Add propagator interface and W3C propagator #85

rghetia · 2019-08-09T21:02:35Z

Partially fixes #8

Binary and B3 propagation will be in another PR..

rghetia · 2019-08-12T22:44:50Z

There is a discussion about moving propagator out of tracer() in the spec
open-telemetry/opentelemetry-specification#112

This PR partially moves propagator to propagation package but Injector/Extractor are still part of tracer package. For example, HTTPPropagator interface is part of propagation package. It uses Injector/Extractor interface defined in trace package but there is not need for those to be in trace pacakge. Injector/Extractor should be moved to propagation package as well.

This PR should be refactored according to the outcome of the spec issue.

rghetia · 2019-08-23T20:51:28Z

since open-telemetry/opentelemetry-specification#112 is fixed by open-telemetry/opentelemetry-specification#209, this PR is ready for review and merge.

krnowak · 2019-08-26T18:04:43Z

propagation/http_trace_context_propagator.go

+
+var _ apipropagation.TextFormatPropagator = httpTraceContextPropagator{}
+
+// CarrierExtractor implements CarrierExtractor method of TextFormatPropagator interface.


That's a bit stuttery. What do you think about CarrierExtractor is a part of an implementation of the TextFormatPropagator interface.?

Docs of other functions sound like this too, so if you think my proposal is fine, please update the other docs too.

I changed it. Will upload the diff soon. LMK if it is ok.

krnowak · 2019-08-26T18:09:57Z

propagation/http_trace_context_propagator.go

+	if ok {
+		return traceContextExtractor{req: req}
+	}
+	return traceContextExtractor{}


Maybe the propagation package, or API propagation package could provide a noop implementations of the interfaces? Then you could use it here and assume that traceContextExtractor's req field can't be nil?

Added noop propagator.

krnowak · 2019-08-26T18:21:15Z

propagation/http_trace_context_propagator.go

+// W3C TraceContext Header and decodes SpanContext from the Header.
+func (tce traceContextExtractor) Extract() (sc core.SpanContext, tm tag.Map) {
+	if tce.req == nil {
+		return core.EmptySpanContext(), tag.NewEmptyMap()


That's a lot of core.EmptySpanContext(), tag.NewEmptyMap() repetition. Maybe add some empty function like func noExtract() (sc core.SpanContext, m tag.Map) {return} (or func noExtract() (core.SpanContext, tag.Map) {return core.EmptySpanContext(), tag.NewEmptyMap()}) and use that?

added noExtract().

krnowak · 2019-08-26T18:35:20Z

api/propagation/propagator.go

+	"go.opentelemetry.io/api/tag"
+)
+
+// TextFormatPropagator is an interface that specifies methods to create CarrierInjector


I'm not sure what this text format pertains to. I see text and binary propagators mentioned in otel spec PR 209, but without any details. The spec of propagators feels to be specified in a way that this interface seems to satisfy the needs of the binary propagator too? If the "text format" is about the contents of the tag.Map, then probably the only thing that needs to be a text are keys and maybe some values. If this is about the way extractors and injectors work (wire format?), I don't see anything limiting them to be textual. So would a name like Propagator be fine too?

There are two primary encoding for SpanContext and tag.Map (Distributed Context)

Binary

TextFormat
TextFormat has different format, like W3C TraceContext and B3.

These formats have different interface requirement hence they are separate.

tedsuo

Sweet! Thanks for getting this latest update out the door @rghetia.

jmacd · 2019-09-05T21:19:26Z

api/trace/pass_through_span.go

+	"go.opentelemetry.io/api/tag"
+)
+
+type PassThroughSpan struct {


I'm questioning whether we need this PassThroughSpan and PassThroughTracer.

My mental model for extracting a context to start a child looks like:

childSpanContext := propagate.Extract(dataFromWire) ctx, childSpan := tracer.Start("childName", trace.ChildOf(childSpanContext))

I guess the key idea here is that we don't need to install the extracted context as the current one in any sense, we only need to pass it to the Start call as an option to signify an extracted parent context.

I'm questioning whether we need this PassThroughSpan and PassThroughTracer.

My mental model for extracting a context to start a child looks like:

childSpanContext := propagate.Extract(dataFromWire) ctx, childSpan := tracer.Start("childName", trace.ChildOf(childSpanContext))

I guess the key idea here is that we don't need to install the extracted context as the current one in any sense, we only need to pass it to the Start call as an option to signify an extracted parent context.

