Why there are pip and conda packages after fresh installation?

Question

All

Windows 10, 64bit, d/l Anaconda 2.5.0 with Python3, 64bit and installed it

After fresh installation i type conda list, and, among packages, I see duplicates like

jupyter                   1.0.0                    py35_1
jupyter-client            4.1.1                     <pip>
jupyter-console           4.1.0                     <pip>
jupyter-core              4.0.6                     <pip>
jupyter_client            4.1.1                    py35_0
jupyter_console           4.1.0                    py35_0
jupyter_core              4.0.6                    py35_0

Is it normal, and why some packages (not all of them, just a few) have duplicates (not quite, there is - vs _) both in conda and pip?

What will happen if I do pip uninstall jupyter-core?

What should be policy toward such packages?

Answer 1

It's a known problem with the output of conda. The packages are installed only once, but due to differing naming conventions, they are listed twice. There is a (now closed) bug report about it: https://github.com/conda/conda/issues/1237

When you ask conda for a list of packages, this is what happens (or used to happen):

• conda knows which packages were installed with conda.
• conda calls pip under the covers to find packages that were installed outside of conda.
• pip lists all packages, regardless of where they came from.
• conda filters the output of pip by removing the packages installed with conda.

Afaict, the problem is with conda-installed packages that contain an underscore. pip replaces underscores with hyphens in its output. Therefore, conda fails to detect that it's a package installed by itself, and lists it as installed by pip. Of course, conda also lists the package with its original, underscored name as installed by conda. Hence the duplicates.