Managing packages
Note
There are many options available for the commands described on this page. For details, see Command reference.
Searching for packages
Use the terminal or an Anaconda Prompt for the following steps.
To see if a specific package, such as SciPy, is available for installation:
conda search scipy
To see if a specific package, such as SciPy, is available for installation from Anaconda.org:
conda search --override-channels --channel defaults scipy
To see if a specific package, such as iminuit, exists in a specific channel, such as http://conda.anaconda.org/mutirri, and is available for installation:
conda search --override-channels --channel http://conda.anaconda.org/mutirri iminuit
Installing packages
Use the terminal or an Anaconda Prompt for the following steps.
To install a specific package such as SciPy into an existing environment "myenv":
conda install --name myenv scipy
If you do not specify the environment name, which in this
example is done by --name myenv
, the package installs
into the current environment:
conda install scipy
To install a specific version of a package such as SciPy:
conda install scipy=0.15.0
To install multiple packages at once, such as SciPy and cURL:
conda install scipy curl
Note
It is best to install all packages at once, so that all of the dependencies are installed at the same time.
To install multiple packages at once and specify the version of the package:
conda install scipy=0.15.0 curl=7.26.0
To install a package for a specific Python version:
conda install scipy=0.15.0 curl=7.26.0 -n py34_env
If you want to use a specific Python version, it is best to use an environment with that version. For more information, see Troubleshooting.
Installing similar packages
Installing packages that have similar filenames and serve similar purposes may return unexpected results. The package last installed will likely determine the outcome, which may be undesirable. If the two packages have different names, or if you're building variants of packages and need to line up other software in the stack, we recommend using Mutex metapackages.
Installing packages from Anaconda.org
Packages that are not available using conda install
can be
obtained from Anaconda.org, a package management service for
both public and private package repositories. Anaconda.org
is an Anaconda product, just like Anaconda and Miniconda.
To install a package from Anaconda.org:
In a browser, go to http://anaconda.org.
To find the package named bottleneck, type
bottleneck
in the top-left box named Search Packages.Find the package that you want and click it to go to the detail page.
The detail page displays the name of the channel. In this example it is the "pandas" channel.
Now that you know the channel name, use the
conda install
command to install the package. In your terminal window or an Anaconda Prompt, run:conda install -c pandas bottleneck
This command tells conda to install the bottleneck package from the pandas channel on Anaconda.org.
To check that the package is installed, in your terminal window or an Anaconda Prompt, run:
conda list
A list of packages appears, including bottleneck.
Note
For information on installing packages from multiple channels, see Managing channels.
Installing non-conda packages
If a package is not available from conda or Anaconda.org, you may be able to find and install the package via conda-forge or with another package manager like pip.
Pip packages do not have all the features of conda packages and we recommend first trying to install any package with conda. If the package is unavailable through conda, try finding and installing it with conda-forge.
If you still cannot install the package, you can try installing it with pip. The differences between pip and conda packages cause certain unavoidable limits in compatibility but conda works hard to be as compatible with pip as possible.
Note
Both pip and conda are included in Anaconda and Miniconda, so you do not need to install them separately.
Conda environments replace virtualenv, so there is no need to activate a virtualenv before using pip.
It is possible to have pip installed outside a conda environment or inside a conda environment.
To gain the benefits of conda integration, be sure to install pip inside the
currently active conda environment and then install packages with that
instance of pip. The command conda list
shows packages installed this way,
with a label showing that they were installed with pip.
You can install pip in the current conda environment with the command
conda install pip
, as discussed in Using pip in an environment.
If there are instances of pip installed both inside and outside the current conda environment, the instance of pip installed inside the current conda environment is used.
To install a non-conda package:
Activate the environment where you want to put the program:
On Windows, in your Anaconda Prompt, run
activate myenv
.On macOS and Linux, in your terminal window, run
conda activate myenv
.
To use pip to install a program such as See, in your terminal window or an Anaconda Prompt, run:
pip install see
To verify the package was installed, in your terminal window or an Anaconda Prompt, run:
conda list
If the package is not shown, install pip as described in Using pip in an environment and try these commands again.
