Batch csv evaluation returns NULL id's about openscoring-python

[vruusmann]

The value of the EvaluationResponse@id attribute is the same as the EvaluationRequest@id attribute. Does your evaluation request JSON object have a non-null Id?

I just tested relevant Openscoring REST API endpoints using the curl command-line application (copy-pasted examples from the README file) and everything works just fine. Have you tested your CSV file with the curl command-line application already?

What is your "requests" Python package version?

[saffrydaffry]

I haven't tried cURL yet, once I try sending a request to /csv I'll post the results. My requests library version is 2.18.4.

Thanks

[saffrydaffry]

I just tested using cURL and the /csv with

curl -X POST --data-binary @features.csv -H "Content-type: text/plain; charset=UTF-8" http://localhost:8080/openscoring/model/testmodel/csv > curl_output.csv

and it looks like from cURL the ID's are processed correctly:

head curl_output.csv
id,true_label,probability_0,probability_1
57977,0,0.9991488452182992,8.511547817011017E-4
31721,0,0.9779992879224841,0.022000712077515847
55943,0,0.9779992879224841,0.022000712077515847
35826,0,0.9929877781321543,0.00701222186784629

This is using the same file as previously. I'm not sure why curl and python would give different results?

[vruusmann]

I have added a unit test to demonstrate that everything works as advertised:
https://github.com/openscoring/openscoring-python/blob/master/openscoring/tests/__init__.py#L12-L51

In my computer, this test passes cleanly with Python 2.7 (+requests 2.10) and Python 3.4 (+requests 2.18):

$ cd openscoring-python
$ nosetests-2.7
$ nosetests-3.4

The output on both cases is this:

Ran 7 tests in 0.281s

OK

Also, there's a new method Openscoring.evaluateDataFrame(id, df), which deals with in-memory Pandas' DataFrame objects (as opposed to CSV files). Their formatting and parsing uses fixed delimiter and quote chars, and therefore should be less error prone.

Anyway, my resolution is that the Openscoring-Python client library is correct, and the problem is located somewhere in your computer.

[saffrydaffry]

I will continue to look at my setup, it does seem like it's something in my system. Thanks for adding the evaluateDataFrame method. I look forward to trying it out.

Thanks for all your help!