pubid 2.0.0.pre.alpha.12 → 2.0.0.pre.alpha.14

data/lib/pubid/itu/builder.rb

@@ -50,6 +50,16 @@ module Pubid
           return sp
         end
 
+        # Radio Regulations — "ITU-R RR (2020)"
+        if data[:radio_regulations]
+          return Identifiers::RadioRegulations.new(
+            sector: Components::Sector.new(sector: data[:sector].to_s),
+            series: Components::Series.new(series: "RR"),
+            date: data[:year] ? build_date(data) : nil,
+            language: data[:language]&.to_s,
+          )
+        end
+
         # Check if this is a supplement identifier
         if data[:supplement_type]
           supp = build_supplement(data)
@@ -153,11 +163,12 @@ module Pubid
         nil
       end
 
-      # Build Special Publication (OB). Sector is silently dropped — OB is a
-      # cross-bureau publication and `Identifier` rejects sector+OB
-      # in its constructor.
+      # Build Special Publication (OB). The sector of the TSB spelling
+      # ("ITU-T OB.1096") is kept so the bulletin renders back as it was cited;
+      # SpecialPublication#== ignores it, since OB is cross-bureau.
       def build_special_publication(data)
         Identifiers::SpecialPublication.new(
+          sector: (Components::Sector.new(sector: data[:sector].to_s) if data[:sector]),
           series: Components::Series.new(series: "OB"),
           code: data[:number] ? build_code(data) : nil,
           date: data[:year] ? build_date(data) : nil,
@@ -344,10 +355,19 @@ module Pubid
         )
       end
 
+      # The Roman month of a bulletin date ("15.III.2016") is stored as the
+      # two-digit month every other ITU date uses; the day marks the spelling.
       def build_date(data)
+        month = if data[:roman_month]
+                  (Identifiers::SpecialPublication::ROMAN_MONTHS.index(data[:roman_month].to_s) + 1)
+                    .to_s.rjust(2, "0")
+                else
+                  data[:month]&.to_s
+                end
         Pubid::Components::Date.new(
           year: data[:year].to_s,
-          month: data[:month]&.to_s,
+          month: month,
+          day: data[:day]&.to_s,
         )
       end
 
@@ -368,3 +388,5 @@ module Pubid
     end
   end
 end
+
+Pubid::Itu::Builder.prepend(Pubid::Builder::AllPartsWrap)

data/lib/pubid/itu/components/code.rb

@@ -20,6 +20,8 @@ module Pubid
       # +subseries+ (dot-separated, flavor-specific) and +parts+
       # (dash-separated).
       class Code < Lutaml::Model::Serializable
+        include ::Pubid::SubsetMatch
+
         attribute :imp_marker, :string
         attribute :number, :string
         attribute :series_suffix, :string

data/lib/pubid/itu/components/designation.rb

@@ -13,6 +13,8 @@ module Pubid
       #
       # Format: SERIES.CODE  (e.g. "Y.1351", "Y.1362-2")
       class Designation < Lutaml::Model::Serializable
+        include ::Pubid::SubsetMatch
+
         attribute :series, Pubid::Itu::Components::Series
         attribute :code, Pubid::Itu::Components::Code
 

data/lib/pubid/itu/components/sector.rb

@@ -8,6 +8,8 @@ module Pubid
       # ITU Sector component
       # Values: R (Radio), T (Telecommunications), D (Development)
       class Sector < Lutaml::Model::Serializable
+        include ::Pubid::SubsetMatch
+
         attribute :sector, :string # R, T, or D
 
         VALID_SECTORS = %w[R T D].freeze

data/lib/pubid/itu/components/series.rb

@@ -8,6 +8,8 @@ module Pubid
       # ITU Series component
       # Examples: BO, V, X, A, etc.
       class Series < Lutaml::Model::Serializable
+        include ::Pubid::SubsetMatch
+
         attribute :series, :string # e.g., BO, V, X, R, SG1, OB
 
         def to_s

data/lib/pubid/itu/identifiers/base.rb

@@ -15,7 +15,7 @@ module Pubid
           raise Pubid::Errors::InvalidInputError, Pubid::INPUT_TOO_LONG_MESSAGE
         end
 
