Conda.jl

This package allows one to use conda as a cross-platform binary provider for Julia for other Julia packages, especially to install binaries that have complicated dependencies like Python.

conda is a package manager which started as the binary package manager for the Anaconda Python distribution, but it also provides arbitrary packages. Instead of the full Anaconda distribution, Conda.jl uses the miniconda Python environment, which only includes conda and its dependencies.

At the julia> prompt, type a ] (close square bracket) to get a Julia package prompt pkg>, where you can type add Conda to install this package.

Once Conda is installed, you can run import Conda to load the package and run a variety of package-management functions:

• Conda.add(package, env; channel=""): install a package from a specified channel (optional);
• Conda.rm(package, env): remove (uninstall) a package;
• Conda.update(env): update all installed packages to the latest version;
• Conda.list(env): list all installed packages.
• Conda.add_channel(channel, env): add a channel to the list of channels;
• Conda.channels(env): get the current list of channels;
• Conda.rm_channel(channel, env): remove a channel from the list of channels;
• experimental: read the section Conda and pip below before using the following
  • Conda.pip_interop(bool, env): config environment to interact with pip
  • Conda.pip(command, package, env): run pip command on packages in environment

The parameter env is optional and defaults to ROOTENV. See below for more info.

Conda environments allow you to manage multiple distinct sets of packages in a way that avoids conflicts and allows you to install different versions of packages simultaneously.

The Conda.jl package supports environments by allowing you to pass an optional env parameter to functions for package installation, update, and so on. If this parameter is not specified, then the default "root" environment (corresponding to the path in Conda.ROOTENV) is used. The environment name can be specified as a Symbol, or the full path of the environment (if you want to use an environment in a nonstandard directory) can be passed as a string.

For example:

using Conda
Conda.add("libnetcdf", :my_env)
Conda.add("libnetcdf", "/path/to/directory")
Conda.add("libnetcdf", "/path/to/directory"; channel="anaconda")

(NOTE: If you are installing Python packages for use with PyCall, you must use the root environment.)

Conda.jl can be used as a Provider for BinDeps with the CondaBinDeps package.

To use a pre-existing Conda installation, first create an environment for Conda.jl and then set the CONDA_JL_HOME environment variable to the full path of the environment. (You have to rebuild Conda.jl and many of the packages that use it after this.) In Julia, run:

julia> run(`conda create -n conda_jl python conda`)

julia> ENV["CONDA_JL_HOME"] = "/path/to/miniconda/envs/conda_jl"  # change this to your path

pkg> build Conda

To use a specific conda executable, set the CONDA_JL_CONDA_EXE environment variable to the location of the conda executable. This conda executable can exist outside of the environment set by CONDA_JL_HOME. To apply the settting, rebuild Conda.jl. In Julia, run:

julia> ENV["CONDA_JL_CONDA_EXE"] = "/path/to/miniconda/bin/conda" # change this to the path of the conda executable

pkg> build Conda

The use of CONDA_JL_CONDA_EXE requires at least version 1.7 of Conda.jl.

As of conda 4.6.0 there is improved support for PyPi packages. Conda is still the recommended installation method however if there are packages that are only availible with pip one can do the following:

julia> Conda.pip_interop(true, env)

julia> Conda.pip("install", "somepackage")

julia> Conda.pip("install", ["somepackage1", "somepackage2"])

julia> Conda.pip("uninstall", "somepackage")

julia> Conda.pip("uninstall", ["somepackage1", "somepackage2])

If the uninstall command is to be used noninteractively, one can use "uninstall -y" to answer yes to the prompts.

By default, the Conda.jl package installs Python 3, and this version of Python is used for all Python dependencies. If you want to use Python 2 instead, set CONDA_JL_VERSION to "2" prior to installing Conda. (This only needs to be done once; Conda subsequently remembers the version setting.)

Once you have installed Conda and run its Miniconda installer, the Python version cannot be changed without deleting your existing Miniconda installation. If you set ENV["CONDA_JL_VERSION"]="2" and run Pkg.build("Conda"), it will tell you how to delete your existing Miniconda installation if needed.

Most users will not need to use Python 2. This is provided primarily for developers wishing to test their packages for both Python 2 and Python, e.g. by setting the CONDA_JL_VERSION variable on TravisCI and/or AppVeyor.

Miniforge is a community based conda installer by conda-forge, a part of NumFOCUS. Using miniforge and conda-forge in general avoids using repo.anaconda.com maintained by Anaconda, Inc which has terms of conditions that you may want to avoid. conda-forge packages are hosted on anaconda.org, but Anaconda, Inc has been providing hosting for free under the terms of conda-forge which is BSD-3-Clause on top of the original license of the software packages. To use miniforge, use the CONDA_JL_USE_MINIFORGE environment variable.

julia> ENV["CONDA_JL_USE_MINIFORGE"] = "1"

pkg> build Conda

Note that Conda.jl 1.6 and above will use miniforge by default on x86_64, aarch64 and ppc64le systems.

If you have a special character in your user name (like an umlaut or an accent) the installation which defaults to directory C:\Users\<username>\.julia\Conda\3 will fail on Windows. A space in your user name will also fail on any platform. This is a known issue. The work-around is to install Miniconda to a user-writable directory outside of the home directory. Before installing Conda.jl, choose a directory without space and without special characters and set the environment variable CONDA_JL_HOME as follows inside a julia session:

ENV["CONDA_JL_HOME"] = raw"C:\Conda-Julia\3"
using Pkg
Pkg.build("Conda")

After restarting Julia, you can verify the new installation directory:

using Conda
@show Conda.ROOTENV

If you use IJulia or PyCall, they need to be re-build:

using Pkg
Pkg.build("PyCall")
Pkg.build("IJulia")

In case there is something wrong with Conda configuration, it is possible to clean the installation by deleting the .julia/conda directory.

Conda has been tested on Linux, OS X, and Windows.

Please report any bug or suggestion as an github issue

The Conda.jl package is licensed under the MIT Expat license, and is copyrighted by Guillaume Fraux and contributors.