Concept for PassThroughTracer was to enable propagating tracecontext without fully enabling tracing. For example, in Istio, envoy-proxy does all the instrumentation. However, it requires trace-context header be forwarded through application. In such case PassThroughTracer would be very useful.

May be it should be pulled out from here for now and can be added later it it is needed.

rghetia · 2019-09-09T18:54:41Z

@jmacd , @tedsuo and @krnowak PTAL.
I added MockTracer and MockSpan under internal directory to test Inject method. MockTracer is also used in current_test.go which I will replace once this is approved and merged.

jmacd · 2019-09-12T17:56:37Z

We discussed in today's SIG call some questions about how to support OT baggage in the bridge for inject/extract. @krnowak will be working on this.

krnowak

Some nits and comments.

This PR seems to useful for implementing the base of the OTEP 42, but otherwise is not really usable at the moment, since there is no connection between the tracer and the propagator, so which propagator should be used when injecting or extracting needs either to be hardcoded or to have some nonOpenTelemetry infrastructure (normally it would be some kind of propagator registry from OTEP 42, but it's not there yet).

krnowak · 2019-09-09T10:07:48Z

api/propagation/noop_propagator.go

+// GetAllKeys returns empty list of strings.
+func (np NoopTextFormatPropagator) GetAllKeys() []string {
+	return []string{}
+}


missing newline

krnowak · 2019-09-09T10:13:34Z

api/trace/pass_through_span.go

+func (ds *PassThroughSpan) SetName(name string) {
+}
+
+// Tracer returns noop implementation of Tracer.


It returns a pass through tracer.

That's an old comment, passthrough tracer seems to be gone.

krnowak · 2019-09-09T10:20:39Z

api/trace/pass_through_tracer.go

+
+// WithSpan wraps around execution of func with noop span.
+func (t PassThroughTracer) WithSpan(ctx context.Context, name string, body func(context.Context) error) error {
+	return body(ctx)


Shouldn't it make a call to Start first to create a span and put it into context? While I see that WithSpan does not allow passing any options, which means that you can't specify a remote context, so Start will return a NoopSpan anyway, I still think that Start should put this NoopSpan into context too - it's predictable and maybe will allow for easier testing of propagators within the single process.

That's an old comment, passthrough tracer seems to be gone.

krnowak · 2019-09-09T10:21:20Z

api/trace/pass_through_tracer.go

+		op(&opts)
+	}
+	if !opts.RemoteSpanContext.IsValid() {
+		return ctx, NoopSpan{}


Shouldn't this put the NoopSpan into the context, just in case to cover some previous span during testing, for example?

That's an old comment, passthrough tracer seems to be gone.

krnowak · 2019-09-16T08:28:28Z

example/http/client/client.go

-			ctx, req, inj := httptrace.W3C(ctx, req)
-
-			trace.Inject(ctx, inj)
+			propagator := propagation.HttpTraceContextPropagator()


For example purposes, I would write it as:

ctx, req = httptrace.W3C(ctx, req) httptrace.Inject(ctx, req)

Otherwise it is confusing that for injection we explicitly get a propagator, but for extraction , we use httptrace.Extract, so we don't use any propagator explicitly.

The example gets away with it, because internally, httptrace.Extract uses the same propagator we use when we inject it here, but I don't think we would want to rely on that.

krnowak · 2019-09-16T09:06:02Z

internal/trace/mock_span.go

+
+var _ apitrace.Span = (*MockSpan)(nil)
+
+// SpancContext returns an invalid span context.


The comment is false. Also s/SpancContext/SpanContext/.

krnowak · 2019-09-16T09:06:31Z

internal/trace/mock_span.go

+func (ms *MockSpan) SetName(name string) {
+}
+
+// Tracer returns noop implementation of Tracer.


That should say that it returns a mock tracer.

krnowak · 2019-09-16T09:10:33Z

internal/trace/mock_tracer.go

+
+// WithSpan wraps around execution of func with noop span.
+func (mt *MockTracer) WithSpan(ctx context.Context, name string, body func(context.Context) error) error {
+	return body(ctx)


Shouldn't that do span := mt.Start(ctx, name) and defer span.Finish() before using the body callback?

