There is not a good and easy way to fully share a virtual environment
across machines. The most convenient solution is first to create the
environment and then re-install this environment in a new
environment. A requirements file can be used for this cause. First
we use pip freeze to save a list of requirements to file:
(venv2)Terminal> pip freeze > requirements.txt
If all the above steps in creating the venv2 environment
have been followed, the requirements file should have the following content:
Django==1.4.1
argparse==1.2.1
wsgiref==0.1.2
yolk==0.4.3
In case one has also installed a Python package from the repository of a version control system, the particular commit version (and of course the URL) is recorded as data for that package.
To replicate exactly the same environment inside another Virtualenv
environment, we create a new environment, say it is called venv3,
copy the requirements.txt file to the venv3 environment, and use
pip to install all the packages and their versions specified in
requirements.txt at once:
(venv3)Terminal> pip install -r requirements.txt
A yolk -l command can be used to check the success of the multiple installations.
A word of caution is necessary here. Distributing a requirements.txt
file produced by pip freeze will not always re-create an environment
by a simple pip install -r requirements.txt.
For example, the output from pip freeze may not account for the
fact that some packages must be installed before others.
When creating a Python environment for doing scientific computing,
numpy is a package that must be installed before most other
packages. Some packages can be challenging to compile via
pip install (ScientificPython is an example, although manual
execution of setup.py runs fine).
Also, many Python scientific computing packages depend on
much non-Python software that cannot be installed by pip.
For such complex environments it is recommended to create a script
that performs the manual installation tasks, but it can utilize
pip to as large extent as possible.