vicary 0.2.0 → 0.2.1

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 030a648056336b3ad6905e692be55844a1a3fbffd16b73a5ef37bcfc275402cc
-  data.tar.gz: 3753c57622c1b2ab6dfb82fa12f79eb862681dcc8377a8fe01094e471d235ad7
+  metadata.gz: 94923f9a64b909c5a6ab25e3d7dc1d8caa41843068e1ede77fe938905a1a5745
+  data.tar.gz: 43e9513288654b871a78115c0f90d86c4f51701944dc01e20df469a62991852d
 SHA512:
-  metadata.gz: 37aa4e0e3ccd3bb7860f97125862e1f7b3b30e63594549f201fec0f42bd92e4883e34a4b0c01cffe0f631e46f6576f4dbfc0e56f89e8d30df9e844575b512eb8
-  data.tar.gz: ae700f44fc829bb7431ffd26a20db4e3dc3c3e15ab0b4e047e5c57ff996b1ac701da0d9023ad2de6167b3047fff929469b948068c77421633ba8935a741a616f
+  metadata.gz: 888df67ec5e7842da4d43a9398ea28edcb6552f47c84fada77bc9912e3dcbc4a56f1ea4f2b19c67dc0c5701eefc9fef6b6e671e7b29f2c034c92fc7d5dc0e129
+  data.tar.gz: 650b3ca105ef6dc6801c0232606077bef713f3c84d55668183707c296233c9d6c6c3438e21dbb5d218457735782fb671c5bb97a7fd15d06242059649c46ca549

data/README.md

@@ -1,7 +1,7 @@
 # vicary (Ruby)
 
-The RubyGems front door. **Not published yet** — the gem name is claimed by a
-pending trusted publisher and the first release has not run.
+The RubyGems front door. **Published** — [`vicary` on RubyGems](https://rubygems.org/gems/vicary),
+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](https://github.com/bwthomas/vicary#readme). What lives here is a
@@ -10,7 +10,7 @@ conformance suite in [`conformance/`](../conformance): for every fixture frame i
 must produce **byte-identical output to the Python implementation, placeholder
 numbering included**.
 