If needed for testing then yes. Otherwise it should be a noop.

krnowak · 2019-09-16T09:14:19Z

propagation/http_trace_context_propagator.go

+// HttpTraceContextPropagator creates a new text format propagator that propagates SpanContext
+// in W3C TraceContext format.
+//
+// The propagator is then used to create CarrierInjector and CarrierExtractor associated with a


This paragraph sounds like a comment from the future. I see no notion of CarrierInjectors or CarrierExtractors in this PR.

comment was from past ;-). Earlier I had CarrierInjectors and CarrierExtractors.
Removed it.

krnowak · 2019-09-16T09:16:32Z

propagation/http_trace_context_propagator_test.go

+
+func TestInjectTraceContextToHTTPReq(t *testing.T) {
+	var id uint64
+	trace.SetGlobalTracer(&mocktrace.MockTracer{


Is there a point in installing the global tracer in the test? The test does not use any API that calls into the global tracer.

I had trace.GlobalTracer().Start() but I don't need that as I can use mockTracer.Start().
Removed installing global tracer.

rghetia · 2019-09-18T16:19:24Z

Some nits and comments.

This PR seems to useful for implementing the base of the OTEP 42, but otherwise is not really usable at the moment, since there is no connection between the tracer and the propagator, so which propagator should be used when injecting or extracting needs either to be hardcoded or to have some nonOpenTelemetry infrastructure (normally it would be some kind of propagator registry from OTEP 42, but it's not there yet).

The connection between tracer and propagator is not necessary. Typically plugins (like http plugins - othttp) would take propagators as an option. Now you could have propagator registry to provide the default propagator but you would still need to be able to override the default on per instance of othttp. For example, a service (A) may be interacting with two other services (B and C). One (C) of them is a legacy service that only accepts B3 Headers. So by default you would have W3C propagator but for B3 propagator for C.

krnowak

Went through the trace context spec and noticed some things.

krnowak · 2019-09-19T09:53:03Z

propagation/http_trace_context_propagator.go

+		return core.EmptySpanContext()
+	}
+	version := int(ver[0])
+	if version > maxVersion {


I went through the spec and noticed this in 3.2.4 Versioning of traceparent:

This specification is opinionated about future versions of trace context. The current version of this specification assumes that future versions of the traceparent header will be additive to the current one.

So this needs to stay as is.

krnowak · 2019-09-19T09:58:44Z

propagation/http_trace_context_propagator.go

+	}
+	sc.SpanID = result
+
+	opts, err := hex.DecodeString(sections[3])


We should check here if len(sections[3]) == 2. This is also in 3.2.4 Versioning of traceparent:

Parse the sampled bit of flags (2 characters from the third dash). Vendors MUST check that the 2 characters are either the end of the string or a dash.

Also, I wonder if we should have a check like if version == 0 && opts[0] > 2 { return core.EmptySpanContext() } as the only specified flag currently is a "sampled" flag for which the first bit is reserved and all other bits should be zero. See 3.2.2.3.3 Other Flags:

The behavior of other flags, such as (00000100) is not defined and is reserved for future use. Vendors MUST set those to zero.

krnowak · 2019-09-19T10:08:28Z

propagation/http_trace_context_propagator.go

+		return core.EmptySpanContext()
+	}
+	sc.TraceID.Low = result
+


Should we add a check if sc.TraceID.Low == 0 && sc.TraceID.High == 0 { return core.EmptySpanContext() }? See first paragraph of 3.2.2.3 trace-id:

All bytes as zero (00000000000000000000000000000000) is considered an invalid value.

Never mind that - I see that we basically check it at the end of the function.

krnowak · 2019-09-19T10:09:57Z

propagation/http_trace_context_propagator.go

+		return core.EmptySpanContext()
+	}
+	result, err = strconv.ParseUint(sections[2][0:], 16, 64)
+	if err != nil {


Should this be err != nil || result == 0? See first paragraph in 3.2.2.3.1 parent-id:

All bytes as zero (0000000000000000) is considered an invalid value.

