responseFields beneathPath does not work with arrays

[bsb319]

Using:

org.springframework.restdocs:spring-restdocs-core:2.0.0.RELEASE
org.springframework.restdocs:spring-restdocs-mockmvc:2.0.0.RELEASE

Error:

org.springframework.restdocs.snippet.SnippetException: The following parts of the payload were not documented:
[ {
  "details" : "test details",
  "name" : "testName"
} ]
Fields with the following paths were not found in the payload: [details, name]

Test Code that causes error:

List<FieldDescriptor> responseFieldDescriptors = new ArrayList<>();
responseFieldDescriptors.add(fieldWithPath("details").type(JsonFieldType.STRING).description("Details to display to the customer."));
responseFieldDescriptors.add(fieldWithPath("name").type(JsonFieldType.STRING).description("Customer's name."));


resultActions.andDo(
        MockMvcRestDocumentation.document("content-example",
                preprocessRequest(prettyPrint()),
                preprocessResponse(prettyPrint()),
                responseFields(beneathPath("information.customers[]").withSubsectionId("customer"), responseFieldDescriptors)
        )
);

Test Code that runs but displays unwanted "[]." in generated table

List<FieldDescriptor> responseFieldDescriptors = new ArrayList<>();
responseFieldDescriptors.add(fieldWithPath("[].details").type(JsonFieldType.STRING).description("Details to display to the customer."));
responseFieldDescriptors.add(fieldWithPath("[].name").type(JsonFieldType.STRING).description("Customer's name."));


resultActions.andDo(
        MockMvcRestDocumentation.document("content-example",
                preprocessRequest(prettyPrint()),
                preprocessResponse(prettyPrint()),
                responseFields(beneathPath("information.customers").withSubsectionId("customer"), responseFieldDescriptors)
        )
);

Sample Response:

{
    "information": {
        "customers": [
            "details": "test details",
            "name": "testName"
        ]
    }
}

[wilkinsona]

Thanks for the detailed reported.

The sample response that you have shown contains two fields:

• information.customers.[].details
• information.customers.[].name

Combining the beneath path and the field path, you are using the following to document the fields:

• information.customers.details
• information.customers.name

Neither field exists so REST Docs is correctly failing as two fields that exist have not been documented and an attempt has been made to document two fields that do not exist.

I guess what you’d like to be able to do is to use information.customers.[] as the beneath path? It may not work right now as I think it may be interpreted as identifying a non-unique part of the payload. Have you tried it?

[bsb319]

Thanks for the response. I had not tried information.customers.[]. However, I have now and get the same error as beneathPath("information.customers[]"). If it combines beneathPath with fieldWithPath I would think the outcome would be a path of information.customers[].details and information.customers[].name.

[wilkinsona]

@bsb319 I see you've edited your original description. When you did so, did you change information.customers to information.customers[]? My response above was written based on the beneath path being information.customers which is what I thought you had originally written.

information.customers[] and information.customers.[] are equivalent so I am not surprised that using the latter did not make any difference if you were, in fact, originally using the former.

[bsb319]

@wilkinsona Aw, yes sorry I did edit the initial post to add more details. I also have a typo in the sample response above as "customers" is invalid JSON.

Thank you for your help, you have lead me down the following path (detailed below) where I have uncovered what I feel is my issue.
I'm going to keep digging, but it seems I need to get my extracted field to be set with a PathType.MULTI. If it is, then the tests pass.
If you have any additional suggestions I greatly welcome them, but regardless thanks for your assistance so far.

spring-restdocs-core-2.0.0.RELEASE

org.springframework.restdocs.payload.FieldPathPayloadSubsectionExtractor.java

@Override
public byte[] extractSubsection(byte[] payload, MediaType contentType) {
    try {
        ExtractedField extractedField = new JsonFieldProcessor().extract(
                this.fieldPath, objectMapper.readValue(payload, Object.class));
        Object value = extractedField.getValue();
        if (value instanceof List && extractedField.getType() == PathType.MULTI) {
            List<?> extractedList = (List<?>) value;
            if (extractedList.size() == 1) {
                value = extractedList.get(0);
            }
            else {
                throw new PayloadHandlingException(this.fieldPath
                        + " does not uniquely identify a subsection of the payload");
            }
        }
        return objectMapper.writeValueAsBytes(value);
    }
    catch (IOException ex) {
        throw new PayloadHandlingException(ex);
    }
}

