fixing access to mml spark repo access from databricks_install in mml…

[gramhagen]

Description

fixing issue pulling in maven repository link in mmlspark lightgbm notebook
i'm still seeing issues locally on windows, but we should try this fix on windows dsvm because my environment is probably not very clean.

Related Issues

#709

Checklist:

• My code follows the code style of this project, as detailed in our contribution guidelines.
• I have added tests.
• I have updated the documentation accordingly.

[review-notebook-app]

Check out this pull request on ReviewNB: https://app.reviewnb.com/Microsoft/Recommenders/pull/714

Visit www.reviewnb.com to know how we simplify your Jupyter Notebook workflows.

[miguelgfierro]

good catch, however, we have to add .get("repo", None), if not spark function will break if we use something similar to what we have originally, where there was only coordinates.

[gramhagen]

i wanted to force an error in that case, should we carry forward if they get out of synch?
also, fwiw if you use dict.get(), None is automatically provided as the default. reference

[miguelgfierro]

if you use dict.get(), None is automatically provided as the default.

yeah, so the idea is if we are using the old dictionary, that doesn't have repo, the spark loader doesn't fail:

MMLSPARK_INFO = {"maven": {"coordinates": "Azure:mmlspark:0.16"}}

I guess that the dictionary that we have now is temporal:

MMLSPARK_INFO = {"maven": {"coordinates": "com.microsoft.ml.spark:mmlspark_2.11:0.16.dev8+2.g6a5318b",
                          "repositories": "https://mmlspark.azureedge.net/maven"}
                }

i wanted to force an error in that case

don't understand very well what you are thinking, can you elaborate?

[gramhagen]

I see so, you're just saying we shouldn't bother updating this notebook if we drop the repo piece in the future (which we're planning on doing), i'm fine with that.

[miguelgfierro]

it happened again that the tests hasn't triggered, this is weird

[gramhagen]

I pushed several prs in close succession, i wonder if there's a limit to the queue of pending tests or some kind of time limit before the test status is no longer accessible?