gitee.com/ks-custle/core-gm@v0.0.0-20230922171213-b83bdd97b62c/httpsnoop/README.md (about) 1 # httpsnoop 2 3 Package httpsnoop provides an easy way to capture http related metrics (i.e. 4 response time, bytes written, and http status code) from your application's 5 http.Handlers. 6 7 Doing this requires non-trivial wrapping of the http.ResponseWriter interface, 8 which is also exposed for users interested in a more low-level API. 9 10 [![GoDoc](https://godoc.org/github.com/felixge/httpsnoop?status.svg)](https://godoc.org/github.com/felixge/httpsnoop) 11 [![Build Status](https://travis-ci.org/felixge/httpsnoop.svg?branch=master)](https://travis-ci.org/felixge/httpsnoop) 12 13 ## Usage Example 14 15 ```go 16 // myH is your app's http handler, perhaps a http.ServeMux or similar. 17 var myH http.Handler 18 // wrappedH wraps myH in order to log every request. 19 wrappedH := http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) { 20 m := httpsnoop.CaptureMetrics(myH, w, r) 21 log.Printf( 22 "%s %s (code=%d dt=%s written=%d)", 23 r.Method, 24 r.URL, 25 m.Code, 26 m.Duration, 27 m.Written, 28 ) 29 }) 30 http.ListenAndServe(":8080", wrappedH) 31 ``` 32 33 ## Why this package exists 34 35 Instrumenting an application's http.Handler is surprisingly difficult. 36 37 However if you google for e.g. "capture ResponseWriter status code" you'll find 38 lots of advise and code examples that suggest it to be a fairly trivial 39 undertaking. Unfortunately everything I've seen so far has a high chance of 40 breaking your application. 41 42 The main problem is that a `http.ResponseWriter` often implements additional 43 interfaces such as `http.Flusher`, `http.CloseNotifier`, `http.Hijacker`, `http.Pusher`, and 44 `io.ReaderFrom`. So the naive approach of just wrapping `http.ResponseWriter` 45 in your own struct that also implements the `http.ResponseWriter` interface 46 will hide the additional interfaces mentioned above. This has a high change of 47 introducing subtle bugs into any non-trivial application. 48 49 Another approach I've seen people take is to return a struct that implements 50 all of the interfaces above. However, that's also problematic, because it's 51 difficult to fake some of these interfaces behaviors when the underlying 52 `http.ResponseWriter` doesn't have an implementation. It's also dangerous, 53 because an application may choose to operate differently, merely because it 54 detects the presence of these additional interfaces. 55 56 This package solves this problem by checking which additional interfaces a 57 `http.ResponseWriter` implements, returning a wrapped version implementing the 58 exact same set of interfaces. 59 60 Additionally this package properly handles edge cases such as `WriteHeader` not 61 being called, or called more than once, as well as concurrent calls to 62 `http.ResponseWriter` methods, and even calls happening after the wrapped 63 `ServeHTTP` has already returned. 64 65 Unfortunately this package is not perfect either. It's possible that it is 66 still missing some interfaces provided by the go core (let me know if you find 67 one), and it won't work for applications adding their own interfaces into the 68 mix. You can however use `httpsnoop.Unwrap(w)` to access the underlying 69 `http.ResponseWriter` and type-assert the result to its other interfaces. 70 71 However, hopefully the explanation above has sufficiently scared you of rolling 72 your own solution to this problem. httpsnoop may still break your application, 73 but at least it tries to avoid it as much as possible. 74 75 Anyway, the real problem here is that smuggling additional interfaces inside 76 `http.ResponseWriter` is a problematic design choice, but it probably goes as 77 deep as the Go language specification itself. But that's okay, I still prefer 78 Go over the alternatives ;). 79 80 ## Performance 81 82 ``` 83 BenchmarkBaseline-8 20000 94912 ns/op 84 BenchmarkCaptureMetrics-8 20000 95461 ns/op 85 ``` 86 87 As you can see, using `CaptureMetrics` on a vanilla http.Handler introduces an 88 overhead of ~500 ns per http request on my machine. However, the margin of 89 error appears to be larger than that, therefor it should be reasonable to 90 assume that the overhead introduced by `CaptureMetrics` is absolutely 91 negligible. 92 93 ## License 94 95 MIT