Where this.fieldPath is "information.customers[]" (also tested with various paths, see below)
Object value is the LinkedHashMap (which is expected)
However, extractedField.getType() equals PathType.SINGLE, which causes value instanceof List && extractedField.getType() == PathType.MULTI to evaluate to false. If this evaluates to true then the tests pass.

The only place that I am finding the PathType get set is by:

org.springframework.restdocs.payload.JsonFieldPath.java

static boolean matchesSingleValue(List<String> segments) {
        Iterator<String> iterator = segments.iterator();
        while (iterator.hasNext()) {
            String segment = iterator.next();
            if ((isArraySegment(segment) && iterator.hasNext())
                    || isWildcardSegment(segment)) {
                return false;
            }
        }
        return true;
    }

I am not able to get the segment to fulfill the isArraySegment or the isWildcardSegment and still be a valid path.

Test Code

List<FieldDescriptor> responseFieldDescriptors = new ArrayList<>();
responseFieldDescriptors.add(fieldWithPath("details").type(JsonFieldType.STRING).description("Details to display to the customer."));
responseFieldDescriptors.add(fieldWithPath("name").type(JsonFieldType.STRING).description("Customer's name."));


resultActions.andDo(
        MockMvcRestDocumentation.document("content-example",
                preprocessRequest(prettyPrint()),
                preprocessResponse(prettyPrint()),
                responseFields(beneathPath("information.customers[]").withSubsectionId("customer"), responseFieldDescriptors)
        )
);

Test also passes with the following if I manually modify extractedField's type while debugging to PathType.MULTI:
beneathPath("information.customers")
beneathPath("information['customers']")
beneathPath("information.customers[]")
beneathPath("information.customers.[]")

Sample Response:

{
    "information": {
        "customers": [{
            "details": "test details",
            "name": "testName"
        }]
    }
}

[ingogriebsch]

We are currently facing the same problem as @bsb319 but we are using Spring Restdocs 1.2.4. Therefore it would be nice if @wilkinsona could give us a hint if this bug will also be solved in this version.

[wilkinsona]

@gbtec-ingogriebsch This issue's still open. It's yet to be solved in any version.

[ingogriebsch]

@wilkinsona This was clear to me as I wrote the comment. To say it the other way: It would be nice if you would also fix this bug in the 1.2.x branch. :)

[wilkinsona]

Ah, I understand now. Yes, in all likelihood the fix will be made in both 1.2.x and 2.0.x although I've yet to figure out what the fix should be…

[Jeff-Walker]

I'm assuming the open issue means that it still isn't fixed. I'm using 2.0.1.RELEASE.

Is there a workaround so I can make this documentation happen? I'm willing to extend the involved classes and make this work, but at this point, I'd like to avoid spending the time to figure out how to do that.

Edit: Tried it with 2.0.2.RELEASE with no change.

[wilkinsona]

I'm assuming the open issue means that it still isn't fixed

Correct.

Is there a workaround so I can make this documentation happen?

I'm not aware of one.

[Jeff-Walker]

I agree with @bsb319 that the problem is in JsonFieldPath.matchesSingleValue(List<String> segments). As it loops over segments, it sees that [] is an array segment, but it requres there to be a second segment after that.

I'm not sure I understand why the iterator.hasNext() is even there. I'm sure there's some use-case that I can't think of, but if we removed that, this case would work as desired. The JsonFieldPath class is used for all paths, so I'm not sure if it works for everything.

