opentracing Extract函数分析

李恒道

看的是MockTracer的代码
在https://github.com/opentracing/opentracing-go/blob/10b1cf09e00bdc84234b8c7a4b4d4e4afe64de87/mocktracer/mocktracer.go#L110

首先根据format找到匹配的解压缩对象，调用Extract

// Extract belongs to the Tracer interface.
func (t *MockTracer) Extract(format interface{}, carrier interface{}) (opentracing.SpanContext, error) {
    extractor, ok := t.extractors[format]
    if !ok {
        return nil, opentracing.ErrUnsupportedFormat
    }
    return extractor.Extract(carrier)
}

我们看new生成部分

这里看到注册了两个Extractor,分别是opentracing.TextMap和opentracing.HTTPHeaders

// New returns a MockTracer opentracing.Tracer implementation that's intended
// to facilitate tests of OpenTracing instrumentation.
func New() *MockTracer {
    t := &MockTracer{
        finishedSpans: []*MockSpan{},
        startedSpans:  []*MockSpan{},
        injectors:     make(map[interface{}]Injector),
        extractors:    make(map[interface{}]Extractor),
    }

    // register default injectors/extractors
    textPropagator := new(TextMapPropagator)
    t.RegisterInjector(opentracing.TextMap, textPropagator)
    t.RegisterExtractor(opentracing.TextMap, textPropagator)

    httpPropagator := &TextMapPropagator{HTTPHeaders: true}
    t.RegisterInjector(opentracing.HTTPHeaders, httpPropagator)
    t.RegisterExtractor(opentracing.HTTPHeaders, httpPropagator)

    return t
}

而注册的代码是

可以看到是传入的第二个参数起到了作用

// RegisterExtractor registers extractor for given format
func (t *MockTracer) RegisterExtractor(format interface{}, extractor Extractor) {
    t.extractors[format] = extractor
}

所以我们应该看textPropagator或者httpPropagator，这里我们选择查看textPropagator

textPropagator := new(TextMapPropagator)

这里可以看到new了一个TextMapPropagator，我们继续往下翻

可以看到Extract代码主要是将carrier断言为opentracing.TextMapReader，然后调用ForeachKey函数进行循环

func (t *TextMapPropagator) Extract(carrier interface{}) (MockSpanContext, error) {
    reader, ok := carrier.(opentracing.TextMapReader)
    if !ok {
        return emptyContext, opentracing.ErrInvalidCarrier
    }
    rval := MockSpanContext{0, 0, true, nil}
    err := reader.ForeachKey(func(key, val string) error {
        lowerKey := strings.ToLower(key)
        switch {
        case lowerKey == mockTextMapIdsPrefix+"traceid":
            // Ids:
            i, err := strconv.Atoi(val)
            if err != nil {
                return err
            }
            rval.TraceID = i
        case lowerKey == mockTextMapIdsPrefix+"spanid":
            // Ids:
            i, err := strconv.Atoi(val)
            if err != nil {
                return err
            }
            rval.SpanID = i
        case lowerKey == mockTextMapIdsPrefix+"sampled":
            b, err := strconv.ParseBool(val)
            if err != nil {
                return err
            }
            rval.Sampled = b
        case strings.HasPrefix(lowerKey, mockTextMapBaggagePrefix):
            // Baggage:
            if rval.Baggage == nil {
                rval.Baggage = make(map[string]string)
            }
            safeVal := val
            if t.HTTPHeaders {
                // unescape errors are ignored, nothing can be done
                if rawVal, err := url.QueryUnescape(val); err == nil {
                    safeVal = rawVal
                }
            }
            rval.Baggage[lowerKey[len(mockTextMapBaggagePrefix):]] = safeVal
        }
        return nil
    })
    if rval.TraceID == 0 || rval.SpanID == 0 {
        return emptyContext, opentracing.ErrSpanContextNotFound
    }
    if err != nil {
        return emptyContext, err
    }
    return rval, nil
}

那么也就是说任何符合TextMapReader接口的都可以传入

type TextMapReader interface {
    ForeachKey(handler func(key, val string) error) error
}

只要创造一个结构符合该接口即可调用Extract函数