HttpBody field causes problems in client libraries

[chrisdunelm]

Some libraries (e.g. Cloud Machine Learning Engine) require un-typed JSON to be passed through to the servre; ignoring the strongly-typed generated code.
There is no obvious way to do this, leading to users having problems:

• https://stackoverflow.com/questions/43638529/google-cloud-machine-learning-api-dot-net-client-predict-request-failing
• https://stackoverflow.com/questions/43805736/google-machine-learning-api-issue

We might want to change the code generator to make this simple to use for APIs that require this behaviour.

There are two known work-arounds, using CMLE as an example:

1. Use the ModifyRequest hook on the service request:

// Create the underlying service. In real use this will probably require authentication
CloudMachineLearningEngineService service = new CloudMachineLearningEngineService();
// Prepare to send the request
string actualBodyJson = "Request JSON";
GoogleCloudMlV1PredictRequest emptyBody = new GoogleCloudMlV1PredictRequest { HttpBody = new GoogleApiHttpBody() };
ProjectsResource.PredictRequest predictRequest = service.Projects.Predict(emptyBody, "projects/xyz");
predictRequest.ModifyRequest = httpRequestMessage =>
{
    httpRequestMessage.Content = new StringContent(actualBodyJson, Encoding.UTF8, "application/json");
};
// Send the request with the body set to 'actualBodyJson'
GoogleApiHttpBody predictResponse = predictRequest.Execute();

2. Using the service HttpClient directly:

// Create the underlying service. In real use this will probably require authentication
CloudMachineLearningEngineService service = new CloudMachineLearningEngineService();
// Prepare to send the request
string actualBodyJson = "Request JSON";
GoogleCloudMlV1PredictRequest emptyBody = new GoogleCloudMlV1PredictRequest { HttpBody = new GoogleApiHttpBody() };
ProjectsResource.PredictRequest predictRequest = service.Projects.Predict(emptyBody, "projects/xyz");
Uri uri = predictRequest.CreateRequest().RequestUri;
HttpContent content = new StringContent(actualBodyJson, Encoding.UTF8, "application/json");
HttpResponseMessage predictResponse = service.HttpClient.PostAsync(uri, content).Result;

[AKhilGarg91]

(Note: edited by jskeet for formatting only)

This solution is also not working for me

cd = GoogleCredential.FromStream(stream);
var service = new CloudMachineLearningEngineService(new BaseClientService.Initializer()
{
    HttpClientInitializer = cd
});
var completeUri = "projects/projectid/models/model_name";
string req = @"{ ""instances"" :[]}";
var httpBody = new GoogleApiHttpBody();
httpBody.Data = req;
httpBody.ContentType = "application/json";
var body = new GoogleCloudMlV1PredictRequest()
{
    HttpBody = httpBody
};

var a = service.Projects.Predict(body, completeUri);

var bodydummy = new GoogleCloudMlV1PredictRequest { HttpBody = new GoogleApiHttpBody() };

var b = service.Projects.Predict(bodydummy, completeUri);

Uri uri = b.CreateRequest().RequestUri;

HttpContent content = new StringContent(req, Encoding.UTF8, "application/json");
HttpResponseMessage predictResponse = service.HttpClient.PostAsync(uri, content).Result;

a.ModifyRequest = httpRequestMessage =>
{
    httpRequestMessage.Content = new StringContent(req, Encoding.UTF8, "application/json");
};
var res = a.Execute();

both methods are giving error. Can somebody tell me why?

With Python client and Google API explorer everything is working fine.

[jskeet]

It's hard to explain what's wrong when we don't know anything about the error you're facing. Please could you give more details?

[AKhilGarg91]

@jskeet with 2nd metthod I amm getting response like this
{StatusCode: 404, ReasonPhrase: 'Not Found', Version: 1.1, Content: System.Net.Http.StreamContent, Headers:
{
X-XSS-Protection: 1; mode=block
X-Frame-Options: SAMEORIGIN
X-Content-Type-Options: nosniff
Alt-Svc: hq=":443"; ma=2592000; quic=51303431; quic=51303339; quic=51303338; quic=51303337; quic=51303335,quic=":443"; ma=2592000; v="41,39,38,37,35"
Date: Thu, 07 Dec 2017 13:31:06 GMT
Server: ESF
Content-Length: 1622
Content-Type: text/html; charset=UTF-8
}}

from first method:

JsonReaderException: Unexpected character encountered while parsing value: <. Path '', line 0, position 0.

[jskeet]

It sounds like you're getting XML back with the first method. I suggest you turn on response logging, so you can see exactly what's happening. Note that if you turn on both request and response logging, you should be able to see exactly what's sent.

The code you've shown in your earlier comment appears to be a mixture of the first and second approaches. Could you post separate comments with what you tried in each approach? Does the project you're specifying definitely exist? (It's not clear whether you actually used "projects/projectid/models/model_name" or a real one.) If you used that example project name, that could be why you're getting a 404.

[AKhilGarg91]

ok i will try that but this is my code. Same thing working in python client and with Google API explorer.

GoogleCredential cd;

        using (var stream = new FileStream("cred.json", FileMode.Open, FileAccess.Read))
        {
            cd = GoogleCredential.FromStream(stream);   
            var service = new CloudMachineLearningEngineService(new BaseClientService.Initializer()
            {
                HttpClientInitializer = cd
            });
            
            //2nd method
            var completeUri = "projects/currexchange-186113/models/trainedcprr";
            string req = @"{ ""instances"" :[{""CBRate"":9,""recievercurrency"":""HKD"",""sendingcurrency"":""EUR""}]}";
          
            var bodydummy = new GoogleCloudMlV1PredictRequest {HttpBody= new GoogleApiHttpBody() };
            var b = service.Projects.Predict(bodydummy, completeUri);
            Uri uri = b.CreateRequest().RequestUri;
            
            HttpContent content = new StringContent(req, Encoding.UTF8, "application/json");
            HttpResponseMessage predictResponse = service.HttpClient.PostAsync(uri, content).Result;




            //first method
            var httpBody = new GoogleApiHttpBody();
            httpBody.Data = req;
            httpBody.ContentType = "application/json";
            var body = new GoogleCloudMlV1PredictRequest()
            {
                HttpBody = httpBody
            };
            var a = service.Projects.Predict(body
               , completeUri);
            a.ModifyRequest = httpRequestMessage =>
            {
                httpRequestMessage.Content = new StringContent(req, Encoding.UTF8, "application/json");
            };
            var res = a.Execute();

        }

[AKhilGarg91]

[jskeet]

Okay - both of those look like they really should work. I'd suggest using either request/response logging or a tool such as Fiddler to see what the actual requests and responses look like.

[chrisdunelm]

@AKhilGarg91 Which version of the Google.Apis.CloudMachineLearningEngine.v1 package are you using? And what platform is this running on (e.g. Windows console app, ASP.NET core, core console app, etc...)
Thanks

[AKhilGarg91]

The latest version. 1.31.0.1061

Windows console App

.Net Framework 4.6.1

VS 2017

[chrisdunelm]

Unfortunately we've just identified a bug in v1.31 that's causing this problem. Downgrading to v1.30 should resolve the problem temporarily.
We're now working on a new bug-fix release, which will be pushed to nuget as soon as possible.

[AKhilGarg91]

Thanks @chrisdunelm I will try with version 1.30.