-        parsed = Parser.parse(normalize_whitespace(identifier))
+        parsed = Pubid::Parg::Backend.parse(:itu, normalize_whitespace(identifier))
         Builder.build(parsed)
       end
 
@@ -101,8 +101,6 @@ module Pubid
         end
 
         super
-
-        validate_ob_no_sector!
       end
 
       # The document number lives on the `code` component for ITU; surface it at
@@ -462,6 +460,16 @@ module Pubid
         model.code ||= Components::Code.new
       end
 
+      # The day is set only by the printed bulletin date ("15.III.2016"), so
+      # it emits only there and no existing index row gains a key.
+      def day_to_kv(model, doc)
+        emit_kv(doc, "day", model.date&.day)
+      end
+
+      def day_from_kv(model, value)
+        date_for(model).day = value.to_s
+      end
+
       def date_for(model)
         model.date ||= Pubid::Components::Date.new
       end
@@ -472,21 +480,6 @@ module Pubid
         str = value.to_s
         LANGUAGES[str] || str
       end
-
-      # OB (Operational Bulletin) is a cross-bureau ITU publication and
-      # must not have a sector. Direct construction with both raises;
-      # the parser silently drops sector for legacy strings like
-      # "ITU-T OB.X" (handled in Builder).
-      def validate_ob_no_sector!
-        return unless series&.series == "OB"
-        return if sector.nil?
-        return if sector.is_a?(Components::Sector) && (sector.sector.nil? || sector.sector.to_s.empty?)
-
-        raise ArgumentError,
-              "OB (Operational Bulletin) is a cross-bureau ITU publication; " \
-              "sector must not be set"
-      end
     end
-
   end
 end

data/lib/pubid/itu/identifiers/radio_regulations.rb

@@ -0,0 +1,27 @@
+# frozen_string_literal: true
+
+module Pubid
+  module Itu
+    module Identifiers
+      # The ITU Radio Regulations — the treaty text revised by each World
+      # Radiocommunication Conference.
+      # Format: ITU-R RR [(YYYY)]
+      # Example: ITU-R RR (2020)
+      #
+      # It has no document number: "RR" is the whole designation, stored as
+      # the series. `#number` returns it, so `root.number` (the relaton-index
+      # key) is not empty.
+      class RadioRegulations < Identifier
+        include StandardSerialization
+
+        def number
+          series&.series
+        end
+
+        def render_base(**_opts)
+          "#{publisher}-#{sector} #{series}#{render_date_suffix}"
+        end
+      end
+    end
+  end
+end

data/lib/pubid/itu/identifiers/special_publication.rb

@@ -4,26 +4,60 @@ module Pubid
   module Itu
     module Identifiers
       # ITU Special Publication — currently models the Operational Bulletin (OB).
-      # OB is a cross-bureau publication (no sector) rendered as
-      # "ITU OB No. {number}" with optional date.
-      #
-      # Pattern: "ITU OB No. 1283 (01/2024)"
+      # OB is a cross-bureau publication. ITU prints it without a sector
+      # ("ITU OB No. 1283 (01/2024)"); the TSB spelling carries one
+      # ("ITU-T OB.1096 (2016)"), which is kept and rendered back. The sector
+      # is not part of the identity: `==`, the URN and the MR slug ignore it,
+      # so both spellings name one bulletin.
       class SpecialPublication < Identifier
         include StandardSerialization
 
