Skip to content

Latest commit

 

History

History
83 lines (49 loc) · 6.02 KB

software-resources.md

File metadata and controls

83 lines (49 loc) · 6.02 KB
Raw

Identifying existing curated resources on software

We can distinguish:

  • Collaborative development platform (e.g. GitHub, GitLab, Bitbucket, ...), this is mostly where the actual software publication and development work takes place today
  • Distribution/package systems (CPAN, CRAN, CTAN, PyPI, Maven Central, NPM, ...)
  • Aggregator for software information (swMath, ASCL service, Papers with code, SciCrunch)
  • Aggregator for software code and package (libraries.io, versioneye, World of Code, SoftwareHeritage)
  • Archive supporting software deposit (Zenodo, SoftwareHeritage, Code Ocean)
  • Tools supporing the collaborative development ecosystem: build/continuous integration service (travis, circleci, Jenkins, GitHub Actions, etc.), test coverage (coverall, codecov, etc.), image system (DockerHub), documentation (readthedoc), etc.
  • Cloud-based computing infrastructure via Jupyter or Jupyter-style Notebooks environment: Google Colab, Azure Notebooks, Kaggle, Amazon Sagemaker, CoCalc, ...

which will be connected somehow to main scholarly services, e.g. for Open Access ones:

  • Scholar article repository (arXiv, bioRxiv, HAL, PMC, ...)
  • Scholar article metadata aggregator/resolver (CrossRef, MAG, WoS, Scopus, ...)
  • Data sharing repository (Zenodo, Dryad, figshare, Harvard Dataverse, WormBase, etc. hundred? others...)
  • Discovery tools (Google Scholar, Semantic Scholar, Primo, Summons, ...)

Resources covering research software

The following lists group resources in term of their level of curation:

Curated research software resources

These resources and service are characterized by the involvement of expert human curation and rely on rich metadata schema. The volume of entries is usually small.

  • swMath: service for software in Mathematics with more than 30.000 curated entries

  • FDI Lab SciCrunch: curated research resources, including software, for identification and citation purposes

  • ASCL service for software in Astrophysics with over 2.000 curated entries

  • TAPoR 3 contains a list of 1500 curated software/tools in Digital Humanities (see https://lehkost.github.io/tools-dh-proceedings/index.html)

  • Wikidata has around 12,000 software entities, excluding video games. The knowlegde base goes beyond software with related entities such as persons (authors, developers), licenses, institutions, etc.

Partially curated research software resources

The metadata descriptions are curated by the authors themselves or crowdsourced. The volume of entries is often larger than the previous category.

  • Zenodo lists some 44.000 software deposits (they are or will be preserved in mirror at some point automatically at SoftwareHeritage). Zenodo metadata is CC-0 license. Software deposits come normally from the Zenodo/GitHub integration. Note that some appears to be dataset generated/used by software (e.g. https://zenodo.org/record/4148730) or documentation on a software (https://zenodo.org/record/4053076). This should be explained by the fact that many GitHub repositories correspond to data or documentation.

In Zenodo, each software version has an independent entry/deposit, but these entries are grouped when related to the same software, e.g. https://zenodo.org/search?page=1&size=20&q=conceptrecid:4070073&all_versions&sort=-version. By default the versions are nicely conflated in the display of search results. There are around 44.000 software and 101.000 versions, given that software and versions have each a distinct DOIs, we are in the range of 150.000 DOIs for software at Zenodo. Finally a REST API and a OAI-PMH service are available for data access.

  • FigShare lists 17K software deposits.

  • Papers with code initiative in Machine Learning, with over 34.000 entries

  • rOpenSci package manager is a community-curated scientific R package environment for build, maintenance, delivery, documentation, etc. (around 350 packages covered)

  • some specialized research software registries: Biocontainers (9.5K tools), Bioconda (7K), Debian Med.

  • Biii is a crowdsourced registry with 1,336 software entries

  • BioTools in Biomedicine (around 17,000 entries), relying on author registration. It covers tools in biomedicine, but also includes databases, services, etc. (note: on covid-19 tools, https://bio.tools/t?domain=covid-19 and it uses keyphrase lookup to find tool mentions in PMC, see https://github.com/bio-tools/pub2tools)

Other minor or inactive resources include CoMSES (800 computational models), mloss.org (for Machine Learning, 674 entries), SBGrid (structural biology, 450 entries), CSDMS portal (Geosciences, 300 entries)

"Raw" research software resources

Some general software service and aggregators providing useful information:

  • libraries.io: aggregator of software package/library objects. Free API requests are subject to a 60/request/minute rate limit based on the api_key. The data is CC-BY-SA, but the tool itself is AGPL.

  • GitHub, Gitlab, Bitbucket

  • General dependency management systems like CRAN, Maven, NPM, etc. can provide interesting metadata (in particular CRAN) and usage information

  • Software Heritage provides in a central place a huge archive of Open source software. World of Code rather offers a sandbox of GitHub mirrors for research studies.

Other

List of programming languages:

A lis of Open Source Sofware licenses: https://libraries.io/licenses