Installing valjean
Requirements
valjean
requires Python >= 3.6. The package dependencies are handled by
the build system.
Getting the code
valjean
is available on GitHub. It can be downloaded using:
$ git clone https://github.com/valjean-framework/valjean.git
Quick start
For those in a rush:
$ python3 -m venv ~/venv-valjean
$ source ~/venv-valjean/bin/activate
(venv-valjean) $ pip install --upgrade pip
(venv-valjean) $ pip install /path/to/valjean
Note
/path/to/valjean
is the location of the package sources.
Using virtual environments
The valjean
package can be installed like a normal Python package, using
the pip package manager.
The recommended way to install valjean
is to use a virtual
environment. At the time of writing, the preferred solution to create virtual
environments is the venv
module from the standard Python distribution:
$ python3 -m venv ~/venv-valjean
$ source ~/venv-valjean/bin/activate
(valjean) $ pip install /path/to/valjean-x.y.z
Note
Ancient versions of pip
(<19.0
) will not be able to install
valjean
, because valjean
uses a pyproject.toml
file
to describe the build, as specified in the PEP 517 and PEP 518
formats. If you are using an old version of pip
, you should upgrade
it (after activating the virtual environment) with:
(venv-valjean) $ pip install --upgrade pip
Prerequisites
valjean
depends on a number of Python packages. You don’t have to do
anything special to install most of these packages, since pip
should take
care of everything. The one exception for the moment is h5py, which will not
work unless the HDF5 library is installed on your machine.
Using conda
It is also possible to use valjean
from a conda
environment. The
first step is to install miniconda or anaconda. The former is a light
installation of python and only required packages will be installed. The latter
is a full installation and can be used offline. Once the installation is done,
you should run:
$ source PATH/TO/CONDA/bin/activate
unless you have set up your shell to do that automatically for you.
The recommended way to install valjean
with conda
is to create a
conda
environment for the package and all of its dependencies:
(base) $ conda create -n MY_ENV python=PY_VERSION
(base) $ conda activate MY_ENV
(MY_ENV) $ conda install -c file://PATH/TO/valjean-DETAILS.tar.bz2 --use-local valjean
DETAILS
stands for vVERSION-NUMBER_HASH_pyPY_VERSION
with:
VERSION
: last tag from valjean in the branch used to build the archiveNUMBER
: number of commits since this tagHASH
: short hash of the commit usedPY_VERSION
: python version used to build the archive, the version used for the installation should be the same.
This procedure should allow to use valjean
from the python interpreter,
from a jupyter notebook or directly with the valjean
command.
Note: only the valjean
package is installed at that step, the others
(numpy, pyparsing, …) will be installed when running valjean
. If
you want to use valjean
directly in python you’ll probably need to
install the required packages using conda install PACKAGE
.
An offline installation is possible adding the --offline
option in the
installation command line. As a consequence updates of packages won’t be
possible, i.e. they will come from the available ones in the local installation
of conda
. The Python version of the package should probably be the default
one of conda
.
The conda package is not editable. If you prefer to use an editable version of
valjean
associated with conda it would probably be easier to use conda
to get the python version, then install the package via pip
or poetry
.
Virtual environnment have to be treated carefully in that case. To get an
editable version offline fully working it might be necessary to install all
(direct and indirect) dependencies.
Checking package integrity
The md5sum
of the archives (pip
or conda
installation) are given in
Tuleap. To check them, just type md5sum MY_ARCHIVE
and compare the obtained
hash with the one stored on Tuleap.