Installing from source#

Getting the source#

More likely you will want to get the source repository to be able to follow the latest changes. In that case, you can use:

git clone https://github.com/dipy/dipy.git

For more information about this see Following the latest source.

After you’ve cloned the repository, you will have a new directory, containing the DIPY pyproject.toml file, among others. We’ll call this directory - that contains the pyproject.toml file - the DIPY source root directory. Sometimes we’ll also call it the <dipy root> directory.

Building and installing#

Install from source (all operating systems - recommended)#

Change directory into the DIPY source root directory.

To clean your directory from temporary file, use:

git clean -fxd

This command will delete all files not present in your github repository.

Then, complete your installation by using this command:

pip install -r requirements/build.txt
pip install --no-build-isolation --user -e .

This command will do the following :

install the requirements for building dipy
remove the old dipy installation if present
build dipy
install dipy locally on your user environment

Install from source for Unix (e.g Linux, macOS)#

Change directory into the DIPY source root directory.

First, install some python build packages:

pip install -r requirements/build.txt

Then, to install for the system:

pip install dipy

Or, to build DIPY in the source tree (locally) so you can run the code in the source tree (recommended for following the latest source) run:

pip install --no-build-isolation --user -e .

add the DIPY source root directory into your PYTHONPATH environment variable. Search google for PYTHONPATH for details or see python module path for an introduction.

When adding dipy to the PYTHONPATH, we usually add the PYTHONPATH at the end of ~/.bashrc or (macOS) ~/.bash_profile so we don’t need to retype it every time. This should look something like:

export PYTHONPATH="/home/user_dir/Devel/dipy:$PYTHONPATH"

After changing the ~/.bashrc or (macOS) ~/.bash_profile try:

source ~/.bashrc

or:

source ~/.bash_profile

so that you can have immediate access to DIPY without needing to restart your terminal.

Ubuntu/Debian#

sudo apt-get install python-dev python-setuptools
sudo apt-get install python-numpy python-scipy
sudo apt-get install cython

then:

sudo pip install nibabel

(we need the latest version of this one - hence pip rather than apt-get).

You might want the optional packages too (highly recommended):

sudo apt-get install ipython python-h5py python-vtk python-matplotlib

Now follow Install from source for Unix (e.g Linux, macOS).

Fedora / Mandriva maybe Redhat#

Same as above but use yum rather than apt-get when necessary.

Now follow Install from source for Unix (e.g Linux, macOS).

Windows#

Anaconda is probably the easiest way to install the dependencies that you need. To build from source, you will also need to install the exact compiler which is used with your specific version of python.

For getting this information, type this command in shell like cmd or Powershell:

python -c "import platform;print(platform.python_compiler())"

This command should print information of this form:

MSC v.1900 64 bit (AMD64)

Now that you find the relevant compiler, you have to install the VisualStudioBuildTools by respecting the following table:

Visual C++ 2008  (9.0)          MSC_VER=1500
Visual C++ 2010 (10.0)          MSC_VER=1600
Visual C++ 2012 (11.0)          MSC_VER=1700
Visual C++ 2013 (12.0)          MSC_VER=1800
Visual C++ 2015 (14.0)          MSC_VER=1900
Visual C++ 2017 (15.0)          MSC_VER=1910
Visual C++ 2019 (16.0)          MSC_VER=1920
Visual C++ 2022 (17.0)          MSC_VER=1930

After the VisualStudioBuildTools installation, restart a command shell and change directory into the DIPY source root directory.

Start to install the build tools:

pip install -r requirements/build.txt

Then to install into your system:

pip install dipy

To install inplace - so that DIPY is running out of the source code directory:

pip install --no-build-isolation --user -e .

(this is the mode we recommend for following the latest source code).

If you get an error saying “unable to find vcvarsall.bat” then you need to check your environment variable PATH or reinstall VisualStudioBuildTools. Setuptools should automatically detect the compiler and use it.

macOS#

Make sure you have Xcode and Anaconda installed.

From here follow the Install from source for Unix (e.g Linux, macOS) instructions.

OpenMP with macOS#

OpenMP is a standard library for efficient multithreaded applications. This is used in DIPY for speeding up many different parts of the library (e.g., denoising and bundle registration). If you do not have an OpenMP-enabled compiler, you can still compile DIPY from source using the above instructions, but it might not take advantage of the multithreaded parts of the code. To be able to compile DIPY from source with OpenMP on macOS, you will have to do a few more things. First of all, you will need to install the Homebrew package manager. Next, you will need to install and configure the compiler. You have two options: using the GCC compiler or the CLANG compiler. This depends on your python installation:

Under Anaconda#

We recommend to install llvm via Anaconda. Run the following:

conda install -c conda-forge llvm-openmp

In case the compiler is not detected automatically, you can specify the compiler by using the environment variable CC.

Under Homebrew Python or python.org Python#

If you are already using the Homebrew Python, or the standard python.org Python, you will need to use the llvm compiler with OMP. Run:

brew install llvm
brew install libomp
export CC="/opt/homebrew/opt/llvm/bin/clang"

In case the compiler or OpenMP are not detected , you can specify some environment variables. For example, you can add the following lines to your ~/.bash_profile:

export CC="/opt/homebrew/opt/llvm/bin/clang"
export CXX="/opt/homebrew/opt/llvm/bin/clang++"
export LDFLAGS="-L/opt/homebrew/opt/libomp/lib -L/opt/homebrew/opt/llvm/lib"
export CPPFLAGS="-I/opt/homebrew/opt/libomp/include -I/opt/homebrew/opt/llvm/include"
export CFLAGS="-I/opt/homebrew/opt/libomp/include -I/opt/homebrew/opt/llvm/include"

Building and installing#

Whether you are using Anaconda or Hombrew/python.org Python, you will need to then run pip install dipy. When you do that, it should now compile the code with this OpenMP-enabled compiler, and things should go faster!

Testing#

If you want to run the tests:

sudo pip install pytest

Then, in the terminal from <dipy root>:

pytest -svv dipy

You can also run the examples in <dipy root>/doc.

Documentation#

To build the documentation in HTML in your computer you will need to do:

sudo pip install sphinx

Then change directory to <dipy root> and:

cd doc
make clean
make -C . html

Tip#

Building the entire DIPY documentation takes a few hours. You may want to skip building the documentation for the examples, which will reduce the documentation build time to a few minutes. You can do so by executing:

make -C . html-no-examples

Troubleshooting#

If you encounter the following error when trying to build the documentation:

tools/build_modref_templates.py dipy reference
*WARNING* API documentation not generated: Can not import dipy
tools/docgen_cmd.py dipy reference_cmd
*WARNING* Command line API documentation not generated: Cannot import dipy
Build API docs...done.
cd examples_built && ../../tools/make_examples.py
Traceback (most recent call last):
  File "../../tools/make_examples.py", line 33, in <module>
    import dipy
ModuleNotFoundError: No module named 'dipy'

it is probably due to a conflict between the picked Sphinx version: this happens when the system’s Sphinx package is used instead of the virtual environment’s Sphinx package, and the former trying to import a DIPY version in the system: the Sphinx package used should correspond to that of the virtual environment where DIPY lives. This can be solved by specifying the path to the Sphinx package in the virtual environment:

make html SPHINXBUILD='python <path_to_sphinx>/sphinx-build'