vicary (Ruby)

The RubyGems front door. Publishedvicary on RubyGems, via trusted publishing, so no API key lives in this repository.

The detector, the data asset and the measured numbers are described in the project README. What lives here is a port, and the bar it has to clear before it is published is the shared conformance suite in conformance/: for every fixture frame it must produce byte-identical output to the Python implementation, placeholder numbering included.

It clears that bar — 38 of 38 masking-required frames, 54 of 54 overall.

require "vicary"

Identity = Struct.new(:first_name, :last_name, :school_name)
identity = Identity.new("Marguerite", "Delacroix-Whitfield", "Westfield High School")

Vicary.redact("My cousin Terrence Okonkwo came over that summer.", identity)
# => "My cousin {NAME_1} came over that summer."

masked, n, restore_map = Vicary.redact_with_report(essay, identity)
Vicary.restore(masked, restore_map) == essay   # => true

Checking it

Three layers, because each catches what the one above it cannot.

command what it says
rake conformance the scoreboard against the 54 frames — the final bar, and a coarse first one
rake gates the nine gates, all nine measured from what the repository ships
rake test the unit suites, including primitives_test.rb: forty-odd primitives over the shared corpus, which says which brick is crooked
rake parity gazetteer verdicts, name by name, against the Python reference
rake redaction_parity masked bytes against the Python reference, on prose no fixture contains

A sixth gate — bare-surname exposure — this gem measures from the surname table shipped in conformance/census/, reporting 1.20% of US surname bearers, the same figure Python and TypeScript report from the same table. VICARY_EVAL_CENSUS_CSV overrides it with your own Census copy, and must be the extracted Names_2010Census.csv from the census.gov 2010 surnames release: a .zip is refused by name rather than read as text, because Ruby's standard library has no zip reader and a binary read parsed as CSV yields zero rows — a lower exposure than the truth, and the wrong direction to fail in silently. The shipped table is gzip, which zlib reads, so that hazard does not arise on the default path.

The last three need an essay corpus no package here ships, and this gem measures them once VICARY_EVAL_CORPUS_TSV points at one: 100% carrier recall (29/29 held-out REDACT spans) and 0.60 over-fired spans per essay (15 across 25), identical to Python and TypeScript, plus its own latency p95 of 8.8–9.9 ms. That last one is worth knowing about: the bar is ≤ 10 ms, and this port runs nearest it of the three — roughly 4× Python — so the latency gate is a live constraint here rather than a formality. The carrier essays are built from offsets recorded in conformance/carrier.json rather than from a reimplementation of Python's RNG, and the suite asserts their sha256.

The last two need the reference interpreter — run just py-setup from the repository root first.

Why there are four and not one, measured on the day the port landed: of eleven deliberate mutations to candidates.rb, the conformance frames caught one. The primitives spec caught seven. Three were inert. The last was a real divergence that both corpora were blind to, because both are single-line and the rule only differs across a newline — which is what redaction_parity and test/dialect_test.rb exist for.

Porting notes

lib/vicary/candidates.rb opens with the regex-dialect differences between Ruby and Python that run through the detector. The short version: ^ and $ mean line in Ruby and string in Python, so every one of them is written \A, \z or \Z; \w, \d and \s are ASCII-only in Ruby and Unicode-aware in Python; and \b — unlike JavaScript's — already agrees with Python, so the explicit lookarounds here are belt-and-braces rather than load-bearing. test/dialect_test.rb pins all of it in both directions.