INSTALL.md

Installation

Requirements

tick currently works on Linux, macOS, and Windows systems and requires Python 3.11 or newer. Please have the required Python dependencies in your Python environment:

• numpy
• scipy
• scikit-learn
• pandas
• dill
• pybind11
• cmake

If you build and install tick via pip these dependencies will automatically be resolved. If not, you can install all of these dependencies easily using:

pip install -r requirements/runtime.txt

Source installations

For source installations, a C++ compiler capable of compiling C++17 source code and CMake 3.24 or newer must be available.

Install using pip

tick is available via pip. In your local Python environment (or global, with sudo rights), do:

pip install tick

Installation may take a few minutes to build and link C++ extensions. At this point tick should be ready to use available (if necessary, you can add tick to the PYTHONPATH as explained below).

Manual Install - Linux -OSX

First you need to clone the repository with

git clone https://github.com/X-DataInitiative/tick.git

and then initialize its submodules (such as cereal) with

git submodule update --init

Install the project directly from the checkout with:

python -m pip install .

This will build all required C++ extensions, and install the extensions in your current site-packages directory.

Manual Install on Windows (Experimental)

Download and install Microsoft Windows Visual Build Tools

https://www.visualstudio.com/downloads/#build-tools-for-visual-studio-2019

Download Python 3.11 or newer and add it to the PATH environment variable.

Download wheel files from http://www.lfd.uci.edu/~gohlke/pythonlibs - this is not required for Anaconda

numpy-1.13.*+mkl-cp36-cp36m-win_amd64.whl
numpydoc-0.7.0-py2.py3-none-any.whl
scipy-0.19.*-cp36-cp36m-win_amd64.whl

Note: To ascertain which version of the wheel files to download you can run

python3 -c "import distutils; from distutils import sysconfig; print(sysconfig.get_config_var('SO'))"

Install the wheel files with pip - this is not required for Anaconda

pip install wheel
pip install numpy-1.13.*+mkl-cp36-cp36m-win_amd64.whl
pip install numpydoc-0.7.0-py2.py3-none-any.whl
pip install scipy-0.19.*-cp36-cp36m-win_amd64.whl

Export Visual Studio environment variables for your architecture

/path/to/VS/VC/Auxiliary/Build/vcvarsall.bat [amd64,x86]

The default path for "vcvarsall.bat" for Visual Studio Build Tools:

C:\Program Files (x86)\Microsoft Visual Studio\2019\BuildTools\VC\Auxiliary\Build\vcvarsall.bat [amd64,x86]

Build and test Tick

python -m pip install .[dev]
python -m pytest -q

Automated windows installation process (Requires CYGWIN or MSYS)

Some scripts have been put together to facilitate in the system setup for Windows This does also require manually downloading the wheel files. This can be used but executing the following file:

./tools/windows/install.sh

In-place installation (for developers)

If you wish to work with the source code of tick it's convenient to have the extensions installed inside the source directory. This way you will not have to install the module for each iteration of the code. Simply do:

python -m pip install -e .[dev]

This will build all extensions and install them in editable mode for the current environment.

Note also that special scripts, intended for developers, are available. ./clean_build_test.sh removes all compiled binaries and runs the full compilation process, with all C++ and Python unit-tests. This can take some time. Similarly, ./build_test.sh runs the full compilation process, without removing first binaries, while full_clean.sh only removes them.

It is possible to build with support for Intel MKL in the tick C++ libraries. This requires having the Intel MKL libraries installed and configuring the CMake build accordingly.

On MacOS it's possible an error may occur when trying to find a "intel_thread.dyld" To fix this try:

export DYLD_LIBRARY_PATH="/opt/intel/lib:$DYLD_LIBRARY_PATH"

Replacing "/opt/intel/lib" to "$MKLROOT/../lib" if necessary.

Upgrading to the latest version of MKL may also resolve this.

Alternative installation method

Also supported is the build tool maiken which works with various compilers on major platforms.

To use maiken, either download a provided binary or compile the source following the instructions in the maiken README @ Part 3: How to build Maiken

Once "mkn" is installed on your system, you should execture the shell script "./mkn.sh" to compile tick.

As with tick in general, this requires python3 and all python pre-requisites. numpy/scipy/etc

This method is provided mainly for developers to decrease compilation times.

To use ccache (or similar) with mkn, edit your ~/.maiken/settings.yaml file so it looks like one here - this can greatly decrease repeated compilation times. ccache is a third party tool which may have to be installed manually on your system. ccache does not work on Windows.

On windows it may be required to configure the maiken settings.yaml manually if the environment variables exported by the Visual Studio batch file "vcvars" are not found, example configurations can be found here

Notes:

Logging from mkn can be enabled with the "KLOG" environment variable

KLOG=1 # INFO
KLOG=2 # ERROR
KLOG=3 # DEBUG

If there is a compile error it should be displayed without KLOG being set.

Python support

tick targets Python 3.11 and newer.