[INSIGHTS-13] Add Algorithmia Insights ability

dinoboy197 commented 3 years ago

Adds the Insights API to the R library per the updates to the R client documentation for Insights.
Makes tests run correctly (they were not running correctly before, so failures were silent!)
Adds CI for running tests and dry-run of submitting to CRAN using RHub
Adds documentation for releases to CRAN
Comments out a couple of broken tests (JIRA filed to fix and re-enable)

This has been tested by:

adding a true CI build of the R client, with:
- running tests in CI
- using R's recommended approach to verifying a packaged submission against CRAN
insights1 testing cluster and correctly pushes insights to the platform (and out to Kafka) with algorithm https://insights1.enthalpy.click/algorithms/algorithmia_admin/traacksept28r

dinoboy197 commented 3 years ago

Moving back to draft as I am add some CI checks and getting things ready for CRAN submission.

dinoboy197 commented 3 years ago

One pre-existing test (loaderThrowsException) is failing now that CI is enabled for tests. Pinged @zeryx for assistance.

dinoboy197 commented 3 years ago

@algorithmiaio/developer-technologies-contractors ping for review - this is finally ready for you to review and approve! :smile:

zeryx commented 3 years ago

Is there no way to test/verify functionality locally @dinoboy197 ? I feel uneasy about oking a build where we can't verify locally, and we had to comment out a number of tests to get things to compile.

dinoboy197 commented 3 years ago

Is there no way to test/verify functionality locally @dinoboy197 ?

I think I'd have to ask you for one test:

AlgorithmHandlerTest.loaderThrowsException was authored by you in this PR

The other two tests were written by James A who is no longer with us, and never went through PR review, so I'm not sure how he tested those:

AlgorithmTest.runWithRawInputOutput and AlgorithmTest.runRaw were written here

It seems they were coded to talk directly to the public marketplace, which isn't a good practice, but unfortunately, that's in the past now.

I feel uneasy about oking a build where we can't verify locally

Do you mean to say that you don't think that CI running the tests in a clean environment is sufficient to prove that the existing tests work?

we had to comment out a number of tests to get things to compile

That's not quite accurate. We commented three tests because they did not pass, not because the runtime or test time code didn't compile.

algorithmiaio / algorithmia-r

[INSIGHTS-13] Add Algorithmia Insights ability #5