Package router provides a playground that routes code execution requests to
run_lang servers.
Package run_lang provides an API server answering code execution queries.
Package images provides run_lang images for different ecosystems.
To run a playground locally, you need to (1) stand up some run_lang servers
and (2) start the router web server.
Prebuilt images for python, javascript, and rust are available on docker under
the ayazhafiz repository and can be stood up without additional configuration.
For example,
docker run --name runpython --publish 9001:9000 ayazhafiz/runpython:latest
docker run --name runjavascript --publish 9002:9000 ayazhafiz/runjavascript:latest
docker run --name runrust --publish 9003:9000 ayazhafiz/runrust:latestMapping a host port to port 9000 in the container is crucial, as this is the port the container's code execution server listens to.
After standing up the containers, list the addresses by which they can be
accessed from the outside world in routes.sh.
Images for each language ecosystem supported by the playground can be found in
the subdirectories of images. Each such image is derived from the
runlang_base image, which describes a common
configuration needed for running the code execution server found in the
./run_lang directory.
Do not modify the Dockerfiles of each language ecosystem directly; rather,
modify the imagegen.yaml (for example, see rust's),
or create a new one for a new language ecosystem, and then run python3 images/imimagegen.py in the root of this repository. This generates a
Dockerfile for the language ecosystem and a packages file used by the
run_lang server during the container's runtime.
An imagegen.yaml file looks like:
# A description of the language.
description: "python 3.7"
# The command to install packages when building this container.
install_pkg_command: pip3 install
# The packages to install when building this container.
# Must be an array, each entry of which will be appended to `install_pkg_command`
# and then executed with the docker `RUN` directive.
# For example, the first package in this list generates
# RUN pip3 install numpy==1.19.4
ecosystem_pkg:
- numpy==1.19.4
- pendulum==2.1.2
- requests==2.25.0
# Docker commands to run before installing packages.
pre_install_pkg:
- RUN apt-get install -y python3.7 python3-pip
# Docker commands to run after installing packages.
# You may not need this, but the key must be present.
post_install_pkg:
- RUN echo 0You can also use the build_images.sh script to generate
and build all images in one go, and use PUSH=true ./build_images.sh to also
publish the built images to Dockerhub (though you will need to change the
repository organization from ayazhafiz if you don't have write access).
Images are also built and published to Dockerhub under ayazhafiz on push to
this repo's master branch thanks to Travis CI.
First install the dependencies:
pip3 install -r router/requirements.txtThere are two ways to start the server:
./start_router.sh: This will start a server suitable for productionsource routes.sh && FLASK_ENV=development FLASK_APP=router flask run --port 8000: this will start a server for use in development, the main nicety being that updates to files inrouterwill be recognized without needing to turn down the server.
Now you have a functional playground at localhost:8000.