type checking: allow numpy.ndarray when expecting Array(Float64) ?

[jgehrcke]

Hey, when bumping from 0.2.3 to 0.2.4 we now get

TypeMismatchError: Code: 53. Type mismatch in VALUES section. Expected Array(Float64) got <class 'numpy.ndarray'>

in code that worked with version 0.2.3. This is because this check was added: ae8aa4a#diff-2e1d5e4524de8a7ffab2ea70f503b827d6e34fe139f8f3531bd0b59c13eedd51R103

use_numpy=True in client settings does not fix the problem.

Converting the numpy array to a Python list works around the problem.

I wonder: a numpy array of floats is type-wise very close to Array(Float64). Should this pass validation?

[xzkostyan]

Hi.

Can you provide snippet that raises this error?

[jgehrcke]

Thank you for looking! I can't get to making a repro right now, but it's really a simple case: create column of type Array(Float64) and insert a list of numpy arrays (so that each row would be a numpy array).

It works (again) when setting types_check=False upon execute().