Create some unit tests about datamations HOT 9 CLOSED

Looping @chisingh in on this, let's coordinate on unit tests for R and python.

One idea that wasn't obvious when this was created, but makes sense now: unit tests can check the specs generated on the backend language side, perhaps not a checksum or hash but checking for certain key fields in the generated json.

from datamations.

It looks like there are some old specs that no longer work, e.g. here:

https://github.com/microsoft/datamations/blob/main/inst/htmlwidgets/test/index.html

Let's get rid of them to avoid confusion going forward.

from datamations.

I've cleaned out the sandbox folder so it only has working specs.

from datamations.

Re: testing, there are a bunch of tests here that check the format / fields of the json on the R end. Definitely not 100% test coverage because of all the shiny, legacy code, etc, but a decent start on the core functionality:

(the middle columns show lines, covered, missed, etc - but the last column is coverage for each file)

I'm still trying to think of the best way to compare the python vs R specs within the testthat framework (which is commonly used for R packages). If the python code can write specs to a json file, we can read those in, generate the R specs, and check that they're the same as a start!

from datamations.

sounds like a good idea on writing python json specs out to files and testing them w/ the R testthat framework.

@chisingh, let's work on this once python is up and running.

from datamations.

I have added unit tests using pytest framework that can also be enabled to run after every commit once issue #104 is closed. Some of these tests use the raw json files provided by @sharlagelfand to verify that both the outputs match.

from datamations.

@chisingh will output json specs from python and @sharlagelfand will read that in and test w/ testthat in R

from datamations.

Just working on testing those specs @chisingh and it looks like there are some discrepancies in ordering of things - in the R specs, meta comes before data, but in the python specs data comes first.

Similarly in the data.values, the order of fields do not match, e.g. in R:

"gemini_id": 20,
"Work": "Academia",
"datamations_x": 1,
"datamations_y": 85.0122219615483,
"datamations_y_tooltip": 85.0122219615483,
"datamations_y_raw": 85.3303201349918,
"Lower": 84.7637160761228,
"Upper": 85.2607278469738

versus in python:

"gemini_id": 20,
"Work": "Academia",
"datamations_x": 1,
"datamations_y": 85.01222196154829,
"datamations_y_raw": 85.33032013499178,
"datamations_y_tooltip": 85.01222196154829,
"Lower": 84.76371607612279,
"Upper": 85.2607278469738

The order of datamations_y_raw and datamations_y_tooltip are flipped here.

Otherwise the specs look equivalent! Just something it might be nice to keep in mind and change if not too big of a hassle, otherwise I will have to match the ordering before each test.

from datamations.

That's great! I will reorder those fields.

from datamations.