+        ROMAN_MONTHS = %w[I II III IV V VI VII VIII IX X XI XII].freeze
+
         def render_base(**_opts)
           number = code&.number
-          result = "#{publisher} #{series} No. #{number}"
-
-          if date
-            result += if date.month
-                        " (#{date.month.to_s.rjust(2, '0')}/#{date.year})"
-                      else
-                        " (#{date.year})"
-                      end
-          end
+          result = if sector
+                     "#{publisher}-#{sector} #{series}.#{number}"
+                   else
+                     "#{publisher} #{series} No. #{number}"
+                   end
+
+          result + render_ob_date
+        end
+
+        # Cross-bureau: the sector is how one bureau cites the bulletin, not
+        # which bulletin it is.
+        def ==(other)
+          return false unless other.instance_of?(self.class)
+
+          series == other.series &&
+            code == other.code &&
+            date == other.date &&
+            language == other.language &&
+            common_text_twin == other.common_text_twin
+        end
+
+        # Keep the MR slug sector-free, like `==` (it was always so, because
+        # the sector used to be dropped).
+        def mr_type
+          nil
+        end
 
-          result
+        private
+
+        # " (MM/YYYY)" or " (YYYY)"; the day-bearing date is the printed
+        # bulletin form " - 15.III.2016", the only spelling that sets a day.
+        def render_ob_date
+          return "" unless date
+
+          if date.day && date.month
+            roman = ROMAN_MONTHS[date.month.to_i - 1]
+            " - #{date.day.to_s.rjust(2, '0')}.#{roman}.#{date.year}"
+          elsif date.month
+            " (#{date.month.to_s.rjust(2, '0')}/#{date.year})"
+          else
+            " (#{date.year})"
+          end
         end
       end
     end

data/lib/pubid/itu/identifiers/standard_serialization.rb

@@ -46,6 +46,8 @@ module Pubid
                 with: { to: :year_to_kv, from: :year_from_kv }
             map "month",
                 with: { to: :month_to_kv, from: :month_from_kv }
+            map "day",
+                with: { to: :day_to_kv, from: :day_from_kv }
             map "language", to: :language
             map "common_text_twin",
                 with: { to: :common_text_twin_to_kv,

data/lib/pubid/itu/identifiers/supplement.rb

@@ -131,6 +131,21 @@ module Pubid
         # Builder#build_supplement) and are deliberately not serialized (see
         # supplement_sector_to_kv), so comparing them would make a parsed
         # identifier unequal to the same identifier rebuilt via from_hash.
+        # A subset match follows the same rules as `==` below: the two
+        # rendering flags are not identity, and when a base is present the
+        # sector/series/code copies of the base are not serialized.
+        SUBSET_BASE_COPIES = %i[sector series code series_word].freeze
+
+        def self.subset_ignored_attributes
+          %i[number_glued slash_joined]
+        end
+
+        def subset_attribute_match?(name, mine, theirs)
+          return true if base && SUBSET_BASE_COPIES.include?(name)
+
+          super
+        end
+
         def ==(other)
           return false unless other.instance_of?(self.class)
 

data/lib/pubid/itu/identifiers.rb

@@ -16,6 +16,7 @@ module Pubid
       autoload :Errata, "#{__dir__}/identifiers/errata"
       autoload :Handbook, "#{__dir__}/identifiers/handbook"
       autoload :Question, "#{__dir__}/identifiers/question"
+      autoload :RadioRegulations, "#{__dir__}/identifiers/radio_regulations"
       autoload :Recommendation, "#{__dir__}/identifiers/recommendation"
       autoload :Report, "#{__dir__}/identifiers/report"
       autoload :SpecialPublication, "#{__dir__}/identifiers/special_publication"

data/lib/pubid/itu/parser.rb

@@ -125,9 +125,48 @@ module Pubid
         dash >> (letter.repeat(1, 3) >> dot >> digits).as(:range_end)
       end
 
-      # Parts
+      # A "-YYYYMM" approval date — the "200307" of "T-REC-T.4-200307-I" and
+      # "ITU-T T.4-200307". ITU's own edition suffix is short ("-5"), so six
+      # digits that read as a plausible year (19xx/20xx) and month (01-12) are
+      # the date, never a part. A six-digit run that fails either test
+      # ("-200313", "-180001") is still a part, as it was before.
+      rule(:yyyymm_year) { (str("19") | str("20")) >> digit >> digit }
+      rule(:yyyymm_month) do
+        (str("0") >> match["1-9"]) | (str("1") >> match["0-2"])
+      end
+      rule(:yyyymm_shape) { yyyymm_year >> yyyymm_month >> digit.absent? }
+
+      # The status letter that trails the date in a publication id — "I" (in
+      # force) or "S" (superseded). It names the state of the edition, not the
+      # edition, so it is parsed and dropped. "S" is also the Spanish language
+      # suffix, so it is a status ONLY in the full "T-REC-…" id, where ITU
+      # always writes one; after an "ITU-T …-YYYYMM" print form only "I" is,
+      # and "-S" stays the language ("ITU-T Z.100-199911-S").
+      rule(:id_status) do
+        dash >> match["IS"] >> match["A-Za-z0-9"].absent?
+      end
+
+      rule(:print_id_status) do
+        dash >> str("I") >> match["A-Za-z0-9"].absent?
+      end
+
+      rule(:yyyymm_date) do
+        dash >> yyyymm_year.as(:year) >> yyyymm_month.as(:month) >>
+          digit.absent?
+      end
+
+      rule(:id_date) { yyyymm_date >> print_id_status.maybe }
+
+      # Either date spelling of a Recommendation.
+      rule(:document_date) { date_part | id_date }
+
+      # "ITU-T REC T.4", "ITU-T REC-T.4" — the redundant type word of ITU's
+      # own URLs. Not captured: a Recommendation is the default type.
+      rule(:rec_word) { str("REC") >> (space | dash) }
+
+      # Parts. The yyyymm guard keeps the approval date out of the part list.
       rule(:part) do
-        dash >> digits.as(:part)
+        dash >> yyyymm_shape.absent? >> digits.as(:part)
       end
 
       rule(:parts) { part.repeat(0).as(:parts) }
@@ -273,6 +312,7 @@ module Pubid
         itu_prefix >>
           sector >>
           space >>
+          rec_word.maybe >>
           series >> dot >>
           code >>
           range_end.maybe >>
@@ -281,7 +321,7 @@ module Pubid
           series_word.maybe >>
           attachment.maybe >>
           version_part.maybe >>
-          date_part.maybe
+          document_date.maybe
       end
 
       rule(:base_without_series) do
@@ -292,7 +332,7 @@ module Pubid
           code_suffixes >>
           attachment.maybe >>
           version_part.maybe >>
-          date_part.maybe
+          document_date.maybe
       end
 
       # A series-code document — "EMC-5", "MES-2", "QOS-2", "IMPL-8",
@@ -311,12 +351,10 @@ module Pubid
       # The number stays in `code.number`, so `root.number` — the field
       # relaton-index bsearches on — is "5" for EMC-5 and "QKD" for SEC-QKD
       # rather than nil.
-      # The OB guard keeps "ITU-T OB-1" a clean parse failure. Without it the
-      # string reaches Builder#build's Recommendation fallback, whose
-      # validate_ob_no_sector! raises an ArgumentError that escapes
-      # Identifier.parse's Parslet::ParseFailed rescue — turning a rejected
-      # input into a crash for callers. It guards "OB" + dash specifically, so
-      # a genuine two-letter mnemonic starting "OB" would still parse.
+      # The OB guard keeps "ITU-T OB-1" a clean parse failure rather than a
+      # Recommendation of a series "OB" — the Operational Bulletin's series
+      # name. It guards "OB" + dash specifically, so a genuine two-letter
+      # mnemonic starting "OB" would still parse.
       rule(:series_code_body) do
         (str("OB") >> dash).absent? >>
           letter.repeat(2).as(:series) >> dash.as(:series_dash) >>
@@ -365,10 +403,8 @@ module Pubid
       # Builder#build's `combined` branch, which builds a CombinedIdentifier and
       # would silently drop the marker — a clean parse failure is better than a
       # Report that comes back as a Recommendation. The OB guard mirrors
-      # series_code_body's: an Operational Bulletin is cross-bureau, and
-      # "Report ITU-T OB.1" would otherwise route to SpecialPublication (marker
-      # dropped) or hit validate_ob_no_sector!, whose ArgumentError escapes
-      # Identifier.parse's Parslet::ParseFailed rescue.
+      # series_code_body's: "Report ITU-T OB.1" would otherwise route to
+      # SpecialPublication with the marker dropped.
       rule(:report_body) do
         (str("OB") >> dot).absent? >>
           (series >> dot).maybe >>
@@ -535,6 +571,7 @@ module Pubid
         itu_prefix >>
           sector >>
           space >>
+          rec_word.maybe >>
           series >> dot >>
           code >>
           range_end.maybe >>
@@ -543,7 +580,7 @@ module Pubid
           series_word.maybe >>
           attachment.maybe >>
           version_part.maybe >>
-          date_part.maybe >>
+          document_date.maybe >>
           language.maybe
       end
 
