PyPI Download Statistics (BigQuery)
Python Software Foundation (PyPI)
Every pip install of every Python package, logged and queryable in SQL.
A public BigQuery table recording each file download from the Python Package Index, with the package, version, installer, Python version, operating system, and country, plus a companion table of metadata for every release. It is the best public window into which software libraries people actually use, and how fast they upgrade.
Access
- Cost
- FreeData is free; querying it uses Google BigQuery, which is free up to a monthly quota and billed beyond.
- License
- Open licenseCreative Commons license (per PyPI documentation)
- Sign-up
- Free account
- How you get it
- APIBigQuery
What's in it
- One row is
- One row per file download (file_downloads) or per release file (distribution_metadata)
- Covers
- Global, with country codes
- Years
- 2016-present
- Updated
- Real time
- Size
- Billions of downloads per month
Quality
Complete log of PyPI downloads straight from the source; downloads are not users, and CI systems and mirrors inflate counts.
- Automated CI installs dominate many packages' counts
- Downloads through mirrors and caches are not seen
- Queries over the full table can be expensive
Connects to
- GH Archive: Compare installs with repository stars and contributions
- NVD (National Vulnerability Database) CVE API: See how quickly downloads move off vulnerable versions
Questions it could answer
- How long does it take for half of a library's downloads to move to a new major version?
- Which Python versions still account for meaningful downloads?
- How fast are AI-related packages growing compared with web frameworks?