Torsions

The torsions server is a wrapper around the TorsionAnalyzer commandline application and is its intermediary into the web. It consists of a django backend server with infrastructure to run asynchornous jobs on a redis cluster, as well as various management functionality for Torsion Library and TorsionAnalyzer data. More information can be found in the following publication:

Patrick Penner, Wolfgang Guba, Robert Schmidt, Agnes Meyder, Martin Stahl, and Matthias Rarey. (2022). The Torsion Library: Semiautomated Improvement of Torsion Rules with SMARTScompare. J.Chem.Inf.Model. https://doi.org/10.1021/acs.jcim.2c00043

The frontend submodule can be found here: https://github.com/rareylab/torsions_frontend

A running instance can be found here: https://torsions.zbh.uni-hamburg.de/

Setup

Get python dependencies with conda.

conda create --name torsions -c anaconda -c conda-forge python=3.8 django celery psycopg2 redis redis-py vine pylint pylint-django coverage
conda activate torsions

At this point either set the environment variables $USER, $PASSWORD and $DATABASE to the ones configured in the torsions/settings.py or replace them with appropriate values below.

torsions requires a postgres database and a separate database user. You can create both with this:

psql -d postgres -c "CREATE ROLE $USER WITH ENCRYPTED PASSWORD '$PASSWORD'; ALTER ROLE $USER WITH LOGIN CREATEDB;"
psql -d postgres -c "CREATE DATABASE $DATABASE;"

Perform the torsions migrations and seed the database with:

python manage.py migrate
python manage.py seed

Ensure the bin directory exists and contains the TorsionAnalyzer binary.

The default configured backend and result system for celery is redis. Redis must be installed and available at the url configured in the torsions/settings.py.

Redis is run with the following:

redis-server

Celery workers can be run with:

celery -A torsions worker --loglevel=INFO

...or in a more debug and IDE friendly way:

python /path/to/conda/envs/torsions/bin/celery -A torsions worker --loglevel=INFO

Code Quality

Criteria	Threshold
pylint	>9.0
coverage (overall)	>90%
coverage (single file)	>80%

Run pylint with the django plugin using:

find torsion* -type f -name "*.py" | xargs pylint --load-plugins pylint_django --django-settings-module=torsions.settings

Run coverage generation:

coverage run manage.py test
coverage html

This will generate an HTML coverage report int htmlcov

Deployment

The preferred deployment server is gunicorn. It can be installed like this:

conda install -c anaconda gunicorn

Two scripts, start.sh and stop.sh, are provided as utilities to start and stop all necessary processes. The necessary processes are the following:

• redis
• celery worker
• gunicorn

Each process has its own directory where it will log to and save its PID file to. All processes are daemonized for an unprivileged user. The assumption is made that if a PID file exists the process is running as intended. This may not always be the case if the process encounters an unexpected error or is configured incorrectly. Configuration is kept as close to default as possible and modified values in default configuration files are marked with 'modified'.

A static torsion library is necessary for the torsion_analyzer. Paths to one can be set in torsion_analyzer/settings.py. The STATIC_TORLIB_URL is especially dependent on how static data is served in the deployment.