QUARRYZ

keeping every version a statistical office published, and failing the build when it quietly issues another.

A statistical office restates the past for years, so a warehouse holding only the latest figure has silently rewritten research that quoted the earlier one. This keeps every version, fails the build when a revision arrives undeclared, and measures what the obvious storage key costs: on 23,943 real published values, keying on the release date destroys 71 of them with no error anywhere. That key was this repository's own recommendation until the measurement was run over the whole corpus instead of one example.

Tests

88

Python

3.11 to 3.13

Release

v0.1.0

Licence

MIT

The demo, as it really ran

$ uv run python examples/what_the_publisher_changed.py

4 series, 535 versions walked, 517 of which changed at least one number, 23,943 recorded changes. Nothing here is fetched.

Every value IKBJ has carried for 2021, and when it said so:

  released    version          value
  2022-02-11  v88             -28761
  2022-04-10  v90             -29107
  2022-10-11  v98             -17572
  2023-01-13  v101            -28039
  2023-10-11  v111             -3518
  2024-10-10  v123             -5119
  2025-10-15  v135               939

  The largest single move is 24,521, on 2023-10-11. The period did not change. The publisher did.

  as at 2023-06       -28039
  as at 2023-12        -3518

What one line of DDL decides, counted over all four series:

  keyed on (series, period)                     keeps  2,533
  keyed on (series, period, vintage as a Date)  keeps 23,872
  keyed on (series, period, vintage, version)   keeps 23,943
  The middle row is not a rounding difference. It is 71 published values, deleted with no error.

Because a release date is not a key. One morning, one period, two published values:

  2016-01-08  v3         -2548   (2015 APR, IKBJ)
  2016-01-08  v4         -2584   (2015 APR, IKBJ)

Output captured on 2026-08-29. It is committed to the repository and a test fails when it stops matching a live run, so this page cannot quietly drift from the code it describes.

Read it

• Source on GitHub the README carries the argument in full
• Release v0.1.0 changelog and tag

The rest of the toolset

• QUACKZ deflating a backtest that only looks good because it was picked out of two hundred
• QUOTEZ market data an agent can read and cannot act on
• QUELLZ measuring what prompt-injection containment costs in utility as well as in attack rate
• QUIDZ refusing the outbound payment that would have gone out twice
• QUESTZ stopping a scraper before it writes a CSV from a page that changed shape
• QUIZZ answering what a statistic said at the time, and refusing when it cannot
• QUARANTINEZ treating an outcome the venue never confirmed as terminal rather than as a retry
• QUENCHZ deciding in the open what a tool server gets free while it is still somebody's subprocess
• QUILTZ proving infrastructure code wrong without a cloud account, and saying what that cannot show
• QUAYZ telling a crash loop from an OOMKill, and naming the failure that no single field finds
• QUASHZ refusing a row whose outcome had not been decided yet when the decision would have been made
• QUALMZ a fixed number of looks at the holdout, where re-running the same configuration does not buy another
• QUEUEZ ordering a feed by its sequence, because on a real recorded session the clock goes backwards
• QUANDARYZ counting the distinct screens a component can settle into when its responses arrive out of order