Reproducibility as a Service (RaaS)

A website that automatically creates Docker images from user-uploaded directories containing a computational analysis. This service supports multiple languages including R, Python, and Julia (to varying degrees).

For detailed set up instructions and other information visit the wiki. For a detailed view into the architecture and evaluation of the system refer to my master's thesis at the UBC Library.

Running RaaS with Docker (recommended)

Make sure to "install" RaaS first and set up the environment as specified in the wiki. This should be a simple process of building the environment image first and initializing the RaaS database in a persistent volume.

Navigate to the raas directory in a terminal run

docker-compose up --build

In your browser navigate to localhost:5000, and you are ready to start using RaaS!

For instructions on how to run RaaS locally from source code visit the wiki.

Instructions for using RaaS

From the build image page users can upload a zip file. When choosing a name for the dataset, the name must be in all-lowercase.

Important Notes for Use

• Dataset names should be all-lowercase when uploaded. Docker will return an error if a name is not all lowercase that may not be clear to the user.
• RaaS will attempt to determine and install all necessary language packages and system libraries.
• However, it is not perfect and some may be missed, especially if a script is uploaded that never explicitly calls 'library' or 'import.'

Headless Mode

Included currently in RaaS is a ''headless'' mode. This is a method of interfacing almost directly with the service from the command line. This should be used for debugging and testing purposes only as it bypasses login information and allows someone to just provide a user ID as a parameter. It is also only available from within the host running the service. To use it, run the headless_raas.py script on the same machine a RaaS instance is running. Some dataset must be specificed, either zip or doi, and a dataset name. Optional parameters include port number for containR, whether to preprocess, and a user ID. For more information run python headless_raas.py -h.

How to support another language?

There are 3 steps you need to follow to support a new language:

1.Create a new object that implement the language_interface. Our standard style is to include a new directory in the app folder with the name language_*lang name*, e.g. language_r. The new object should go in a file in this directory, and any other files that might be needed for this language specifically. The methods you would need to implement from the language_interface are

 script_analysis

 build_docker_file

 create_report

The script_analysis function should return a StaticAnalysisResults object as defined in the languageinterface.py script. It should have a list of language packages and list of system libraries to install. Anything else that might be needed when writing the dockerfile should be in the lang_specific variable.

2.In the app/starter.py add an if condition to call your language object

3.In the app/forms.py add your language name to the front end selection box

If you are stuck and confused about any of these steps, check the existing implementations for R and Python as a reference.