My colleague and I are having a head-scratcher moment. We built an internal library that are trying to get working on two machines. The machines are nearly identical in hardware spec.
On my machine i can do this:
>conda create --no-default-packages -n library_test_env python=3.8
>conda activate library_test_env
>pip install git+https_link_to_repository
Which will then install the library and the dependencies. When I do conda list it doesn't list all the dependencies, it does however show some of them but not our own library. My colleague gets a much longer list, which does contains the dependencies and the library.
When i do pip list i get a much longer list than my colleague, which includes the library and the remaining dependencies.
Here is the catch; a specific function in the library uses sqlalchemy to connect to the company database and run a query. The query returns a timeseries which contains a UTC timestamp column.
On my machine the dtype of this column is datetime64 as you would expect, on his machine it is read as an object. Which causes the logic to fail.
We are scratching our heads here. Apart from the different behavior of conda on our machines I just realised he is using windows 11 and I am using windows 10. Apart from the dependencies showing up differently when using conda list and pip list the listed versions are identical.
What could be a possible cause of this?
