Problem passing JSON long format parameters to Matlab Production Server

I am using MATLAB production server to run functions via REST API. I have a function with 5 input parameters. When I pass parameters using JSON small format (e.g. 5 double) using POST/synchronous method then function returns my result fine. This is my Python code:

    conn = http.client.HTTPConnection("serverURL:9910")
    headers = { "Content-Type": "application/json"}

    body = json.dumps({ "nargout": 1, "rhs" : ['run', inputs] })
    conn.request("POST", "/optimFctExample/model", body, headers)

    response = conn.getresponse()

If

inputs = [1,2,3,4,5]

then it accepts the rhs just fine. However, I would like to pass more complex parameters than double, and I understand I need to use JSON large format for this.

I tested using

input= [{
 "mwtype": "double",
 "mwsize": [1,1],
 "mwdata": [1]
 },{
 "mwtype": "double",
 "mwsize": [1,1],
 "mwdata": [2]
},{
 "mwtype": "double",
 "mwsize": [1,1],
 "mwdata": [3]
},{
 "mwtype": "double",
 "mwsize": [1,1],
 "mwdata": [4]
},{
 "mwtype": "double",
 "mwsize": [1,1],
 "mwdata": [5]
}]

but then the server returns a 400 error of type:

Error while parsing request:
[err] [worker:31] Expected a data inside special float objects to be a string

What am I doing wrong? is it the way I encaplustae the JSON long format? If I try to pass the original long format lhs (I got from GET) to the rhs, then it works, so the problem is not with the inputs call itself, but rather with the rhs format here. Thanks for your help.

Answer 1

[Kojiro Saito]

Instead of mixing small and large JSON formats, please put 'run' parameter to large JSON. The following will work.

input= [{
"mwtype": "char",
"mwsize": [1,3],
"mwdata": ["run"]
},{
"mwtype": "double",
"mwsize": [1,1],
"mwdata": [1]
},{
"mwtype": "double",
"mwsize": [1,1],
"mwdata": [2]
},{
"mwtype": "double",
"mwsize": [1,1],
"mwdata": [3]
},{
"mwtype": "double",
"mwsize": [1,1],
"mwdata": [4]
},{
"mwtype": "double",
"mwsize": [1,1],
"mwdata": [5]
}]
body = json.dumps({ "nargout": 1, "rhs" : inputs })

Answer 2

[Zebulon]

Thanks for the advice, but I get this error:

Too many input arguments.

I should have precised the structure of the model .m fct. It takes two arguments:

function [output] = model(command, inputs)

command is a char type ('run' or 'info') and inputs is an array of MATLAB data types (double or char). From model I dispatch the inputs to 'run' or 'info' functions depending of the command word. Hence the need for two rhs arguemnts there. I suppose I need to encapsulate inputs in a mwdata of type struct or cell then?

Answer 3

[Zebulon]

I answer to myself: this is working:

    command = { "mwtype": "char",
                "mwsize": [1,3],
                "mwdata": ["run"]
                }

    inputs = { 'mwdata': [ -1,
                          8,
                          "tut",
                          5,
                         { 'mwdata': [ -3, 2],
                           'mwsize': [1, 2],
                           'mwtype': 'cell' }
                         ],
                'mwsize': [1, 5],
                'mwtype': 'cell'}

    body = json.dumps({ "nargout": 1, "rhs" : [command, inputs] })

Note that the inputs have to be encapsulated in a cell array or a structure, and then from the .m file access the data using MATLAB cell data acess, e.g. inputs{1} , inputs{2}, inputs{5}{1} etc.

Alternatively, you may use a structure:

     inputs = { 'mwdata': {'value': [ {'mwdata': [-3], 'mwsize': [1, 1], 'mwtype': 'double'},
                          {'mwdata': [6], 'mwsize': [1, 1], 'mwtype': 'double'},
                          { 'mwdata': ['tut'],
                                         'mwsize': [1, 3],
                                         'mwtype': 'char'},
                          {'mwdata': [5], 'mwsize': [1, 1], 'mwtype': 'double'},
                          { 'mwdata': [ { 'mwdata': [-3],
                                          'mwsize': [1, 1],
                                          'mwtype': 'double'},
                                        { 'mwdata': [2],
                                          'mwsize': [1, 1],
                                          'mwtype': 'double'}],
                            'mwsize': [1, 2],
                            'mwtype': 'cell'}]},
                 'mwsize': [1, 5],
                 'mwtype': 'struct'}

Then you may access data using inputs(1).value and inputs(5).value{1} for instance.

That would be great if the MATLAB Prod server documentation had better examples.