Never mind that, I see that we check it at the end of the function.

krnowak · 2019-09-19T10:20:23Z

propagation/http_trace_context_propagator.go

+	if len(sections) < 4 {
+		return core.EmptySpanContext()
+	}
+


Should we really check if all the characters in at least first four sections are either digits or lowercase letters from a to f? That's what the W3C trace context spec says in several places, like

HEXDIGLC = DIGIT / "a" / "b" / "c" / "d" / "e" / "f" ; lowercase hex character

or

Vendors MUST ignore the traceparent when the parent-id is invalid (for example, if it contains non-lowercase hex characters).

added the check.

krnowak · 2019-09-19T10:29:03Z

propagation/http_trace_context_propagator_test.go

+		{
+			name:   "future version",
+			header: "02-4bf92f3577b34da6a3ce929d0e0e4736-00f067aa0ba902b7-01",
+			wantSc: core.SpanContext{


Yeah, that's what spec says, so it the test should stay as it is.

krnowak · 2019-09-19T10:47:19Z

propagation/http_trace_context_propagator.go

+	if err != nil || len(opts) < 1 {
+		return core.EmptySpanContext()
+	}
+	sc.TraceOptions = opts[0]


I think this should be sc.TraceOptions = opts[0] &^ core.TraceOptionUnused.

krnowak · 2019-09-19T10:49:51Z

propagation/http_trace_context_propagator_test.go

+			header: "00-4bf92f3577b34da6a3ce929d0e0e4736-00f067aa0ba902b7",
+			wantSc: core.EmptySpanContext(),
+		},
+		{


Some suggestions for the tests:

"wrong length of version", "0000-4bf92f3577b34da6a3ce929d0e0e4736-00f067aa0ba902b7-01"

"bogus version", "qw-4bf92f3577b34da6a3ce929d0e0e4736-00f067aa0ba902b7-01"

"invalid version", "ff-4bf92f3577b34da6a3ce929d0e0e4736-00f067aa0ba902b7-01"

"wrong length of trace id", "00-4bf92f3577b34da6a3ce929d0e0e47-00f067aa0ba902b7-01"

"bogus trace id", "00-qwf92f3577b34da6a3ce929d0e0e4736-00f067aa0ba902b7-01"

"wrong length of parent id", "00-4bf92f3577b34da6a3ce929d0e0e4736-00f067aa0ba902-01"

"bogus parent id", "00-4bf92f3577b34da6a3ce929d0e0e4736-qwf067aa0ba902b7-01"

"valid trace id, zero parent id", "00-4bf92f3577b34da6a3ce929d0e0e4736-0000000000000000-01"

"valid parent id, zero trace id", "00-00000000000000000000000000000000-00f067aa0ba902b7-01"

"wrong length of options", "00-4bf92f3577b34da6a3ce929d0e0e4736-00f067aa0ba902b7-0101"

"invalid options", "00-4bf92f3577b34da6a3ce929d0e0e4736-00f067aa0ba902b7-02"

"bogus options", "00-4bf92f3577b34da6a3ce929d0e0e4736-00f067aa0ba902b7-qw"

"ignored future options", "aa-4bf92f3577b34da6a3ce929d0e0e4736-00f067aa0ba902b7-ff"

should result in valid span context with only sampling bit set in options

Not sure about those yet:

"version with invalid hex characters", "0A-4bf92f3577b34da6a3ce929d0e0e4736-00f067aa0ba902b7-01"

"trace id with invalid hex characters", "00-4BF92f3577b34da6a3ce929d0e0e4736-00f067aa0ba902b7-01"

"parent id with invalid hex characters", "00-4BF92f3577b34da6a3ce929d0e0e4736-00F067AA0ba902b7-01"

"options with invalid hex characters", "01-4BF92f3577b34da6a3ce929d0e0e4736-00F067AA0ba902b7-A1"

added all test cases including upper-case hex characters.
Also added future additional data "aa-4bf92f3577b34da6a3ce929d0e0e4736-00f067aa0ba902b7-ff-XYZxsf09"

krnowak · 2019-09-19T10:51:27Z

propagation/http_trace_context_propagator_test.go

+	if diff := cmp.Diff(got, want); diff != "" {
+		t.Errorf("GetAllKeys: -got +want %s", diff)
+	}
+}


Missing newline at the end of the file.

krnowak · 2019-09-19T10:55:56Z

propagation/http_trace_context_propagator_test.go

+			},
+			wantHeader: "00-4bf92f3577b34da6a3ce929d0e0e4736-0000000000000002-00",
+		},
+		{


I wonder if there should be a test case with a valid span context having options like 0xff, so the expected header should have 01 as options, not ff.

with version '0' I added test to check that option should not be >2. For future version I added other values of option but expected value to be 01 or 00.