@@ -556,24 +593,71 @@ module Pubid
           code_suffixes >>
           attachment.maybe >>
           version_part.maybe >>
-          date_part.maybe >>
+          document_date.maybe >>
+          language.maybe
+      end
+
+      # ITU's publication id — "T-REC-T.4-200307-I",
+      # "R-REC-BO.1130-5-202602-I": <sector>-REC-<number>[-<edition>]-<YYYYMM>
+      # [-<status>], the name ITU gives each edition in its URLs and PDF
+      # files. It builds the plain Recommendation it names and renders in the
+      # print form ("ITU-T T.4 (07/2003)"). The date is required: without it
+      # the string names no edition. No other rule starts with a bare sector
+      # letter, so the slot is free.
+      rule(:publication_id) do
+        sector >> dash >> str("REC") >> dash >>
+          series >> dot >> code >> yyyymm_date >> id_status.maybe >>
           language.maybe
       end
 
+      # The Radio Regulations — "ITU-R RR", "ITU-R RR (2020)", and the URL
+      # spelling "ITU-R RR-2020". Always ITU-R. The trailing any.absent? is
+      # load-bearing: PEG ordered choice never re-enters the alternation once
+      # an alternative succeeds, so a partial match on "ITU-R RR.1" must fail
+      # here and fall through to with_series.
+      rule(:radio_regulations) do
+        itu_prefix >> str("R").as(:sector) >> space >>
+          str("RR").as(:radio_regulations) >>
+          (date_part | (dash >> digit.repeat(4, 4).as(:year))).maybe >>
+          language.maybe >> any.absent?
+      end
+
       # OB (Operational Bulletin) — Special Publication.
-      # OB is a cross-bureau ITU publication; sector, when present in legacy
-      # strings like "ITU-T OB.1096", is silently dropped by the builder.
+      # OB is a cross-bureau ITU publication. The TSB spelling carries a
+      # sector ("ITU-T OB.1096 (2016)"); the builder keeps it, and it renders
+      # back, but it is not part of the bulletin's identity.
       rule(:ob_series) { str("OB").as(:series) }
 
       rule(:ob_dot_body) { dot >> number }
       rule(:ob_no_body) { space >> str("No.") >> space >> number }
+      # "ITU OB 1000" — metanorma-itu's docidentifier ("Annex to ITU OB %").
+      # Accepted as an input spelling only; it renders "ITU OB No. 1000", the
+      # form ITU's own bulletin site uses.
+      rule(:ob_bare_body) { space >> number }
+
+      # The date as a bulletin prints it — "ITU-T OB.1096 - 15.III.2016": day,
+      # Roman month, year. The months are tried longest first, because PEG
+      # takes the first alternative that matches and "I" would otherwise win
+      # on "III"; "XIII" matches "XII", then fails on the required dot.
+      rule(:roman_month) do
+        %w[XII XI X IX VIII VII VI V IV III II I]
+          .map { |m| str(m) }.reduce(:|)
+      end
+
+      rule(:ob_roman_date) do
+        str(" - ") >> digit.repeat(2, 2).as(:day) >> dot >>
+          roman_month.as(:roman_month) >> dot >>
+          digit.repeat(4, 4).as(:year)
+      end
+
+      rule(:ob_date) { date_part | ob_roman_date }
 
       rule(:ob_with_sector) do
         itu_prefix >>
           (sector >> space).maybe >>
           ob_series >>
-          (ob_dot_body | ob_no_body) >>
-          date_part.maybe >>
+          (ob_dot_body | ob_no_body | ob_bare_body) >>
+          ob_date.maybe >>
           language.maybe
       end
 
