You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
You can use Hamilton without numpy so there's no reason to have it as a dependency.
By searching numpy in hamilton/**.py, there are 3 results:
hamilton/base.py where NumpyMatrixResult is defined. This could be moved to the hamilton/plugins/h_numpy.py. We would need to solve the import pattern.
hamilton/telemetry.py there's a check for the base.NumpyMatrixBuilder. We can probably just remove or stitch with the above change?
hamilton/log_setup.py where np.seterr is used to ignore some numpy error. This should be safe to delete?
By searchin numpy in tests/**.py there are more results. Numpy is used to test the 3 previous elements, but also as an input for certain tests and for @check_output stuff.
The easy fix is to keep numpy in the test dependencies.
We could move certain tests to tests/plugins/test_h_numpy.py
Also, we would have to make sure that the examples have an explicit dependency on numpy if they are using it.
The text was updated successfully, but these errors were encountered:
looks like the same could be said about pandas? it does pop up in more places, but none look core
Yes! It's only left there for legacy reasons. One thing I've been meaning to do is add a hamilton[raw] that doesn't install it or numpy. The library should have zero dependencies. That's dependent on this.
You can use Hamilton without numpy so there's no reason to have it as a dependency.
By searching
numpy
inhamilton/**.py
, there are 3 results:hamilton/base.py
whereNumpyMatrixResult
is defined. This could be moved to thehamilton/plugins/h_numpy.py
. We would need to solve the import pattern.hamilton/telemetry.py
there's a check for thebase.NumpyMatrixBuilder
. We can probably just remove or stitch with the above change?hamilton/log_setup.py
wherenp.seterr
is used to ignore some numpy error. This should be safe to delete?By searchin
numpy
intests/**.py
there are more results. Numpy is used to test the 3 previous elements, but also as an input for certain tests and for@check_output
stuff.tests/plugins/test_h_numpy.py
Also, we would have to make sure that the examples have an explicit dependency on numpy if they are using it.
The text was updated successfully, but these errors were encountered: