Description
As reproducibility gains traction in the data science and research communities, the need to package code, data and the computational environment is growing.
There are many tools that address different aspects of this type of packaging, such as Jupyter Notebooks for literate programming, Docker for containerising and porting computational environments, and so on. But they represent barriers to reproducibility as each one requires time and effort to learn.
Project Binder integrates Notebooks and Docker for generating reproducible computational analyses and combines them with a web-based interface and cloud orchestration engines. This means that analysts do not have to worry about all the moving parts so long as they have followed basic software best practices: their code is version controlled and they've captured the dependencies the analysis needs to run. Binder then hosts the compute in the cloud and makes it easily shareable by providing a unique URL to the code repository, without imposing additional overheads on the analyst.
During this talk, Sarah will introduce Binder (the service), BinderHub (the technological infrastructure) and mybinder.org (a public instance of a Binder service, free for anyone to use) and demonstrate how it can be used to share Python environments and analyses.