Conda: installing local development package into single conda environment

问题

If I were using a virtualenv, I would activate my project's virtual environment then install the package I am developing in develop mode. Something like the following:

workon superbad
pip install -e fnawesome

This allows my package fnawesome to be accessible with any code updates in my superbad virtual environment. If I switch to any other environment, including the default environment, superbad id not accessible.

How are people doing similar setups using conda?

回答1:

You can configure a list of default packages that will be installed into any conda environment automatically

conda config --add create_default_packages pip --add create_default_packages ipython

will make it so that conda create will always include pip and ipython in new environments (this command is the same as adding

create_default_packages:
  - ipython
  - pip

to your .condarc file).

To create an environment without these, use conda create --no-default-packages.

回答2:

Okay, I figured out the issue behind the question.

If you create a conda environment, make sure to include pip and ipython. Otherwise, it will not setup the path to point to environment specific versions of these utilities.

so:

conda create -n superbad scikit-learn
source activate superbad
pip install -e fnawesome  # (installs in default env b/c pip is global pip)
ipython  # runs global ipython with access to global site packages
python # runs the environment's python with no access to fnawesome

this works as expected:

conda create -n superbad scikit-learn pip ipython
source activate superbad
pip install -e fnawesome  # installing into superbad site packages
ipython  # runs superbad ipython
python  # runs the environment's python with access to fnawesome
source deactivate
ipython # no access to fnawesome