Fixed B3 Propagator

B3 propagator was returning `None` instead of invalid context after last
change. This resulted in many apps crashing when B3 failed to extract a
valid context from carrier. This PR patches it to not make apps crash
and return an invalid context when one cannot be extracted from the
carrier.

CHANGELOG.md

@@ -19,6 +19,8 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
   ([#1721](https://github.com/open-telemetry/opentelemetry-python/pull/1721))
 - Update bootstrap cmd to use exact version when installing instrumentation packages. 
   ([#1722](https://github.com/open-telemetry/opentelemetry-python/pull/1722))
+- Fix B3 propagator to never return None.
+  ([#1750](https://github.com/open-telemetry/opentelemetry-python/pull/1750))
 
 
 ## [1.0.0](https://github.com/open-telemetry/opentelemetry-python/releases/tag/v1.0.0) - 2021-03-26

propagator/opentelemetry-propagator-b3/src/opentelemetry/propagators/b3/__init__.py

@@ -97,6 +97,8 @@ def extract(
             or self._trace_id_regex.fullmatch(trace_id) is None
             or self._span_id_regex.fullmatch(span_id) is None
         ):
+            if context is None:
+                return trace.set_span_in_context(trace.INVALID_SPAN, context)
             return context
 
         trace_id = int(trace_id, 16)
@@ -119,7 +121,8 @@ def extract(
                     trace_flags=trace.TraceFlags(options),
                     trace_state=trace.TraceState(),
                 )
-            )
+            ),
+            context,
         )
 
     def inject(

propagator/opentelemetry-propagator-b3/tests/test_b3_format.py

@@ -215,6 +215,31 @@ def test_flags_and_sampling(self):
 
         self.assertEqual(new_carrier[FORMAT.SAMPLED_KEY], "1")
 
+    def test_derived_ctx_is_returned_for_success(self):
+        """Ensure returned context is derived from the given context."""
+        old_ctx = {"k1": "v1"}
+        new_ctx = FORMAT.extract(
+            {
+                FORMAT.TRACE_ID_KEY: self.serialized_trace_id,
+                FORMAT.SPAN_ID_KEY: self.serialized_span_id,
+                FORMAT.FLAGS_KEY: "1",
+            },
+            old_ctx,
+        )
+        self.assertIn("current-span", new_ctx)
+        for k, v in old_ctx.items():
+            self.assertIn(k, new_ctx)
+            self.assertEqual(new_ctx[k], v)
+
+    def test_derived_ctx_is_returned_for_failure(self):
+        """Ensure returned context is derived from the given context."""
+        old_ctx = {"k2": "v2"}
+        new_ctx = FORMAT.extract({}, old_ctx)
+        self.assertNotIn("current-span", new_ctx)
+        for k, v in old_ctx.items():
+            self.assertIn(k, new_ctx)
+            self.assertEqual(new_ctx[k], v)
+
     def test_64bit_trace_id(self):
         """64 bit trace ids should be padded to 128 bit trace ids."""
         trace_id_64_bit = self.serialized_trace_id[:16]
@@ -334,3 +359,12 @@ def test_fields(self):
             inject_fields.add(call[1][1])
 
         self.assertEqual(FORMAT.fields, inject_fields)
+
+    def test_extract_none_context(self):
+        """Given no trace ID, do not modify context"""
+        old_ctx = None
+
+        carrier = {}
+        new_ctx = FORMAT.extract(carrier, old_ctx)
+        self.assertIsNotNone(new_ctx)
+        self.assertEqual(new_ctx["current-span"], trace_api.INVALID_SPAN)