percona-pg-activity

A dashboard of Percona's contributions to PostgreSQL, rebuilt daily from public sources.

Live page: https://percona-lab.github.io/percona-pg-activity/

Four tabs:

• Code by companies (2026, filterable by month range): changes authored and reviewed, company ranking, engineers, areas of PostgreSQL, and every change with Percona credit (summary, roles, PostgreSQL versions, links), exportable as Markdown or CSV.
• Code by people (same range filter, or the current release cycle): everyone credited in PostgreSQL commits, ranked by changes authored + reviewed, committers included by default, Percona engineers highlighted; scatter, top 20 and a sortable, exportable table.
• Events & community (since 2022, by year): talks, organiser / program / Code of Conduct committee roles and volunteering at PostgreSQL events, with slides and recordings; postgresql.org contributor and release-note recognition; Planet PostgreSQL; mailing-list activity from Hackorum; meetups and open source projects.
• Blog posts: PostgreSQL posts on percona.com and percona.community.

How it works

.github/workflows/update.yml runs daily at 05:17 UTC, on every push to main, or manually (Actions → Update PG activity dashboard → Run workflow). It:

1. runs collect_community.py → community.json (HTTP responses are cached between runs with actions/cache, so a warm run only re-fetches current events and feeds),
2. clones PostgreSQL's recent history and runs build_dashboard_data.py → data.json and build_ranking_data.py → ranking.json,
3. runs percona_activity.py --company-statistics --company-chart → chart.svg, stats.md,
4. publishes everything with index.html to GitHub Pages.

File	Purpose
percona_activity.py	Identity and company attribution for commits (COMPANY_DOMAINS, COMPANY_NAME_OVERRIDES, MANUAL_IDENTITIES); also a standalone CLI report
build_dashboard_data.py	Builds data.json: per-change details, versions, areas, categories, monthly company counts
build_ranking_data.py	Builds ranking.json: monthly authored / reviewed / reported counts for every contributor, committers included
collect_community.py	Builds community.json from conference sites, blogs, Planet PostgreSQL, Hackorum, postgresql.org
community/people.json	Percona people to look for at events, with optional since/until dates
community/manual.json	Hand-curated meetups, community days, volunteering and projects no website lists
index.html	The dashboard (plain HTML/JS, no dependencies)

Community sources

Source	What we take
PostgreSQL Europe events (postgresql.eu): PGConf.EU, PGConf.DE, PGConf.PL, FOSDEM PGDay, PGDay Paris, Nordic PGDay, PGDay Lowlands, PGDay Napoli, PGDay Med	talks (schedule.xml), speaker companies, organisers and committees (/organisation/), slides and video links
PgUS (postgresql.us), PGDay UK (pgday.uk), PGConf.dev (pgevents.ca)	same, via the same conference system
Swiss PGDay (pgday.ch)	talks and committees from its own site
PGConf India (pgconf.in)	talks with slides/video, core committee
postgresql.org event list	coverage check: conferences no dedicated collector reads
Radar (every conference from the coverage check, plus community/watch.json)	Indico (CERN PGDay) and eventyay (FOSSASIA PGDay) via their APIs; any other site by looking for people.json names on its schedule / programme / speaker pages, shown as "Listed on the event website". Names found next to "Percona" that aren't in people.json go to review
percona.com/blog, percona.community	PostgreSQL posts (RSS)
Planet PostgreSQL	Percona feeds and recent post count
Hackorum (/team/percona)	mailing-list messages, patches, per year
postgresql.org	contributors page, release-note acknowledgments

How a person is counted at an event: the event lists them with company Percona, or they are in community/people.json, the event is inside their since/until window, and the event does not list another employer. Speaker profiles show a person's current company, so they are only used as evidence for events in the last ~18 months. Entries dropped because a source names another employer are listed under review in community.json.

Volunteering

Volunteer schedules on the conference system (/events/<event>/volunteer/) need a login, so the daily build can't read them. Import them locally with your own session instead:

python3 import_volunteering.py --session 'www.postgresql.eu=<sessionid>' --session 'postgresql.us=<sessionid>'
python3 import_volunteering.py --session 'www.postgresql.eu=<sessionid>' --write   # merge into community/volunteering.csv

<sessionid> is the sessionid cookie of that site while you're logged in (browser developer tools → Cookies); each site (postgresql.eu, postgresql.us, pgday.uk, pgevents.ca) has its own. The script only sees events where you were a volunteer or admin, and lists the Percona people on those events' volunteer lists. Review the CSV diff, then commit. Volunteering at events outside this system (FOSDEM, meetups) goes in community/manual.json.

New conference series on these systems are picked up automatically. To add a different kind of site, add a collector function in collect_community.py.

Updating attribution

Someone missing, or attributed to the wrong company? Edit the mappings in percona_activity.py and push; the page rebuilds within a few minutes. Change types ("Bug fix", "New feature", …) and areas are heuristics from the commit subject and changed paths: see CATEGORY_RULES and AREA_RULES in build_dashboard_data.py.

Running locally

With Task: task serve clones PostgreSQL into .cache/pg, builds site/ and serves it on http://127.0.0.1:8000/ (HOST=0.0.0.0 inside a container, PG=/path/to/postgres to use an existing clone). task build also collects community.json. By hand:

cd /path/to/postgres            # needs refs/remotes/origin/{master,REL_*_STABLE}
python3 /path/to/build_dashboard_data.py --after 2025-12-31 --output /tmp/site/data.json
cp /path/to/index.html /tmp/site/ && python3 -m http.server -d /tmp/site 8000

Requires Python 3.10+, no extra packages.

Roadmap

The page is meant to grow into a single Percona PostgreSQL community dashboard: talks at PostgreSQL events, blog posts (percona.com and percona.community), mailing-list activity, and extension/ecosystem work.