Tabb Labs · Public datasets

Data you can query straight from a URL.

Public datasets published by Tabb Labs, served as Parquet files and read over HTTPS by Malloy and DuckDB. They back the Malloyyo work.

Everything here is public data or synthetic.

Read it in one line

No download step

DuckDB reads a Parquet file over HTTPS directly. Point a query at any file URL below.

SELECT * FROM 'https://tabblabs.net/datasets/osrs/servers.parquet' LIMIT 10;

The datasets

04 datasets · 14 files

Old School RuneScape service tickets

Synthetic

Four made-up servers, modelled on the shapes of a real service ledger: refunds, re-posts, cut rates that change over time, clients who get renamed, and tickets that run across several days.

Amounts are in US dollars, converted from in-game gold at $0.765 per million GP (Old School Bond at $8.99 for about 11.75M GP, as of 2026-10-06).

Contains no real people and no real business data
Five related tables: servers, clients, cut rates, helpers, payments

Source: synthetic data published by Tabb Labs.

FileSize

osrs/servers.parquet715 B

osrs/clients.parquet6.8 KB

osrs/cut_rates.parquet615 B

osrs/helpers.parquet582 B

osrs/payments.parquet45.9 KB

US public libraries, FY2018 to FY2024

US public data

Public library systems across the United States from the Public Libraries Survey, fiscal years 2018 to 2024, with FY2023 branches and outlets, a system to county mapping, and county population.

Visits, borrowing and digital use
Programs, spending and funding
library_systems_by_year: 64,730 rows, one per system per year, 9,377 systems
library_counties: library system to county by fiscal year, 67,416 rows
county_population: county totals 2010 to 2025, 50,294 rows
library_systems (FY2023): 9,252 library systems
library_branches (FY2023): 17,586 branches and outlets

Source: Institute of Museum and Library Services (IMLS), Public Libraries Survey, FY2018 to FY2024. US Census Bureau, Population Estimates (public domain). US public data.

FileSize

libraries/library_systems_by_year.parquet4.8 MB

libraries/library_counties.parquet392 KB

libraries/county_population.parquet224 KB

libraries/library_systems.parquet1.4 MB

libraries/library_branches.parquet823 KB

US domestic flights, May to July 2026

US public data

Reporting-carrier on-time performance for US domestic flights, one file per month.

Delays, cancellations and delay causes
Aircraft tail numbers (public registration marks)
1,851,282 flights in total, 611,735 of them in May

Source: Bureau of Transportation Statistics (BTS), Reporting Carrier On-Time Performance. US public data.

FileSize

flights/flights_2026_05.parquet9.7 MB

flights/flights_2026_06.parquet9.7 MB

flights/flights_2026_07.parquet10.2 MB

All three months1,851,282 rows

US transit ridership, 2002 to August 2026

US public data

Monthly ridership for US transit agencies, one row per agency, mode, type of service and month.

Monthly trips
Vehicle revenue miles and hours, peak vehicles
Agencies with city, state, urbanized area and 2020 population
372,801 rows

Source: Federal Transit Administration (FTA), National Transit Database, Complete Monthly Ridership (with Adjustments and Estimates), plus 2024 NTD agency information for names and cities. US public data (Public Domain U.S. Government).

FileSize

transit/ntd_monthly.parquet3.2 MB

Format

Parquet over HTTPS

Each file is a single Parquet table at its own URL under tabblabs.net/datasets/.

Tools

Malloy and DuckDB

Malloy models and DuckDB queries read the files in place, with no copy or import first.

Provenance

Public or synthetic

The library, flight and transit data are US public data from the agencies named above. The service-ticket data is synthetic.