install_python_pyenv.rst

install_python_pyenv.rst

An fully FOSS alternative to a conda environment is pyenv. The only real drawback to pyenv is that you must be able to compile Python itself.

On ubuntu this can be setup as follows:

### Based on code in https://github.com/Erotemic/local/blob/main/tools/pyenv_ext/pyenv_ext_commands.sh

 # Install requirements for building Python (apt-specific command, might be different for other distros)
 sudo apt-get install -y \
     make build-essential libssl-dev zlib1g-dev \
     libbz2-dev libreadline-dev libsqlite3-dev wget curl llvm libncurses5-dev \
     libncursesw5-dev xz-utils tk-dev libffi-dev liblzma-dev python-openssl

 # Download pyenv
 export PYENV_ROOT="$HOME/.pyenv"
 if [[ ! -d "$PYENV_ROOT" ]]; then
     git clone https://github.com/pyenv/pyenv.git $PYENV_ROOT
     (cd $PYENV_ROOT && src/configure && make -C src)
 fi

 # We will need to add something similar in your bashrc
 if [ -d "$PYENV_ROOT" ]; then
     export PATH="$PYENV_ROOT/bin:$PATH"
     eval "$($PYENV_ROOT/bin/pyenv init -)"
     source $PYENV_ROOT/completions/pyenv.bash
 fi

 # Compiling with optimizations makes Python run ~20% faster:
 # For more info see:
 # https://github.com/docker-library/python/issues/160#issuecomment-509426916
 # https://gist.github.com/nszceta/ec6efc9b5e54df70deeec7bceead0a1d
 # https://clearlinux.org/news-blogs/boosting-python-profile-guided-platform-specific-optimizations
 CHOSEN_PYTHON_VERSION=3.10.5

 PROFILE_TASK="-m test.regrtest --pgo test_array test_base64 test_binascii test_binhex test_binop test_c_locale_coercion test_csv test_json test_hashlib test_unicode test_codecs test_traceback test_decimal test_math test_compile test_threading test_time test_fstring test_re test_float test_class test_cmath test_complex test_iter test_struct test_slice test_set test_dict test_long test_bytes test_memoryview test_io test_pickle"

 PYTHON_CONFIGURE_OPTS="--enable-shared --enable-optimizations --with-computed-gotos --with-lto"

 PYTHON_CFLAGS="-march=native -O2 -pipe" 

 PROFILE_TASK=$PROFILE_TASK \
 PYTHON_CFLAGS="$PYTHON_CFLAGS" \
 PYTHON_CONFIGURE_OPTS="$PYTHON_CONFIGURE_OPTS" \
 pyenv install $CHOSEN_PYTHON_VERSION --verbose

 # Set your global pyenv version, so your prefix maps correctly.
 pyenv shell $CHOSEN_PYTHON_VERSION
 pyenv global $CHOSEN_PYTHON_VERSION

 # Create the virtual environment
 PYENV_PREFIX=$(pyenv prefix)
 python -m venv $PYENV_PREFIX/envs/pyenv$CHOSEN_PYTHON_VERSION

 # Add this to your bashrc so you have access to the pyenv command
 # and optionally auto-start in a virtual environment

 #### START BASHRC PART ###
 echo "#### ADD THIS TO YOUR BASH RC ####"
 BASHRC_CONTENTS='

 # Add the pyenv command to our environment if it exists
 export PYENV_ROOT="$HOME/.pyenv"
 if [ -d "$PYENV_ROOT" ]; then
     export PATH="$PYENV_ROOT/bin:$PATH"
     eval "$($PYENV_ROOT/bin/pyenv init -)"
     source $PYENV_ROOT/completions/pyenv.bash
     export PYENV_PREFIX=$(pyenv prefix)
 fi

 # Optionally auto-activate the chosen pyenv pyenv environment
 CHOSEN_PYTHON_VERSION=3.10.5
 if [ -d "$PYENV_PREFIX/envs/pyenv$CHOSEN_PYTHON_VERSION" ]; then
     source $PYENV_PREFIX/envs/pyenv$CHOSEN_PYTHON_VERSION/bin/activate
 fi
 '
 echo "#### ADD THE ABOVE TO YOUR BASH RC ####"
 echo "$BASHRC_CONTENTS"
 #### END BASHRC PART ####

Create WATCH environment with Pyenv

First create and activate a new virtual environment (note this could be done with conda as well).

If using pyenv installation instructions, then a virtual environment can be created with the standard venv module. Assuming you have installed Python with pyenv the following will create a virtual environment.

CHOSEN_PYTHON_VERSION=3.10.5
# Set your shell to use this pyenv shim
pyenv shell $CHOSEN_PYTHON_VERSION

# Create the virtual environment
python -m venv $(pyenv prefix)/envs/pyenv-watch

# Activate the virtual environment
source $(pyenv prefix)/envs/pyenv-watch/bin/activate

Once you are in a virtual environment (managed by either conda or pyenv), the WATCH Python module can then be installed with pip via the following command, where /path/to/watch-repo is the absolute path to the directory containing to WATCH repo.

NOTE: It is important you install the module with the editable (-e) flag, otherwise changes you make to the module, will not be reflected when you run your scripts.

pip install -e /path/to/watch-repo

This is more commonly done as

cd /path/to/watch-repo
pip install -e .

This installation process is also scripted in the top-level run_developer_setup.sh script and takes care of issues that can arise with opencv-python.

After the watch module has been installed to your python environment, it can be imported from anywhere regardless of the current working directory as long as the virtual environment was installed in is active.