I also experimented with adding a class like FieldPathPayloadSubsectionExtractor but explicitly for array subsections. It basically drops the check for PathType.MULTI. It gets the first element of the List that comes back from extract() which then will return the object that AbstractFieldsSnippet wants.

I'm willing to submit a pull request. I think the JsonFieldPath change is cleaner and makes sense to me, but the new subsection extractor affects fewer scenarios.

[Jeff-Walker]

I've been looking at the unit tests for FieldPathPayloadSubsectionExtractor and comparing those to the JsonPath rules. You could say that it's working as designed, there's a test case for this, but it expects to get the whole list back, I find this surprising. This makes the path a.[] return the same as a, which may be what's expected.

spring-restdocs/spring-restdocs-core/src/test/java/org/springframework/restdocs/payload/FieldPathPayloadSubsectionExtractorTests.java

Lines 73 to 84 in d46703d

	@Test
	@SuppressWarnings("unchecked")
	public void extractSingleElementArraySubsectionOfJsonMap()
	throws JsonParseException, JsonMappingException, IOException {
	byte[] extractedPayload = new FieldPathPayloadSubsectionExtractor("a.[]")
	.extractSubsection("{\"a\":[{\"b\":5}]}".getBytes(),
	MediaType.APPLICATION_JSON);
	List<Map<String, Object>> extracted = new ObjectMapper()
	.readValue(extractedPayload, List.class);
	assertThat(extracted.size(), is(equalTo(1)));
	assertThat(extracted.get(0).get("b"), is(equalTo((Object) 5)));
	}

The a.[] construct isn't valid JsonPath according to Jayway, but it looks like the intent is for it to be equivalent to a.[*], which is the same as a, so that matches. But, in JsonPath, you can use a[0] to get the first object/value in the array. So, using the example in the test of {"a": [{"b":5}]} it would give me {"b":5} which would solve this problem for us.

Looking at JsonFieldProcessor, when doing traverse(), it just returns the collection when it hits []. It should instead do more work to see what's between the brackets, whether or not we're trying to get the array, as in a.[] or some element of the array, as in a.[0].

With this implemented, a new test for this would be:

  @Test
  @SuppressWarnings("unchecked")
  public void __extractSingleElementArraySubsectionOfJsonMap()
      throws JsonParseException, JsonMappingException, IOException {
    byte[] extractedPayload = new FieldPathPayloadSubsectionExtractor("a.[0]")
        .extractSubsection("{\"a\":[{\"b\":5}]}".getBytes(),
            MediaType.APPLICATION_JSON);
    Map<String, Object> extracted = new ObjectMapper()
        .readValue(extractedPayload, Map.class);
    assertThat(extracted.get("b"), is(equalTo((Object) 5)));
  }

I haven't looked deep enough to know what the effort of implementing this is, but what do you think, @wilkinsona ?

[wilkinsona]

Thanks for taking a look, @Jeff-Walker. a.[] is indeed intended to be equivalent to a.[*] in JSON path. I don't think that introducing support for a.[n] will address the original problem that @bsb319 has reported. Looking at their original example, I believe they'd like every item in the array to have the same structure. Using an index to identify a specific item in the array would not give them what they want.

My hope is that this problem can be fixed with a change to JsonFieldPath that doesn't require the introduction of any new features. Once this week's SpringOne Platform is over, I hope to be able to get a solid block of time to make some progress on figuring out what that fix should be.

[Jeff-Walker]

Understood. That's why wrote it up before making a pull request.

For reference, here's a link to my work-around.
https://gist.github.com/Jeff-Walker/5602b416ea84b21c64bc0bbca16910ab

It's a copy of FieldPathPayloadSubsectionExtractor that grabs the first element of a list. It's ad-hoc and not a full solution, but it get's my docs to look like I want them without doing a custom snippet template.

Since [].a and [].b work, maybe it's all a matter of presentation? Again, that was more or less my problem.

Anyway, even if what I've said doesn't help with the real issue, if there's anything I can do to help, let me know.

Also, thanks for this great product.