PFHub ReImplementation For Fair Data Collection

Daniel Wheeler

2022-05-11

OVERView

• Using GitHub workflow (extreme FAIRness)
  • Examples from other communities
    • Nixpkgs
    • Conda-Forge
    • NIST Code Portal
• Schema Tools
  • The CodeMeta Project
  • ASDF - Advanced Scientific Data Format
  • Boutiques
• Python-PFHub (FAIRer than JS)
• PFHub example submission using issue templates

Nixpkgs

• Nixpkgs is a large community
  • 100k packages
  • 5k issues
  • 3.1k PRs
  • 4k contributers
• Completely GitHub based
• Single repository
• 100s of CI workflows
• Takes submissions from 100s of users everyday that require human interaction

Nixpkgs

Example submission on Nixpkgs

• Checks to ensure compliance
• Automated assignment

Nixpkgs

Example submission on Nixpkgs

• Checks to ensure compliance
• Automated assignment
• Human interaction

Nixpkgs

Example submission on Nixpkgs

• Checks to ensure compliance
• Automated assignment
• Human interaction
• Human CIs
• Automated CIs

ASIDE: USING NIX for DATA workflow

• Nix used to orchestrate data workflows
• Not just build workflows
• Functional storage
• Completely reproducible

USING nix in PFHub

• PFHub uses Nix for all builds
• Python-pfhub will also have Pip and Conda builds
• Uses Cachix to cache builds
• All CI builds with Nix

Conda-Forge

• Different model to Nixpkgs
  • 16k repositories!!!
  • Users merge (not admins)
• but all GitHub based

NIST Code portal

• Akin to PFHub
  • Jekyll frontend
  • CMS-free
• builds NIST's code.json daily
• All GitHub Actions

SCHEMA: codemeta Project

• Simple standard for code metadata (from science community)
• Includes: 6 basic categories of data (software, discoverability, development, run-time, versions, other)
• Plan to use this with PFHub
• Metadata builder tools include web, cli, python

SCHEMA: ASDF

• Settled on YAML in 2015 for astronomical data (many other choices)
• ASCII and binary data in same file
  • include simple editable data files
  • supports compressed Numpy arrays
• Number of readers available
• No standard data model
• See "ASDF: A new data format for astronomy", Greenfield et al.

SCHEMA: Boutiques

• Not a workflow language
• Formal command line description
• Specify inputs and outputs

Boutiques output for "echo" command

PYTHON-PFHUB

• Python package that deals with all data transformations and aggregations
• All data exported as Pandas dataframes
• Easy for others to augment, develop, change
• Everything in Python
  • Plotly has improved Python support
• Everything working outside of website setting

Python-PFHUB

PYTHON-PFHUB

pFHUb Submissions

1. Started with simple YAML file
   • Fill out YAML file by hand
   • Submit pull-request
   • CIs + human checks in pull-request
2. Next iteration included an upload form
   • Fill out sophisticated form
   • Submit and Staticman app submits pull-request
3. Currently working on using GitHub issue template
   • Issue template is a simple form (not sophisticated)
   • On submission launches GitHub Action (parses form and submits pull-request)
4. CLI tool?

DEMO

https://github.com/usnistgov/pfhub/issues/new/choose

Discussion

• Upload mechanism (CLI?, GitHub issue templates?)
• Schema
• File type (ASDF?)