Trino reports

Trino reports

Reports about the Trino project, published at trinodb.github.io/reports: what needs attention now, what to expect when you contribute, how CI is doing, and who the community is. They’re generated every day from GitHub, Git, and Slack data.

Page For Queries
Overview everyone: what needs attention, the project in numbers, the last month sql/overview
Contributing contributors: what to expect after opening a pull request sql/contributing
Maintaining maintainers: pull requests waiting for a decision, the review work sql/maintaining
CI/CD the merge queue and nightly runs, durations, runner time sql/ci-cd
Tests broken and flaky tests, new failures, causes of failed jobs sql/tests
Community growth, regions, companies, releases, issues, Slack sql/community
Clients and tools the other repositories: open pull requests and issues, reviewers sql/repositories
Airlift libraries Trino is built on: open pull requests and issues, reviewers sql/airlift
Hall of fame new contributors, milestones, top authors, reviewers, and mergers sql/hall-of-fame

Something wrong, or missing? Open an issue, or change a report: every section of a page is one SQL file.

How it works

Everything is SQL, run by Trino. Two small Java programs only run the SQL files, in order:

GitHub and Slack APIs -> trino-openapi connector -> sync/Sync.java runs sync/github/*.sql  -> Iceberg tables in S3
                                                    every 2 hours, only what changed

Iceberg tables + trinodb/trino Git repository   -> sql/Reports.java runs sql/<page>/*.sql -> reports/<page>/index.md
                         (trino-git connector)      daily, printed with the Trino CLI         -> GitHub Pages
  • The sync copies data from the APIs with the trino-openapi connector, which turns OpenAPI specs into tables. Every step is a MERGE from an API table into an Iceberg table. Sync.java needs only the Trino JDBC driver. See how the sync works.
  • The reports are queries over those tables, and over the Git history read by the trino-git connector. Reports.java runs them with the client of the Trino CLI, renders their results as tables, and the workflow commits the pages, which GitHub Pages publishes with Jekyll.
  • There’s no database server. The tables are files in S3, and the workflows compact them.

Find your way around

Path What’s there
sql/<page>/ the queries of one page, one file for every section, in the order of the page
sql/setup/ tables and functions that all reports use, like pr_link and sparkline
sql/pages.properties the pages and their titles
sql/Reports.java generates the pages
sync/ the sync program, its steps for GitHub and Slack, the API specs, and table care
bin/ scripts to start Trino, run the sync, and generate the pages
catalog/, etc/ the configuration of Trino and of its catalogs
.github/workflows/ the schedules of the sync, the reports, and the table maintenance
reports/ the generated pages; edit only index.md

The workflows run the same scripts from bin/ as you do locally, with the versions in bin/versions.env.

Change or add a report

Add a .sql file to the directory of a page, or change one:

  • The first line is a comment with the title. The next comment lines are the description: what the report shows, how to read it, and what to do about it.
  • The result is a table. If the file has the word chart anywhere, like in a column title, it’s printed in a monospace font instead, so characters line up between rows. Draw with the functions from sql/setup/01-functions.sql, like hbar, sparkline, stacked, and shade.
  • Links are values of link(label, url), or of pr_link, user_link, and section_link, which links to a report on another page by its title. Mix them with text using || and plain(text), and join a list of them with join_links.
  • Leave out bots with is_bot, and read pull requests, their events, CI runs, and contributors from the tables built in sql/setup, so all pages count the same things.

Try it locally, as described below, and open a pull request. The pages are generated again when it’s merged.

To add a page, add its directory to sql/pages.properties, to nav_pages in _config.yml, and to reports/index.md.

Run it locally

You need Docker and Java 25. The tables are in a private S3 bucket, so build a copy of recent history on your disk, with any GitHub token, and generate a page from it:

export GITHUB_TOKEN=$(gh auth token)
./bin/run-trino.sh --local git github
export SYNC_TRINO_URL=jdbc:trino://localhost:8080
# a few weeks of trinodb/trino, without CI; 2 days take about 4 minutes, see sync/README.md
SYNC_STEPS_DIR=sync/github SYNC_OWNER=trinodb SYNC_REPO=trino SYNC_SINCE=2026-10-15 \
    SYNC_STEPS=tables,commits,issues,issue_comments,pulls,pull_commits,pull_stats,review_comments,reviews,users \
    ./bin/sync.sh
# empty Slack tables, which some reports read
SYNC_STEPS_DIR=sync/slack SYNC_STEPS=tables ./bin/sync.sh
./bin/reports.sh overview

The page is in reports/overview/index.md, and reports about longer periods show only what the copy has. Run a query with docker exec -it trino trino --catalog trinocicd --schema github, and stop Trino with docker rm -f trino.

With access to the AWS account, start Trino on the real tables instead, with the AWS credentials and region in the environment, and generate all pages:

export AWS_REGION=eu-central-1 AWS_ACCESS_KEY_ID AWS_SECRET_ACCESS_KEY AWS_SESSION_TOKEN
./bin/run-trino.sh git
./bin/reports.sh

Design notes

  • The reports read copies of the data, not the APIs, because the APIs limit requests, and keep only recent data: GitHub returns jobs only for runs from the last 2 months, and the free Slack plan shows only the last 90 days.
  • The copies are Iceberg tables in S3, because they’re cheap and need no server.
  • The sync doesn’t use materialized views, because every API needs its own way of reading only what changed.