I can't see those tests anywhere.

for version 0, anything other than 01 should fail

{ name: "trace-flag unused bits set", header: "00-ab000000000000000000000000000000-cd00000000000000-09", },

For future version accept 01 and 00

{ name: "future options with sampled bit set", header: "02-4bf92f3577b34da6a3ce929d0e0e4736-00f067aa0ba902b7-09", wantSc: core.SpanContext{ TraceID: traceID, SpanID: spanID, TraceOptions: core.TraceOptionSampled, }, }, { name: "future options with sampled bit cleared", header: "02-4bf92f3577b34da6a3ce929d0e0e4736-00f067aa0ba902b7-08", wantSc: core.SpanContext{ TraceID: traceID, SpanID: spanID, }, },

Oh, I meant the test for the inject part, where as input you have a span context like:

core.SpanContext{ TraceID: core.TraceID{High: 0x4bf92f3577b34da6, Low: 0xa3ce929d0e0e4736}, SpanID: uint64(0x00f067aa0ba902b7), TraceOptions: 0xff, }

and the expected generated "traceparent" HTTP header (for now) would be 00-4bf92f3577b34da6a3ce929d0e0e4736-00f067aa0ba902b7-01 (not ff at the end).

rghetia · 2019-09-20T18:05:57Z

@krnowak PTAL.

krnowak

Nits mostly, but I think there are some tests for getting a header out of span context missing.

krnowak · 2019-09-20T18:48:44Z

Makefile

@@ -33,6 +33,9 @@ precommit: $(TOOLS_DIR)/golangci-lint  $(TOOLS_DIR)/misspell $(TOOLS_DIR)/string
 	PATH="$(abspath $(TOOLS_DIR)):$${PATH}" go generate ./...
 	$(TOOLS_DIR)/golangci-lint run --fix # TODO: Fix this on windows.
 	$(TOOLS_DIR)/misspell -w $(ALL_DOCS)
+
+.PHONY: mod-tidy


I'm kinda on the fence with this. I guess it's fine as long as you make mod-tidy a prerequisite of the precommit target. That way CI can run go mod tidy and yell at you when the go.mod or go.sum files change.

krnowak · 2019-09-20T18:50:07Z

propagation/http_trace_context_propagator.go

+	}
+
+	if !traceCtxRegExp.MatchString(h) {
+		fmt.Printf("header does not match regex %s\n", h)


I guess it's some debugging leftover.

krnowak · 2019-09-20T18:52:22Z

propagation/http_trace_context_propagator_test.go

+			header: "00-ab000000000000000000000000000000-cd00000000000000-0100",
+		},
+		{
+			name:   "bogus version length",


Please drop the length word here.

Did you forget to push the fix?

I probably missed when I did search/replace.

krnowak · 2019-09-20T18:52:37Z

propagation/http_trace_context_propagator_test.go

+			header: "qw-00000000000000000000000000000000-0000000000000000-01",
+		},
+		{
+			name:   "bogus trace ID length",


Did you forget to push the fix?

krnowak · 2019-09-20T18:52:44Z

propagation/http_trace_context_propagator_test.go

+			header: "00-qw000000000000000000000000000000-cd00000000000000-01",
+		},
+		{
+			name:   "bogus span ID length",


Did you forget to push the fix?

krnowak · 2019-09-20T18:53:06Z

propagation/http_trace_context_propagator_test.go

+			header: "00-ab000000000000000000000000000000-cd00000000000000-qw",
+		},
+		{
+			name:   "upper case version length",


krnowak · 2019-09-20T18:53:12Z

propagation/http_trace_context_propagator_test.go

+			header: "A0-00000000000000000000000000000000-0000000000000000-01",
+		},
+		{
+			name:   "upper case trace ID length",


krnowak · 2019-09-20T18:53:19Z

propagation/http_trace_context_propagator_test.go

+			header: "00-AB000000000000000000000000000000-cd00000000000000-01",
+		},
+		{
+			name:   "upper case span ID length",


krnowak · 2019-09-20T18:53:25Z

propagation/http_trace_context_propagator_test.go

+			header: "00-ab000000000000000000000000000000-CD00000000000000-01",
+		},
+		{
+			name:   "upper case trace flag length",


krnowak · 2019-09-20T18:55:23Z

propagation/http_trace_context_propagator_test.go

+			},
+			wantHeader: "00-4bf92f3577b34da6a3ce929d0e0e4736-0000000000000002-00",
+		},
+		{


I can't see those tests anywhere.

- added Default Tracer which simply propagates SpanContext. - added CopyOfRemote option to simply create remote span.

