FLAME GPU Website

This repository contains the source for the FLAME GPU website, built with Jekyll, and using the Minimal Mistakes theme.

The website is hosted on GitHub Pages and can be found at https://flamegpu.github.io.

Development

Install Dependencies

1. Install Ruby

   • On Windows, this installer can be used https://rubyinstaller.org/

   • On Linux, follow the instructions according to your distribution e.g. for Debian/Ubuntu:

     sudo apt install ruby-full

2. Install bundler (via a terminal):

   gem install bundler

3. Install other dependencies:

   cd path/to/clone/of/this/repo
   bundle config set path vendor/bundle
   bundle install

Note: if you get an error related to the public_suffix package, try installing and updating bundler before rebuilding the site:

gem install public_suffix --version 3.0.3
bundler update

Updating Dependencies

Ensure ruby packages are up to date, to avoid differences between local and GitHub/GitHub Actions builds:

bundle update --all

Serving a Local Copy of the Website

To build and serve a local copy of the website, run

bundle exec jekyll serve

The website can then be found at http://127.0.0.1:4000

Note that if you are running Ruby 2.7 then you will see lots of deprecation warnings until Jekyll 4.1 is released.

Building HTML files

bundle exec jekyll build

Generated HTML files can be found in _site.

Publications and Citations

The website uses a data file to dynamically display details of FLAME GPU publications and citations respectively. To populate these data files there is a provided python script (citations.py) which uses the scholarly Pythoin package to scrape the details from the web. To run this script use the instructions below.

Note: The script is not run as part of the site build as it will likely fail due to aggressive bot detection

Install dependencies

You can either install the scholarly Python package or do this via an environment manager such as conda. E.g.

conda create --name scholarly --file requirements.txt

Then active your conda environment

activate scholarly

Configuring the FLAME GPU papers

The title of each FLAME GPU publication is hard coded in the citations.py file within the flame_pubs array. Each of these paper titles will have the full publication details retrieved form the web which will be stored in json format within data\publications.yml.

The 'data\citations.yml' file will be populated by retrieving papers which site any of the FLAME GPU publications (e.g. papers in flame_pubs). The citations are then ordered by number of citations as a rough metric of the significance of the paper (more citations is better).

Running the citations script

In order to run the citations scrip (from the activated conda environment) run the following;

python citations.py

Important: The scholarly library uses web scraping and google scholar is very aggressive towards bots. It is highly likely that when you script runs you will get a pop-up browser window to complete a captcha to proceed. Using the scraping code may lead to IP blocking so do so at your own risk (or via a proxy/VPN).