-It clears that bar — 36 of 36 masking-required frames, 52 of 52 overall.
+It clears that bar — 38 of 38 masking-required frames, 54 of 54 overall.
 
 ```ruby
 require "vicary"
@@ -31,7 +31,7 @@ Three layers, because each catches what the one above it cannot.
 
 | command | what it says |
 |---|---|
-| `rake conformance` | the scoreboard against the 52 frames — the final bar, and a coarse first one |
+| `rake conformance` | the scoreboard against the 54 frames — the final bar, and a coarse first one |
 | `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 |

data/lib/vicary/conformance.rb

@@ -119,7 +119,7 @@ module Vicary
       # The primitives spec — the layer underneath the frames.
       #
       # `frames.json` scores finished output, which is the right final bar and a
-      # poor first one: a port with nothing implemented scores 0 of 36 and learns
+      # poor first one: a port with nothing implemented scores 0 of 38 and learns
       # nothing about which of the forty-odd primitives underneath is wrong.
       # `primitives.json` is that missing layer, generated from the Python
       # functions and byte-compared against a fresh export by

data/lib/vicary/structured.rb

@@ -19,11 +19,12 @@ module Vicary
   #
   # **Order is the contract, not an optimisation.** The first pattern to claim a
   # span wins, and placeholder indices follow mint order, so reordering these
-  # tables changes the output bytes even when it changes no verdict. Identity
-  # runs first (an address line can otherwise swallow a surname); EMAIL before
-  # PHONE; SSN and CARD before the generic digit runs; ZIP and AGE last, because
-  # both are bare digits and would claim characters belonging to a phone, card or
-  # address.
+  # tables changes the output bytes even when it changes no verdict. EMAIL and URL
+  # run first, because a school-issued address and a profile URL *contain* the
+  # writer's name and both are anchored too tightly to take one out of prose;
+  # identity next (an address line can otherwise swallow a surname); SSN and CARD
+  # before the generic digit runs; ZIP and AGE last, because both are bare digits
+  # and would claim characters belonging to a phone, card or address.
   #
   # ## Regex dialect
   #
@@ -141,13 +142,34 @@ module Vicary
     # Date of birth, explicitly labelled.
     DOB = %r{\b(?:date\s+of\s+birth|d\.?o\.?b\.?|born\s+on)\s*:?\s*\d{1,2}[/-]\d{1,2}[/-]\d{2,4}\b}i
 
-    # (placeholder kind, pattern) in application order.
+    # The two structured patterns that run BEFORE identity interpolation, because
+    # their match text can legitimately *contain* the writer's own name:
+    # `first.last@district.org` and a profile URL ending in a name slug. Identity
+    # interpolation is a literal-name substitution, so running it first left these
+    # shredded rather than masked — `{NAME_2}.{NAME_1}{USERNAME_1}.k12.oh.us`
+    # instead of `{EMAIL_1}`, with the domain tail surviving in the clear and the
+    # span unrestorable on the round trip.
+    #
+    # Putting them first is safe in the direction that matters, and that asymmetry
+    # is the whole argument. Both are anchored on structure a name cannot supply —
+    # EMAIL needs an `@` and a dotted TLD, URL needs a scheme or a `www.` — so
+    # neither can reach into prose and take a bare surname out of it. Patterns
+    # that *could* still run after identity.
+    STRUCTURED_BEFORE_IDENTITY = [
+      ["EMAIL", EMAIL],
+      ["URL", URL_PATTERN],
+    ].freeze
+
+    # (placeholder kind, pattern) in application order, running AFTER identity
+    # interpolation.
     #
     # CARD is handled separately because it needs the Luhn gate; ZIP and AGE run
     # after it for the reason in the module docstring.
+    #
+    # Numbering is unaffected by which of these two tables a pattern sits in: the
+    # minter counts per kind, so `{EMAIL_1}` is the first email whether emails are
+    # matched before or after names.
     STRUCTURED = [
-      ["EMAIL", EMAIL],
-      ["URL", URL_PATTERN],
       ["US_SOCIAL_SECURITY_NUMBER", SSN],
       ["IP_ADDRESS", IP],
       ["PHONE", PHONE],
@@ -286,10 +308,15 @@ module Vicary
         masked = text
         n = 0
 
-        # Identity patterns run FIRST: a name is the span most likely to be
+        # Identity patterns run early: a name is the span most likely to be
         # partially consumed by a looser pattern (an address line can swallow a
         # surname), and masking it first makes that impossible.
-        (identity_patterns(identity) + STRUCTURED).each do |kind, pattern|
+        #
+        # Early, not first. Email and URL precede it, because those two are the
+        # patterns whose own match text contains a name — see
+        # STRUCTURED_BEFORE_IDENTITY for why that direction is the safe one.
+        patterns = STRUCTURED_BEFORE_IDENTITY + identity_patterns(identity) + STRUCTURED
+        patterns.each do |kind, pattern|
           masked, count = minter.substitute(kind, pattern, masked)
           n += count
         end

data/lib/vicary/version.rb

@@ -6,5 +6,5 @@ module Vicary
   # Shared across all three front doors on purpose: one detector, one number. A
   # gem 0.3.0 that corresponds to nothing on PyPI cannot be reasoned about, and
   # the parity claim is between *versions*, not between package names.
-  VERSION = "0.2.0"
+  VERSION = "0.2.1"
 end

data/lib/vicary.rb

@@ -17,7 +17,7 @@
 #
 # Three layers check that claim, and each catches what the one above it cannot:
 #
-# * `rake conformance` scores the 52 frames — the final bar, and a coarse first
+# * `rake conformance` scores the 54 frames — the final bar, and a coarse first
 #   one;
 # * `rake test` runs `test/primitives_test.rb`, forty-odd primitives over the
 #   shared `primitives.json` corpus, which says *which brick is crooked*;

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: vicary
 version: !ruby/object:Gem::Version
-  version: 0.2.0
+  version: 0.2.1
 platform: ruby
 authors:
 - Blake Thomas
 autorequire:
 bindir: bin
 cert_chain: []
-date: 2026-08-11 00:00:00.000000000 Z
+date: 2026-08-12 00:00:00.000000000 Z
 dependencies: []
 description: |
   Finds the names a student writes about — classmates, teachers, relatives — and