Closed andy-esch closed 6 years ago
Takin' a look!
Thanks for the review @talos !
For 1. above, I may not have said it very clearly, but this PR's goal is to get rid of returning null
s in favor of terminating the flow, so all good here.
@talos, I added basic usage of decorators here that reduces the redundancy and increases the maintainability of the code :) Please take a look and let me know what you think. Do you still think that I should take a look at functools
?
It looks like the functools.wrap
just carries over the function metadata? In this case, that doesn't seem like an important property.
@talos, anything else before I hand this PR over?
👍
This fixes a problem where empty lists of data were passed to algorithms. Before the data analysis provider was written, the default was to return
[(None, None, ...)]
(an empty tuple with the correct number of columns). With this PR, I'm choosing a noisy error if the input data has so many null that the data cannot be analyzed.@talos, could you do a CR, please? Specifically, I'm interested in your opinion of the
verify_data
data function, and whether I should be raising an exception and handling it in thetry, except
block instead.closes #143, closes #156