(w http.ResponseWriter, r *http.Request)
type Route struct {
handlers map[string]RouteHandler
}
func (r *Route) AddRoute(path string, handler RouteHandler) {
r.handlers[path] = handler
}
Now all routes has to worry about is handling routes. It doesn't know anything about your specific application logic.
I was trying to get my http.res & http.req functions in a controllers file.
Now that we've rearranged the file structure, you can do that. You can, if you like, define app/controllers.go to organize your code.
// src/app/controllers.go
package app
import(
"fmt"
"net/http"
)
func (a *App) getSomething(w http.ResponseWriter, r *http.Request) {
fmt.Println("Something!")
}
app/app.go and app/controllers.go are in the same package. They have the same path and the same name. So app/controllers.go can add methods to App.
You have an error because your files belong to different packages. Everything related to one struct must be in the same package.
It is possible to declare struct and its methods in different files, but they must belong same package (be in same folder).
I'm using gorilla/context in a web app. The example in the docs looks like:
func MyHandler(w http.ResponseWriter, r *http.Request) {
//...
val := context.Get(r, foo.MyKey)
//...
}
How can I unit test a handler that works like this? The only way I've managed so far is to use the context package inside my tests. I'm thinking at the moment that I could inject a context parameter into the handler but then I'm not conforming to the HandlerFunc interface.
This is a classic cross-cutting concerns example.
You are using a 3rd party to magically handle input params for your unit under test. By that very definition, you are going to have to do some extra setup to prep the context for the state you want.
When it comes to Go http handlers (which the convention is to KISS it), you shouldn't need to "reach out of context" of your func to get extra data: keep all the data you need within your single func.
Personally, I try to avoid corrupting my handlers like this. I think I've only used gorilla's context once out of the dozens of big sites I've built. And that was basically to get around a cached response, just to keep the data refesh to the end user. Of which I simply ignored in my unit tests, as it was out-of-scrope of what I was testing.
Instead, I use middle-ware wrappers to setup the basics I want in my handles and modify the handler's signature accordingly.
caching
logging
authentication and authorization
json marshaling
context (e.g. expected User{} object loaded from DB)
...etc. I would create a middle-ware that wraps your handler when you register it with mux that uses gorilla/context to lookup your cookie or userid or something, hidrates that user object from cache, DB, redis, etc, and then calls your handler that would have a signature like:
func MyHandler(u User, p Page, w http.ResponseWriter, r *http.Request) {
// u is the User object loaded from middle-ware
// p is your global Page object... etc
}
That way, your unit tests inject only the items you need under test.
And you can integration test your middle-ware on a QA server with expected User and Page objects in a datastore.
The way my team does it is to add a name to the route handler and then in the tests we call that route by name.
This is how to add a route:
r.HandleFunc("/<route>", MyHandler).Methods("GET").Name("MyHandlerByName")
Then this is how to test it
r.Get("MyHandlerByName")
One way to test handles is modify the way in which they are created. for example, Creating a function that return a http.HandlerFunc, this function can have parameters. You can mock the values that you send to the function
Without parameters
func State() http.HandlerFunc {
return http.HandlerFunc(func(pResponse http.ResponseWriter, r *http.Request) {
// your code
})
}
With Parameters
func State(pParam1,pParam2,pParam3 ...) http.HandlerFunc {
return http.HandlerFunc(func(pResponse http.ResponseWriter, r *http.Request) {
// your code using pParam1,pParam2,pParam3
})
}
The mapping will be
http.HandleFunc("/State", State())
or
http.HandleFunc("/State", State(value1,value2,value3 ....))
I want to have different middleware for different path. My current implementation is from this link
UserRouter := mux.NewRouter().StrictSlash(true)
AdminRouter := mux.NewRouter().StrictSlash(true)
Router.HandleFunc("/apps/{app_name}/xyz", Handler).Methods("GET")
I created three different routers, so that I can assosiate them with different path and middleware
nUserPath := negroni.New(middleware.NewAuthMiddleWare())
nUserPath.UseHandler(UserRouter)
nAdminPath := negroni.New()
nAdminPath.UseHandler(AdminRouter)
I created two different negroni instances and passed them the respective routers. As I wanted all this to run part of the same application on the same port so I created a Wrapper Router and negroni instance and associated them with the existing like below
BaseRouter := mux.NewRouter().StrictSlash(true)
BaseRouter.Handle(UserBasePath,nUserPath) // UserBasePath is `/apps`
BaseRouter.Handle(HealthCheck,nUserPath) // HealthCheck is `/health`
BaseRouter.Handle(AdminBasePath,nAdminPath) // AdminBasePath is `/Admin`
n := negroni.New(middleware.NewLogger()) // attached other common middleware here
n.UseHandler(router.BaseRouter)
n.Run(":8080")
Issues faced in this approach:
When I run /health it runs properly but when I run /apps/{app_name}/something I get a 404: Not Found
Note : I went through other approaches mentioned in below link but they don't satisfy my need.
- Route-specific Middlewares with Negroni
So, the issue with the above implementation is that BaseRouter.Handle() method take a path and not a path_matcher/template so all the url's which has path_length more than one were not working.
I figured out two ways to achieve what I needed:
First approach
// Create a rootRouter
var rootRouter *mux.Router = mux.NewRouter()
// Create as many subRouter you want with some prefix
var appsBasePath string = "/apps"
var adminBasePath string = "/admin"
upRouter := rootRouter.PathPrefix(appsBasePath).Subrouter()
apRouter := rootRouter.PathPrefix(adminBasePath).Subrouter()
// Register all the paths and mention middleware specifically for all of them
// Here middleware is a method with signature as
// func middleware( http.Handler) http.HandlerFunc {}
upRouter.Path("/test").Methods("POST").Handler(middleware(http.HandlerFunc(func(w http.ResponseWriter, req *http.Request){
fmt.Fprintf(w, "Welcome to the home page!")
})))
n := negroni.New(middleware.NewLogger()) // attached other common middleware here
n.UseHandler(rootRouter)
n.Run(":8080")
Second approach
This is extension/solution of the original issue in the question
// Replace BaseRouter.handle() as below
// as PathPrefix takes a template so it won't have issue that we were facing
BaseRouter.PathPrefix(UserBasePath).Handler(nUserPath)
Thing to remember here is that within negroni nUserPath RequestContext of the middleware attached will be different from that of the actual router's HandlerMethod
Note:
By path length I mean something like this -- /abc or /abc/ has path_length=1 and /abc/xyz has path_length=2