- also remove PassThroughTracer

rghetia requested review from bogdandrutu, jmacd, pjanotti, sjkaris and tedsuo as code owners August 9, 2019 21:02

rghetia force-pushed the api_tc_propagation branch from 9411593 to 74ff726 Compare August 13, 2019 19:11

rghetia mentioned this pull request Aug 13, 2019

API: Add b3 propagator #91

Merged

songy23 mentioned this pull request Aug 14, 2019

Remove BinaryFormat and HttpTextFormat from Tracer open-telemetry/opentelemetry-specification#214

Closed

rghetia force-pushed the api_tc_propagation branch 2 times, most recently from 94c9d5d to 107023d Compare August 23, 2019 20:50

rghetia force-pushed the api_tc_propagation branch from 91bcbd8 to 35f8c48 Compare August 26, 2019 17:37

krnowak reviewed Aug 26, 2019

View reviewed changes

rghetia added the do not review label Aug 30, 2019

rghetia force-pushed the api_tc_propagation branch from 066ce9a to 4f5f4ea Compare September 5, 2019 05:36

rghetia removed the do not review label Sep 5, 2019

tedsuo reviewed Sep 5, 2019

View reviewed changes

jmacd reviewed Sep 5, 2019

View reviewed changes

rghetia mentioned this pull request Sep 12, 2019

W3C Correlation Context Propagator #124

Closed

jmacd approved these changes Sep 12, 2019

View reviewed changes

krnowak suggested changes Sep 16, 2019

View reviewed changes

krnowak reviewed Sep 19, 2019

View reviewed changes

rghetia force-pushed the api_tc_propagation branch 2 times, most recently from 65820f7 to 63ae45b Compare September 19, 2019 21:26

krnowak suggested changes Sep 20, 2019

View reviewed changes

rghetia added 26 commits September 23, 2019 11:06

remove Extract api from trace.

fdaf0f0

remove Extract interface for tracer.

7137e7b

fix copyright.

d9c6157

fix variable names and comments.

1e2f4ea

replace INVALID_SPAN_CONTEXT with EmptySpanContext function.

8215b36

move inject/extract out of trace.

ae02d90

fix tag.Map.

ddeb734

make carrier as interface instead of http.Request.

ba6cda4

rename structs and update doc comments..

0c84304

add doc.go

034c2fc

update doc.

2b279af

add noop propagator.

8d3cced

add new propagation api with Supplier interface.

a55726d

- added Default Tracer which simply propagates SpanContext. - added CopyOfRemote option to simply create remote span.

remove old propagator.

8fe1085

rename propagator to TextFormatPropagator.

db8f60b

rename default tracer/span as pass_through tracer/span.

9c3ecb6

add test for pass through tracer.

e5e1a93

add missing interface to pass through tracer.

ecd6ba7

return SpanContext instead of contex.Context from Extract interface.

2a5fc9e

- also remove PassThroughTracer

fix review comments.

c8a05ec

add more test cases for traceContext extraction.

d759fec

remove tidy temporarily from circle-ci target to avoid build failure.

418b0f4

allow header ending in dash '-'.

f271dfc

add inject test for non-zero value other than 01 for traceoption

4ca80c3

add AddLink and Link interface to MockSpan

064af5e

fix running go mod tidy on every build.

ed41e82

rghetia force-pushed the api_tc_propagation branch from 046dfa9 to ed41e82 Compare September 23, 2019 18:07

rghetia merged commit 83935b2 into open-telemetry:master Sep 23, 2019

rghetia deleted the api_tc_propagation branch October 8, 2019 20:05

krnowak mentioned this pull request Oct 9, 2019

REQUEST: New membership for krnowak open-telemetry/community#213

Closed

6 tasks


		var _ apipropagation.TextFormatPropagator = httpTraceContextPropagator{}

		// CarrierExtractor implements CarrierExtractor method of TextFormatPropagator interface.


		var _ apitrace.Span = (*MockSpan)(nil)

		// SpancContext returns an invalid span context.

Add propagator interface and W3C propagator #85

Add propagator interface and W3C propagator #85

Conversation

rghetia commented Aug 9, 2019

rghetia commented Aug 12, 2019

rghetia commented Aug 23, 2019

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

tedsuo left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

rghetia commented Sep 9, 2019

jmacd commented Sep 12, 2019

krnowak left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

rghetia commented Sep 18, 2019

krnowak left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

rghetia commented Sep 20, 2019

krnowak left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment