feat: extract/inject datadog tracestate field p (#115)

Add support for W3C datadog tracestate field `p`.

Author: dmehala

examples/http-server/common/tracingutil.h

@@ -16,7 +16,12 @@ class HeaderWriter final : public datadog::tracing::DictWriter {
   explicit HeaderWriter(httplib::Headers& headers) : headers_(headers) {}
 
   void set(std::string_view key, std::string_view value) override {
-    headers_.emplace(key, value);
+    auto found = headers_.find(std::string(key));
+    if (found == headers_.cend()) {
+      headers_.emplace(key, value);
+    } else {
+      found->second = value;
+    }
   }
 };
 

examples/http-server/proxy/proxy.cpp

@@ -49,37 +49,38 @@ int main() {
   auto forward_handler = [&tracer, &upstream_client](
                              const httplib::Request& req,
                              httplib::Response& res) {
-    auto span = tracer.create_span();
-    span.set_name("forward.request");
-    span.set_resource_name(req.method + " " + req.path);
-    span.set_tag("network.origin.ip", req.remote_addr);
-    span.set_tag("network.origin.port", std::to_string(req.remote_port));
-    span.set_tag("http.url_details.path", req.target);
-    span.set_tag("http.route", req.path);
-    span.set_tag("http.method", req.method);
+    tracingutil::HeaderReader reader(req.headers);
+    auto span = tracer.extract_or_create_span(reader);
+    span->set_name("forward.request");
+    span->set_resource_name(req.method + " " + req.path);
+    span->set_tag("network.origin.ip", req.remote_addr);
+    span->set_tag("network.origin.port", std::to_string(req.remote_port));
+    span->set_tag("http.url_details.path", req.target);
+    span->set_tag("http.route", req.path);
+    span->set_tag("http.method", req.method);
 
     httplib::Error er;
     httplib::Request forward_request(req);
     forward_request.path = req.target;
 
     tracingutil::HeaderWriter writer(forward_request.headers);
-    span.inject(writer);
+    span->inject(writer);
 
     upstream_client.send(forward_request, res, er);
     if (er != httplib::Error::Success) {
       res.status = 500;
-      span.set_error_message(httplib::to_string(er));
+      span->set_error_message(httplib::to_string(er));
       std::cerr << "Error occurred while proxying request: " << req.target
                 << "\n";
     } else {
       tracingutil::HeaderReader reader(res.headers);
-      auto status = span.read_sampling_delegation_response(reader);
+      auto status = span->read_sampling_delegation_response(reader);
       if (auto error = status.if_error()) {
         std::cerr << error << "\n";
       }
     }
 
-    span.set_tag("http.status_code", std::to_string(res.status));
+    span->set_tag("http.status_code", std::to_string(res.status));
   };
 
   httplib::Server server;

src/datadog/extracted_data.h

@@ -23,6 +23,10 @@ struct ExtractedData {
   bool delegate_sampling_decision = false;
   Optional<int> sampling_priority;
   // If this `ExtractedData` was created on account of `PropagationStyle::W3C`,
+  // then `datadog_w3c_parent_id` contains the parts of the "tracestate"
+  // refering to the latest datadog parent ID.
+  Optional<std::string> datadog_w3c_parent_id;
+  // If this `ExtractedData` was created on account of `PropagationStyle::W3C`,
   // then `additional_w3c_tracestate` contains the parts of the "tracestate"
   // header that are not the "dd" (Datadog) entry. If there are no other parts,
   // then `additional_w3c_tracestate` is null.

src/datadog/extraction_util.cpp

@@ -295,6 +295,7 @@ ExtractedData merge(const std::vector<ExtractedData>& contexts) {
       });
 
   if (other != contexts.end()) {
+    result.datadog_w3c_parent_id = other->datadog_w3c_parent_id;
     result.additional_w3c_tracestate = other->additional_w3c_tracestate;
     result.additional_datadog_w3c_tracestate =
         other->additional_datadog_w3c_tracestate;

src/datadog/tags.cpp

@@ -1,7 +1,5 @@
 #include "tags.h"
 
-#include "string_util.h"
-
 namespace datadog {
 namespace tracing {
 namespace tags {
@@ -32,11 +30,10 @@ const std::string process_id = "process_id";
 const std::string language = "language";
 const std::string runtime_id = "runtime-id";
 const std::string sampling_decider = "_dd.is_sampling_decider";
+const std::string w3c_parent_id = "_dd.parent_id";
 
 }  // namespace internal
 
-bool is_internal(StringView tag_name) { return starts_with(tag_name, "_dd."); }
-
 }  // namespace tags
 }  // namespace tracing
 }  // namespace datadog

src/datadog/tags.h

@@ -5,6 +5,7 @@
 
 #include <string>
 
+#include "string_util.h"
 #include "string_view.h"
 
 namespace datadog {
@@ -36,11 +37,14 @@ extern const std::string process_id;
 extern const std::string language;
 extern const std::string runtime_id;
 extern const std::string sampling_decider;
+extern const std::string w3c_parent_id;
 }  // namespace internal
 
 // Return whether the specified `tag_name` is reserved for use internal to this
 // library.
-bool is_internal(StringView tag_name);
+inline bool is_internal(StringView tag_name) {
+  return starts_with(tag_name, "_dd.");
+}
 
 }  // namespace tags
 }  // namespace tracing

src/datadog/trace_segment.cpp

@@ -418,10 +418,11 @@ bool TraceSegment::inject(DictWriter& writer, const SpanData& span,
         writer.set(
             "traceparent",
             encode_traceparent(span.trace_id, span.span_id, sampling_priority));
-        writer.set("tracestate",
-                   encode_tracestate(sampling_priority, origin_, trace_tags,
-                                     additional_datadog_w3c_tracestate_,
-                                     additional_w3c_tracestate_));
+        writer.set(
+            "tracestate",
+            encode_tracestate(span.span_id, sampling_priority, origin_,
+                              trace_tags, additional_datadog_w3c_tracestate_,
+                              additional_w3c_tracestate_));
         break;
       default:
         assert(style == PropagationStyle::NONE);

src/datadog/tracer.cpp

@@ -175,7 +175,7 @@ Expected<Span> Tracer::extract_span(const DictReader& reader,
   }
 
   auto [trace_id, parent_id, origin, trace_tags, delegate_sampling_decision,
-        sampling_priority, additional_w3c_tracestate,
+        sampling_priority, datadog_w3c_parent_id, additional_w3c_tracestate,
         additional_datadog_w3c_tracestate, style, headers_examined] =
       merge(extracted_contexts);
 
@@ -277,6 +277,10 @@ Expected<Span> Tracer::extract_span(const DictReader& reader,
     }
   }
 
+  if (datadog_w3c_parent_id) {
+    span_data->tags[tags::internal::w3c_parent_id] = *datadog_w3c_parent_id;
+  }
+
   Optional<SamplingDecision> sampling_decision;
   if (sampling_priority) {
     SamplingDecision decision;

src/datadog/w3c_propagation.cpp

@@ -221,6 +221,14 @@ void parse_datadog_tracestate(ExtractedData& result, StringView datadog_value) {
           (*result.sampling_priority > 0) == (priority > 0)) {
         result.sampling_priority = priority;
       }
+    } else if (key == "p") {
+      if (value.size() != 16) {
+        // chaff!
+        pair_begin = pair_end == end ? end : pair_end + 1;
+        continue;
+      }
+
+      result.datadog_w3c_parent_id = std::string(value);
     } else if (starts_with(key, "t.")) {
       // The part of the key that follows "t." is the name of a trace tag,
       // except without the "_dd.p." prefix.
@@ -300,6 +308,7 @@ Expected<ExtractedData> extract_w3c(
     return result;
   }
 
+  result.datadog_w3c_parent_id = "0000000000000000";
   extract_tracestate(result, headers);
 
   return result;
@@ -326,11 +335,14 @@ std::string encode_traceparent(TraceID trace_id, std::uint64_t span_id,
 }
 
 std::string encode_datadog_tracestate(
-    int sampling_priority, const Optional<std::string>& origin,
+    uint64_t span_id, int sampling_priority,
+    const Optional<std::string>& origin,
     const std::vector<std::pair<std::string, std::string>>& trace_tags,
     const Optional<std::string>& additional_datadog_w3c_tracestate) {
   std::string result = "dd=s:";
   result += std::to_string(sampling_priority);
+  result += ";p:";
+  result += hex_padded(span_id);
 
   if (origin) {
     result += ";o:";
@@ -382,12 +394,14 @@ std::string encode_datadog_tracestate(
 }
 
 std::string encode_tracestate(
-    int sampling_priority, const Optional<std::string>& origin,
+    uint64_t span_id, int sampling_priority,
+    const Optional<std::string>& origin,
     const std::vector<std::pair<std::string, std::string>>& trace_tags,
     const Optional<std::string>& additional_datadog_w3c_tracestate,
     const Optional<std::string>& additional_w3c_tracestate) {
-  std::string result = encode_datadog_tracestate(
-      sampling_priority, origin, trace_tags, additional_datadog_w3c_tracestate);
+  std::string result =
+      encode_datadog_tracestate(span_id, sampling_priority, origin, trace_tags,
+                                additional_datadog_w3c_tracestate);
 
   if (additional_w3c_tracestate) {
     result += ',';

src/datadog/w3c_propagation.h

@@ -37,7 +37,8 @@ std::string encode_traceparent(TraceID trace_id, std::uint64_t span_id,
 
 // Return a value for the "tracestate" header containing the specified fields.
 std::string encode_tracestate(
-    int sampling_priority, const Optional<std::string>& origin,
+    uint64_t span_id, int sampling_priority,
+    const Optional<std::string>& origin,
     const std::vector<std::pair<std::string, std::string>>& trace_tags,
     const Optional<std::string>& additional_datadog_w3c_tracestate,
     const Optional<std::string>& additional_w3c_tracestate);

test/test_span.cpp

@@ -660,6 +660,7 @@ TEST_CASE("injecting W3C tracestate header") {
   //   - sampling priority
   //   - origin
   //   - trace tags
+  //   - parent id
   //   - extra fields (extracted from W3C)
   //   - all of the above
   // - character substitutions:
@@ -704,79 +705,79 @@ TEST_CASE("injecting W3C tracestate header") {
     {__LINE__, "sampling priority",
      {{"x-datadog-trace-id", "1"}, {"x-datadog-parent-id", "1"},
       {"x-datadog-sampling-priority", "2"}},
-     "dd=s:2"},
+     "dd=s:2;p:$parent_id"},
 
     {__LINE__, "origin",
      {{"x-datadog-trace-id", "1"}, {"x-datadog-parent-id", "1"},
       {"x-datadog-origin", "France"}},
       // The "s:-1" comes from the 0% sample rate.
-     "dd=s:-1;o:France"},
+     "dd=s:-1;p:$parent_id;o:France"},
 
     {__LINE__, "trace tags",
      {{"x-datadog-trace-id", "1"}, {"x-datadog-parent-id", "1"},
       {"x-datadog-tags", "_dd.p.foo=x,_dd.p.bar=y,ignored=wrong_prefix"}},
       // The "s:-1" comes from the 0% sample rate.
-     "dd=s:-1;t.foo:x;t.bar:y"},
+     "dd=s:-1;p:$parent_id;t.foo:x;t.bar:y"},
 
     {__LINE__, "extra fields",
      {{"traceparent", traceparent_drop}, {"tracestate", "dd=foo:bar;boing:boing"}},
     // The "s:0" comes from the sampling decision in `traceparent_drop`.
-    "dd=s:0;foo:bar;boing:boing"},
+    "dd=s:0;p:$parent_id;foo:bar;boing:boing"},
 
     {__LINE__, "all of the above",
      {{"traceparent", traceparent_drop},
       {"tracestate", "dd=o:France;t.foo:x;t.bar:y;foo:bar;boing:boing"}},
     // The "s:0" comes from the sampling decision in `traceparent_drop`.
-    "dd=s:0;o:France;t.foo:x;t.bar:y;foo:bar;boing:boing"},
+    "dd=s:0;p:$parent_id;o:France;t.foo:x;t.bar:y;foo:bar;boing:boing"},
 
