json_to_metadata: apply on_missing for all rules for non-object json …

…payload (envoyproxy#28935)

Signed-off-by: kuochunghsu <kuochunghsu@pinterest.com>

docs/root/configuration/http/http_filters/json_to_metadata_filter.rst

@@ -46,7 +46,7 @@ comes from the owning HTTP connection manager.
   :header: Name, Type, Description
   :widths: 1, 1, 2
 
-  rq_success, Counter, Total requests that succeed to parse the json body
+  rq_success, Counter, Total requests that succeed to parse the json body. Note that a pure string or number body is treated as a successful json body which increases this counter.
   rq_mismatched_content_type, Counter, Total requests that mismatch the content type
   rq_no_body, Counter, Total requests without content body
   rq_invalid_json_body, Counter, Total requests with invalid json body

source/extensions/filters/http/json_to_metadata/filter.cc

@@ -289,6 +289,19 @@ void Filter::processBody(const Buffer::Instance* body, const Rules& rules,
   }
 
   Json::ObjectSharedPtr body_json = std::move(result.value());
+  // A pure string or number is considered a valid application/json body, but it is not a JSON
+  // object. Therefore, we treat this case as 'on_missing' for all rules in the absence of any
+  // key-value pairs to match.
+  if (!body_json) {
+    ENVOY_LOG(
+        debug,
+        "Apply on_missing for all rules on a valid application/json body but not a json object.");
+    handleAllOnMissing(rules, request_processing_finished_);
+    // This JSON body is valid and successfully parsed.
+    success.inc();
+    return;
+  }
+
   StructMap struct_map;
   for (const auto& rule : rules) {
     const auto& keys = rule.keys_;

test/extensions/filters/http/json_to_metadata/filter_test.cc

@@ -754,6 +754,40 @@ TEST_F(FilterTest, OnMissingForArray) {
   EXPECT_EQ(getCounterValue("json_to_metadata.rq_invalid_json_body"), 0);
 }
 
+TEST_F(FilterTest, OnMissingSecondLayerString) {
+  initializeFilter(R"EOF(
+request_rules:
+  rules:
+  - selectors:
+    - key: foo
+    - key: bar
+    on_present:
+      metadata_namespace: envoy.lb
+      key: baz
+    on_missing:
+      metadata_namespace: envoy.lb
+      key: version
+      value: "missing"
+)EOF");
+  const std::string request_body =
+      R"delimiter(
+        {"foo": "James, James, Morrison, Morrison, weather-beaten, off-key"})delimiter";
+
+  const std::map<std::string, std::string> expected = {{"version", "missing"}};
+
+  EXPECT_EQ(Http::FilterHeadersStatus::StopIteration,
+            filter_->decodeHeaders(incoming_headers_, false));
+
+  EXPECT_CALL(decoder_callbacks_, streamInfo()).WillRepeatedly(ReturnRef(stream_info_));
+  EXPECT_CALL(stream_info_, setDynamicMetadata("envoy.lb", MapEq(expected)));
+  testRequestWithBody(request_body);
+
+  EXPECT_EQ(getCounterValue("json_to_metadata.rq_success"), 1);
+  EXPECT_EQ(getCounterValue("json_to_metadata.rq_mismatched_content_type"), 0);
+  EXPECT_EQ(getCounterValue("json_to_metadata.rq_no_body"), 0);
+  EXPECT_EQ(getCounterValue("json_to_metadata.rq_invalid_json_body"), 0);
+}
+
 TEST_F(FilterTest, NoRequestContentType) {
   initializeFilter(request_config_yaml_);
 
@@ -838,6 +872,58 @@ TEST_F(FilterTest, InvalidJsonPayload) {
   EXPECT_EQ(getCounterValue("json_to_metadata.rq_invalid_json_body"), 1);
 }
 
+TEST_F(FilterTest, OnMissingQuotedString) {
+  initializeFilter(request_config_yaml_);
+  const std::string request_body = R"delimiter("")delimiter";
+  const std::map<std::string, std::string> expected = {{"version", "unknown"}};
+
+  EXPECT_EQ(Http::FilterHeadersStatus::StopIteration,
+            filter_->decodeHeaders(incoming_headers_, false));
+  EXPECT_CALL(decoder_callbacks_, streamInfo()).WillRepeatedly(ReturnRef(stream_info_));
+  EXPECT_CALL(stream_info_, setDynamicMetadata("envoy.lb", MapEq(expected)));
+  testRequestWithBody(request_body);
+
+  EXPECT_EQ(getCounterValue("json_to_metadata.rq_success"), 1);
+  EXPECT_EQ(getCounterValue("json_to_metadata.rq_mismatched_content_type"), 0);
+  EXPECT_EQ(getCounterValue("json_to_metadata.rq_no_body"), 0);
+  EXPECT_EQ(getCounterValue("json_to_metadata.rq_invalid_json_body"), 0);
+}
+
+TEST_F(FilterTest, OnMissingQuotedJsonObject) {
+  initializeFilter(request_config_yaml_);
+  const std::string request_body =
+      R"delimiter("{\"model\": \"gpt-3.5-turbo\",\"temperature\": 0.2,\"stream\": false}")delimiter";
+  const std::map<std::string, std::string> expected = {{"version", "unknown"}};
+
+  EXPECT_EQ(Http::FilterHeadersStatus::StopIteration,
+            filter_->decodeHeaders(incoming_headers_, false));
+  EXPECT_CALL(decoder_callbacks_, streamInfo()).WillRepeatedly(ReturnRef(stream_info_));
+  EXPECT_CALL(stream_info_, setDynamicMetadata("envoy.lb", MapEq(expected)));
+  testRequestWithBody(request_body);
+
+  EXPECT_EQ(getCounterValue("json_to_metadata.rq_success"), 1);
+  EXPECT_EQ(getCounterValue("json_to_metadata.rq_mismatched_content_type"), 0);
+  EXPECT_EQ(getCounterValue("json_to_metadata.rq_no_body"), 0);
+  EXPECT_EQ(getCounterValue("json_to_metadata.rq_invalid_json_body"), 0);
+}
+
+TEST_F(FilterTest, OnMissingPureNumber) {
+  initializeFilter(request_config_yaml_);
+  const std::string request_body = R"delimiter(5566)delimiter";
+  const std::map<std::string, std::string> expected = {{"version", "unknown"}};
+
+  EXPECT_EQ(Http::FilterHeadersStatus::StopIteration,
+            filter_->decodeHeaders(incoming_headers_, false));
+  EXPECT_CALL(decoder_callbacks_, streamInfo()).WillRepeatedly(ReturnRef(stream_info_));
+  EXPECT_CALL(stream_info_, setDynamicMetadata("envoy.lb", MapEq(expected)));
+  testRequestWithBody(request_body);
+
+  EXPECT_EQ(getCounterValue("json_to_metadata.rq_success"), 1);
+  EXPECT_EQ(getCounterValue("json_to_metadata.rq_mismatched_content_type"), 0);
+  EXPECT_EQ(getCounterValue("json_to_metadata.rq_no_body"), 0);
+  EXPECT_EQ(getCounterValue("json_to_metadata.rq_invalid_json_body"), 0);
+}
+
 // TODO(kuochunghsu): Planned to support trimming.
 TEST_F(FilterTest, InvalidJsonForAdditionalPrefixSuffix) {
   initializeFilter(request_config_yaml_);