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;Conda.pip_interop(bool, env)
: config environment to interact with pip
Conda.pip(command, package, env)
: run pip
command on packages in environmentThe 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.