fix: Use the same correlation_id in BeforeExecution and `AfterExe…

…cution` events if the API schema contains an error that causes an `InvalidSchema` exception during test execution.

docs/changelog.rst

@@ -4,6 +4,11 @@ Changelog
 `Unreleased`_ - TBD
 -------------------
 
+**Fixed**
+
+- Use the same ``correlation_id`` in ``BeforeExecution`` and ``AfterExecution`` events if the API schema contains an error that
+  causes an ``InvalidSchema`` exception during test execution.
+
 `3.13.1`_ - 2022-02-10
 ----------------------
 

src/schemathesis/runner/impl/core.py

@@ -130,7 +130,8 @@ def _run_tests(
                 feedback = Feedback(self.stateful, operation)
                 # Track whether `BeforeExecution` was already emitted.
                 # Schema error may happen before / after `BeforeExecution`, but it should be emitted only once
-                before_execution_emitted = False
+                # and the `AfterExecution` event should have the same correlation id as previous `BeforeExecution`
+                before_execution_correlation_id = None
                 try:
                     for event in run_test(
                         operation,
@@ -143,7 +144,7 @@ def _run_tests(
                     ):
                         yield event
                         if isinstance(event, events.BeforeExecution):
-                            before_execution_emitted = True
+                            before_execution_correlation_id = event.correlation_id
                         if isinstance(event, events.Interrupted):
                             return
                     # Additional tests, generated via the `feedback` instance
@@ -162,7 +163,7 @@ def _run_tests(
                         results,
                         data_generation_method,
                         recursion_level,
-                        emit_before_execution=not before_execution_emitted,
+                        before_execution_correlation_id=before_execution_correlation_id,
                     )
             else:
                 # Schema errors
@@ -204,7 +205,7 @@ def handle_schema_error(
     data_generation_method: DataGenerationMethod,
     recursion_level: int,
     *,
-    emit_before_execution: bool = True,
+    before_execution_correlation_id: Optional[str] = None,
 ) -> Generator[events.ExecutionEvent, None, None]:
     if error.method is not None:
         assert error.path is not None
@@ -218,9 +219,11 @@ def handle_schema_error(
             data_generation_method=data_generation_method,
         )
         result.add_error(error)
-        correlation_id = uuid.uuid4().hex
-        # It might be already emitted
-        if emit_before_execution:
+        # It might be already emitted - reuse its correlation id
+        if before_execution_correlation_id is not None:
+            correlation_id = before_execution_correlation_id
+        else:
+            correlation_id = uuid.uuid4().hex
             yield events.BeforeExecution(
                 method=method,
                 path=error.full_path,
@@ -244,7 +247,7 @@ def handle_schema_error(
         )
         results.append(result)
     else:
-        # When there is no `method`, then the schema error may cover multiple operations and we can't display it in
+        # When there is no `method`, then the schema error may cover multiple operations, and we can't display it in
         # the progress bar
         results.generic_errors.append(error)
 

test/cli/test_commands.py

@@ -1997,13 +1997,20 @@ def test_get_exit_code(swagger_20, capsys):
 
 
 @pytest.mark.parametrize("location", ("path", "query", "header", "cookie"))
-def test_missing_content_and_schema(cli, testdir, empty_open_api_3_schema, location):
+def test_missing_content_and_schema(cli, tmp_path, testdir, empty_open_api_3_schema, location):
+    debug_file = tmp_path / "debug.jsonl"
     # When an Open API 3 parameter is missing `schema` & `content`
     empty_open_api_3_schema["paths"] = {
         "/foo": {"get": {"parameters": [{"in": location, "name": "X-Foo", "required": True}]}}
     }
     schema_file = testdir.makefile(".json", schema=json.dumps(empty_open_api_3_schema))
-    result = cli.run(str(schema_file), "--dry-run", "--validate-schema=false", "--hypothesis-max-examples=1")
+    result = cli.run(
+        str(schema_file),
+        f"--debug-output-file={debug_file}",
+        "--dry-run",
+        "--validate-schema=false",
+        "--hypothesis-max-examples=1",
+    )
     lines = result.stdout.split("\n")
     # Then CLI should show that this API operation errored
     assert lines[10].startswith("GET /foo E")
@@ -2013,3 +2020,7 @@ def test_missing_content_and_schema(cli, testdir, empty_open_api_3_schema, locat
         lines[14] == f'InvalidSchema: Can not generate data for {location} parameter "X-Foo"! '
         "It should have either `schema` or `content` keywords defined"
     )
+    # And emitted Before / After event pairs have the same correlation ids
+    with debug_file.open(encoding="utf-8") as fd:
+        events = [json.loads(line) for line in fd]
+    assert events[1]["correlation_id"] == events[2]["correlation_id"]