优化

internal/service/introdb_test.go

@@ -1,9 +1,12 @@
 package service
 
 import (
+	"context"
 	"net/http"
 	"net/http/httptest"
+	"sync/atomic"
 	"testing"
+	"time"
 
 	"go.uber.org/zap"
 )
@@ -160,3 +163,142 @@ func TestIntroDBFetchParsesBody(t *testing.T) {
 		t.Fatalf("spans = %#v", spans)
 	}
 }
+
+// 实测：从生产机连发 45 个请求有 15 个被返回 429，加间隔重试后其中 10 个成功。
+// 所以限流值得一次重试，否则那部分播放会静默少掉「跳过片头」按钮。
+func TestIntroDBFetchRetriesRateLimit(t *testing.T) {
+	var calls int32
+	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, _ *http.Request) {
+		if atomic.AddInt32(&calls, 1) == 1 {
+			w.WriteHeader(http.StatusTooManyRequests)
+			return
+		}
+		w.Header().Set("Content-Type", "application/json")
+		_, _ = w.Write([]byte(introDBTVPayload))
+	}))
+	defer server.Close()
+
+	svc := NewIntroDBService(zap.NewNop()).SetBaseURL(server.URL).SetRetryDelay(0)
+	spans, err := svc.Fetch(t.Context(), 1396, 1, 1)
+	if err != nil {
+		t.Fatalf("fetch: %v", err)
+	}
+	if len(spans) != 2 {
+		t.Fatalf("spans = %#v, want the retried response", spans)
+	}
+	if got := atomic.LoadInt32(&calls); got != 2 {
+		t.Fatalf("calls = %d, want 2 (the original plus one retry)", got)
+	}
+}
+
+func TestIntroDBFetchGivesUpAfterRetryBudget(t *testing.T) {
+	var calls int32
+	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, _ *http.Request) {
+		atomic.AddInt32(&calls, 1)
+		w.WriteHeader(http.StatusTooManyRequests)
+	}))
+	defer server.Close()
+
+	svc := NewIntroDBService(zap.NewNop()).SetBaseURL(server.URL).SetRetryDelay(0)
+	if _, err := svc.Fetch(t.Context(), 1396, 1, 1); err == nil {
+		t.Fatal("a persistent 429 must surface as an error so the caller keeps its cache")
+	}
+	// 只重试一次：限流通常不是靠密集重试解决的，而调用方的等待预算有限。
+	if got := atomic.LoadInt32(&calls); got != 2 {
+		t.Fatalf("calls = %d, want 2 (one retry, then give up)", got)
+	}
+}
+
+func TestIntroDBFetchDoesNotRetryNotFound(t *testing.T) {
+	var calls int32
+	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, _ *http.Request) {
+		atomic.AddInt32(&calls, 1)
+		w.WriteHeader(http.StatusNotFound)
+	}))
+	defer server.Close()
+
+	svc := NewIntroDBService(zap.NewNop()).SetBaseURL(server.URL).SetRetryDelay(0)
+	spans, err := svc.Fetch(t.Context(), 424242, 0, 0)
+	if err != nil || len(spans) != 0 {
+		t.Fatalf("spans = %#v err = %v, want a cached miss", spans, err)
+	}
+	if got := atomic.LoadInt32(&calls); got != 1 {
+		t.Fatalf("calls = %d, want 1: 404 means \"no data\", it is not worth retrying", got)
+	}
+}
+
+// 调用方预算不够时不能为了重试干等：Emby 只给 5 秒，等下去会把
+// 「少一个跳过按钮」升级成「请求超时」。
+func TestIntroDBFetchSkipsRetryWhenCallerBudgetIsSpent(t *testing.T) {
+	var calls int32
+	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, _ *http.Request) {
+		atomic.AddInt32(&calls, 1)
+		w.WriteHeader(http.StatusTooManyRequests)
+	}))
+	defer server.Close()
+
+	svc := NewIntroDBService(zap.NewNop()).SetBaseURL(server.URL).SetRetryDelay(time.Minute)
+	ctx, cancel := context.WithTimeout(t.Context(), 50*time.Millisecond)
+	defer cancel()
+
+	started := time.Now()
+	if _, err := svc.Fetch(ctx, 1396, 1, 1); err == nil {
+		t.Fatal("want an error when the budget is spent")
+	}
+	if elapsed := time.Since(started); elapsed > 2*time.Second {
+		t.Fatalf("Fetch waited %s, want it to give up promptly", elapsed)
+	}
+	if got := atomic.LoadInt32(&calls); got != 1 {
+		t.Fatalf("calls = %d, want 1 without a retry", got)
+	}
+}
+
+func TestIntroDBRetryWaitPrefersRetryAfterHeader(t *testing.T) {
+	svc := NewIntroDBService(zap.NewNop())
+	if got := svc.retryWait(2 * time.Second); got != 2*time.Second {
+		t.Fatalf("retryWait(2s) = %s, want the server's value", got)
+	}
+	if got := svc.retryWait(0); got != introDBRetryDelay {
+		t.Fatalf("retryWait(0) = %s, want the default %s", got, introDBRetryDelay)
+	}
+	// 服务端可以要求等很久，但一次播放不值得为它挂住几十秒。
+	if got := svc.retryWait(10 * time.Minute); got != introDBMaxRetryDelay {
+		t.Fatalf("retryWait(10m) = %s, want it capped at %s", got, introDBMaxRetryDelay)
+	}
+}
+
+func TestParseIntroDBRetryAfter(t *testing.T) {
+	cases := []struct {
+		value string
+		want  time.Duration
+	}{
+		{"2", 2 * time.Second},
+		{" 3 ", 3 * time.Second},
+		{"", 0},
+		{"abc", 0},
+		{"-5", 0},
+		{"0", 0},
+	}
+	for _, tc := range cases {
+		if got := parseIntroDBRetryAfter(tc.value); got != tc.want {
+			t.Fatalf("parseIntroDBRetryAfter(%q) = %s, want %s", tc.value, got, tc.want)
+		}
+	}
+}
+
+func TestIntroDBRetryableStatus(t *testing.T) {
+	retryable := []int{http.StatusTooManyRequests, http.StatusServiceUnavailable}
+	for _, status := range retryable {
+		if !introDBRetryableStatus(status) {
+			t.Fatalf("status %d should be retryable", status)
+		}
+	}
+	// 404 由 fetchOnce 单独处理；500 之类的服务端故障重试也不会变好，
+	// 却会白占调用方的等待预算。
+	notRetryable := []int{http.StatusNotFound, http.StatusInternalServerError, http.StatusBadRequest}
+	for _, status := range notRetryable {
+		if introDBRetryableStatus(status) {
+			t.Fatalf("status %d should not be retryable", status)
+		}
+	}
+}