Closed rdmontgomery closed 3 years ago
As for the failing tests, you might get lucky by pinning pandas<1.1
but admittedly there is some tech debt in the CI pipelines. The version pin likely needs to be specified in both .travis.yml
and appveyor.yml
as the requirements.txt
is just a formality in the build.
This approach isn't ideal right now because I'm not sure what other args might be passed to the Data
constructor that we wouldn't want passed onto pd.read_csv
. The alternate constructor would circumvent this, but I figured it might come down to usability preferences. Putting it back up for discussion.
Test incoming!
Start with the easiest approach and see how much more we want.