Package Installation

To install Halotools, you can either use pip or clone the repo from GitHub and build the source code. Either way, be sure to read the Dependencies section prior to installation.

Using pip

The simplest way to install the latest release of the code is with pip. Before installation, be sure you have installed the package dependencies described in the Dependencies section. If you will be Installing Halotools using a virtual environment, activate the environment before installing with pip:

pip install halotools

Alternatively, you can install using conda:

conda install -c astropy halotools

Either pip or conda will install the latest official release of the code. If instead you want the latest master branch, you will need to build the code from source following the instructions in the next section.


Consider installing Halotools into a virtual environment. Setting this up is completely straightforward and takes less than a minute, even if this is your first time using a virtual environment. Using a virtual environment simplifies not just the current installation but also package upgrades and your subsequent workflow. If you use conda to manage your python distribution, you can find explicit instructions in the Installing Halotools using a virtual environment section of the documentation.

Building from source

If you don’t install the latest release using pip, you can instead clone the cource code and call the setup file. This is the most common way to install Halotools if you want versions of the code that have been updated since the latest official release. In this case, after installation it is particularly important that you follow the instructions in the Verifying your installation section below.

Before installation, be sure you have installed the package dependencies described in the Dependencies section. If you will be Installing Halotools using a virtual environment, activate the environment before following the instructions below. The first step is to clone the halotools repository:

git clone
cd halotools

Installing one of the official releases

All official releases of the code are tagged with their version name, e.g., v0.5. To install a particular release:

git checkout v0.5
python install

This will install the v0.5 release of the code. Other official release versions (e.g., v0.1) can be installed similarly.

Installing the most recent master branch

If you prefer to use the most recent version of the code:

git checkout master
python install

This will install the master branch of the code that is currently under development. While the features in the official releases have a stable API, new features being developed in the master branch may not. However, the master branch may have new features and/or performance enhancements that you may wish to use for your science application. A concerted effort is made to ensure that only thoroughly tested and documented code appears in the public master branch, though Halotools users should be aware of the distinction between the bleeding edge version in master and the official release version available through pip.


Optional: If you need to fine-tune the optimization of an especially performance-critical science application, we recommend that you install the package from source. This will give you the opportunity to manually throw your own compiler flags that are enabled by your version of gcc. For example, throwing the -Ofast and -march=native flags can improve performance of the mock_observables functions by 10-40% (with zero impact on the performance of the mock-making algorithm implemented in empirical_models). To compile Halotools with these flags thrown, simply add two new elements to the extra_compiler_args list in every source code file named the string '-Ofast' and the string '-march=native'. When you’ve made these modifications to the code, install Halotools by following the Building fom source instructions above using with your locally modified source code. Alternatively, if you have an older version of gcc that does not support the default choice for these flags made by Halotools, you may need to remove the flag causing the installation problem.


If you install halotools using pip, then most of your dependencies will be handled for you automatically. The only additional dependency you may need is:

  • h5py: 2.5 or later

The h5py package is used for fast I/O of large simulated datasets.

If you did not use pip, then you should be aware of the following strict requirements:

Any of the above can be installed with either pip or conda.

Verifying your installation

After installing the code and its dependencies, fire up a Python interpreter and check that the version number matches what you expect:

import halotools

If the version number is not what it should be, this likely means you have a previous installation that is superseding the version you tried to install. This should be accomplished by doing pip uninstall halotools before your new installation, but you may need to uninstall the previous build “manually”. Like all python packages, you can find the installation location as follows:

import halotools

This will show where your active version is located on your machine. You can manually delete this copy of Halotools prior to your new installation to avoid version conflicts. (There may be multiple copies of Halotools in this location, depending on how may times you have previously installed the code - all such copies may be deleted prior to reinstallation).

Once you have installed the package, see Getting started with Halotools for instructions on how to get up and running.

Testing your installation

To verify that your Halotools installation runs properly, navigate to some new working directory and execute the test suite. If you installed Halotools into a virtual environment as described in the Installing Halotools using a virtual environment section of the documentation, activate the environment before spawning a python session and executing the code below.

For halotools versions v0.6 and later, there is a test_installation feature that runs a few simple tests scattered throughout the code base:

import halotools
halotools.test_installation()  #  v0.6 and later

For earlier versions, you will need to run the full test suite, which is more memory intensive and takes several minutes to run:

halotools.test()  #  v0.5 and earlier

Whether you installed the master branch or a release branch, the message that concludes the execution of the test suite should not indicate that there were any errors or failures. A typical acceptable test suite report will read something like “445 passed, 45 skipped in 383.2 seconds”. If you installed the master branch, your message may read something like “475 passed, 4 xfailed in 374.3 seconds”. The xfail marker is shorthand for “expected failure”; tests marked by xfail do not indicate a bug or installation problem; instead, this indicates that there is a new feature that has only been partially implemented. If you encounter problems when running the test suite, please be sure you have installed the package dependencies first before raising a Github Issue and/or contacting the Halotools developers.


See Installation Troubleshooting for solutions to known installation-related problems.