Installing commercial packages
Installing a commercial package such as IOPro is the same as installing any other package. In your terminal window or an Anaconda Prompt, run:
conda install --name myenv iopro
This command installs a free trial of one of Anaconda's commercial packages called IOPro, which can speed up your Python processing. Except for academic use, this free trial expires after 30 days.
Viewing a list of installed packages
Use the terminal or an Anaconda Prompt for the following steps.
To list all of the packages in the active environment:
conda list
To list all of the packages in a deactivated environment:
conda list -n myenv
Listing package dependencies
To find what packages are depending on a specific package in your environment, there is not one specific conda command. It requires a series of steps:
List the dependencies that a specific package requires to run:
conda search package_name --info
Find your installation’s package cache directory:
conda info
Find package dependencies. By default, Anaconda/Miniconda stores packages in ~/anaconda/pkgs/ (or ~/opt/pkgs/ on macOS Catalina). Each package has an index.json file which lists the package’s dependencies. This file resides in ~anaconda/pkgs/package_name/info/index.json.
Now you can find what packages depend on a specific package. Use grep to search all index.json files as follows:
grep package_name ~/anaconda/pkgs/*/info/index.json
The result will be the full package path and version of anything containing the <package_name>.
Example:
grep numpy ~/anaconda3/pkgs/*/info/index.json
Output from the above command:
/Users/testuser/anaconda3/pkgs/anaconda-4.3.0-np111py36_0/info/index.json: numpy 1.11.3 py36_0
/Users/testuser/anaconda3/pkgs/anaconda-4.3.0-np111py36_0/info/index.json: numpydoc 0.6.0 py36_0
/Users/testuser/anaconda3/pkgs/anaconda-4.3.0-np111py36_0/info/index.json: numpy 1.11.3 py36_0
Note this also returned “numpydoc” as it contains the string “numpy”. To get a more specific result set you can add < and >.
Updating packages
Use conda update
command to check to see if a new update is
available. If conda tells you an update is available, you can
then choose whether or not to install it.
Use the terminal or an Anaconda Prompt for the following steps.
To update a specific package:
conda update biopython
To update Python:
conda update python
To update conda itself:
conda update conda
Note
Conda updates to the highest version in its series, so Python 3.8 updates to the highest available in the 3.x series.
To update the Anaconda metapackage:
conda update conda
conda update anaconda
Regardless of what package you are updating, conda compares versions and then reports what is available to install. If no updates are available, conda reports "All requested packages are already installed."
If a newer version of your package is available and you wish to
update it, type y
to update:
Proceed ([y]/n)? y
Preventing packages from updating (pinning)
Pinning a package specification in an environment prevents
packages listed in the pinned
file from being updated.
In the environment's conda-meta
directory, add a file
named pinned
that includes a list of the packages that you
do not want updated.
EXAMPLE: The file below forces NumPy to stay on the 1.7 series, which is any version that starts with 1.7. This also forces SciPy to stay at exactly version 0.14.2:
numpy 1.7.*
scipy ==0.14.2
With this pinned
file, conda update numpy
keeps NumPy at
1.7.1, and conda install scipy=0.15.0
causes an error.
Use the --no-pin
flag to override the update restriction on
a package. In the terminal or an Anaconda Prompt, run:
conda update numpy --no-pin
Because the pinned
specs are included with each conda
install, subsequent conda update
commands without
--no-pin
will revert NumPy back to the 1.7 series.
Adding default packages to new environments automatically
To automatically add default packages to each new environment that you create:
Open Anaconda Prompt or terminal and run:
conda config --add create_default_packages PACKAGENAME1 PACKAGENAME2
Now, you can create new environments and the default packages will be installed in all of them.
You can also edit the .condarc file with a list of packages to create by default.
You can override this option at the command prompt with the --no-default-packages
flag.
Removing packages
Use the terminal or an Anaconda Prompt for the following steps.
To remove a package such as SciPy in an environment such as myenv:
conda remove -n myenv scipy
To remove a package such as SciPy in the current environment:
conda remove scipy
To remove multiple packages at once, such as SciPy and cURL:
conda remove scipy curl
To confirm that a package has been removed:
conda list