     {__LINE__, "replace invalid characters in origin",
      {{"x-datadog-trace-id", "1"}, {"x-datadog-parent-id", "1"},
       {"x-datadog-origin", "France, is a country=nation; so is 台北."}},
       // The "s:-1" comes from the 0% sample rate.
-     "dd=s:-1;o:France_ is a country~nation_ so is ______."},
+     "dd=s:-1;p:$parent_id;o:France_ is a country~nation_ so is ______."},
 
     {__LINE__, "replace invalid characters in trace tag key",
      {{"x-datadog-trace-id", "1"}, {"x-datadog-parent-id", "1"},
       {"x-datadog-tags", "_dd.p.a;d台北x =foo,_dd.p.ok=bar"}},
       // The "s:-1" comes from the 0% sample rate.
-     "dd=s:-1;t.a_d______x_:foo;t.ok:bar"},
+     "dd=s:-1;p:$parent_id;t.a_d______x_:foo;t.ok:bar"},
 
     {__LINE__, "replace invalid characters in trace tag value",
      {{"x-datadog-trace-id", "1"}, {"x-datadog-parent-id", "1"},
       {"x-datadog-tags", "_dd.p.wacky=hello fr~d; how are คุณ?"}},
       // The "s:-1" comes from the 0% sample rate.
-     "dd=s:-1;t.wacky:hello fr_d_ how are _________?"},
+     "dd=s:-1;p:$parent_id;t.wacky:hello fr_d_ how are _________?"},
 
     {__LINE__, "replace equal signs with tildes in trace tag value",
      {{"x-datadog-trace-id", "1"}, {"x-datadog-parent-id", "1"},
       {"x-datadog-tags", "_dd.p.base64_thingy=d2Fra2EhIHdhaw=="}},
       // The "s:-1" comes from the 0% sample rate.
-     "dd=s:-1;t.base64_thingy:d2Fra2EhIHdhaw~~"},
+     "dd=s:-1;p:$parent_id;t.base64_thingy:d2Fra2EhIHdhaw~~"},
 
     {__LINE__, "oversized origin truncates it and subsequent fields",
      {{"x-datadog-trace-id", "1"}, {"x-datadog-parent-id", "1"},
       {"x-datadog-origin", "long cat is looooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooong"},
       {"x-datadog-tags", "_dd.p.foo=bar,_dd.p.honk=honk"}},
       // The "s:-1" comes from the 0% sample rate.
-     "dd=s:-1"},
+     "dd=s:-1;p:$parent_id"},
 
     {__LINE__, "oversized trace tag truncates it and subsequent fields",
      {{"x-datadog-trace-id", "1"}, {"x-datadog-parent-id", "1"},
       {"x-datadog-tags", "_dd.p.foo=bar,_dd.p.long_cat_is=looooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooong,_dd.p.lost=forever"}},
       // The "s:-1" comes from the 0% sample rate.
-     "dd=s:-1;t.foo:bar"},
+     "dd=s:-1;p:$parent_id;t.foo:bar"},
 
     {__LINE__, "oversized extra field truncates itself and subsequent fields",
      {{"traceparent", traceparent_drop},
       {"tracestate", "dd=foo:bar;long_cat_is:looooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooong;lost:forever"}},
      // The "s:0" comes from the sampling decision in `traceparent_drop`.
-     "dd=s:0;foo:bar"},
+     "dd=s:0;p:$parent_id;foo:bar"},
 
     {__LINE__, "non-Datadog tracestate",
      {{"traceparent", traceparent_drop},
       {"tracestate", "foo=bar,boing=boing"}},
      // The "s:0" comes from the sampling decision in `traceparent_drop`.
-     "dd=s:0,foo=bar,boing=boing"},
+     "dd=s:0;p:$parent_id,foo=bar,boing=boing"},
   }));
   // clang-format on
 
@@ -797,6 +798,9 @@ TEST_CASE("injecting W3C tracestate header") {
   const auto found = writer.items.find("tracestate");
   REQUIRE(found != writer.items.end());
 
+  test_case.expected_tracestate.replace(
+      test_case.expected_tracestate.find("$parent_id"),
+      sizeof("$parent_id") - 1, hex_padded(span->id()));
   REQUIRE(found->second == test_case.expected_tracestate);
 
   REQUIRE(logger->error_count() == 0);