Citation: N/A (2018/08/31) Easing the burden of code review. Nature Methods (RSS)
DOI (original publisher): 10.1038/s41592-018-0137-5
Semantic Scholar (metadata): 10.1038/s41592-018-0137-5
Sci-Hub (fulltext): 10.1038/s41592-018-0137-5
Internet Archive Scholar (search for fulltext): Easing the burden of code review
Download: https://doi.org/10.1038/s41592-018-0137-5
Tagged: Computer Science (RSS) computational science (RSS)

Summary

Code review is time-consuming for already overworked reviewers.
Authors packing their code and data in executable containers, tagged with metadata helps ease the burden of setting it up.
{Nature Methods, Nature Biotechnicology, Nature Machine Intelligence} started using CodeOcean platform, which:
- integrates the metadata, code, data and dependencies into a single 'compute capsule'
- displays all inputs and results on a web interface
- stores metadata, discoverable by search engine
- offers 100 hours of free compute for academics
CodeOcean mitigates capsule generation time by:
- including preinstalled dependencies
- import packages from GitHub automatically (why not use language package manager?)
- include files by hand
CodeOcean is limited by:
- No support for multi-CPU, GPU, or other hardware clusters
- Not optimized for large datasets

Easing the burden of code review