@@ -585,7 +669,7 @@ module Pubid
           str("Operational Bulletin").as(:_op_bull) >>
           space >> str("No.") >> space >>
           number >>
-          date_part.maybe >>
+          ob_date.maybe >>
           language.maybe
       end
 
@@ -671,13 +755,15 @@ module Pubid
           handbook |
           numeric_question |
           letter_question |
+          radio_regulations |
           with_series |
           contribution |
           # Unreachable earlier: special_publication needs the literal "OB",
           # handbook/numeric_question need leading digits, letter_question
           # needs series >> dot, and contribution needs "-C" after the series.
           series_code_identifier |
-          without_series
+          without_series |
+          publication_id
       end
 
       # Common-text form: an ITU identifier followed by "| ISO/IEC ...".

data/lib/pubid/itu/urn_generator.rb

@@ -16,7 +16,10 @@ module Pubid
       def generate_base_urn
         parts = ["urn", "itu"]
 
-        if identifier.sector
+        # An Operational Bulletin is cross-bureau: its sector is a spelling,
+        # not identity (see SpecialPublication#==), so it stays out of the URN.
+        if identifier.sector &&
+            !identifier.is_a?(Identifiers::SpecialPublication)
           sector = identifier.sector.to_s
           parts << sector.to_s.downcase
         else
@@ -56,7 +59,11 @@ module Pubid
 
         if identifier.date
           date = identifier.date
-          if date&.year && date.month
+          # Only an Operational Bulletin's printed date carries a day
+          # ("15.III.2016"); it is in `==`, so it must reach the URN too.
+          if date&.year && date.month && date.day
+            parts << "#{date.day}/#{date.month}/#{date.year}"
+          elsif date&.year && date.month
             parts << "#{date.month}/#{date.year}"
           elsif date&.year
             parts << date.year.to_s

data/lib/pubid/jcgm/CLAUDE.md

@@ -0,0 +1,7 @@
+# JCGM flavor notes
+
+JCGM meetings and URN parsing.
+
+These notes were part of the root `CLAUDE.md`. Read them before you change `lib/pubid/jcgm/` or `spec/pubid/jcgm/`. The root file keeps the cross-flavor contract that every flavor obeys.
+
+- **JCGM meeting year is optional (`JCGM 11st Meeting`)**: a meeting's trailing ` (YYYY)` group is a *separable* component, not part of its identity — JCGM numbers its meetings in one sequence, so the ordinal alone names the event. `rule(:meeting_identifier)` wraps the **whole** group in `.maybe` (not the year inside it, so a dangling `(`, an empty `()` or a truncated `(200)` still fails), mirroring `date_portion.maybe` in `rule(:base)`; `render_meeting` appends the group only when `date` is present. No builder or serialization change was needed: `Builder::Base#assign_attributes` iterates the parsed hash, so an absent `:date` key is simply never assigned, and `SingleIdentifier#emit_kv` already drops a nil value — a dateless meeting is `{_type, number}` and a dated one is byte-identical to before, so **no index migration follows**. **The load-bearing third surface is `UrnParser#parse_meeting_urn`**: `UrnGenerator#generate_meeting_urn` had *always* guarded the date and emitted the short `urn:jcgm:meeting:11`, but the parser rebuilt the ungrammatical `"JCGM 11st Meeting ()"` from it — so the dateless form must be made optional on the URN read-back too, or the round-trip the relaton index depends on breaks. **The whole `UrnParser` was converted to build identifiers directly from the URN segments** (the BIPM pattern; 5 of 38 flavors construct directly, the rest use the shared `flavor_parse` helper, which stays for them). Rendering a string and re-parsing it flattens the segments and silently loses whatever the string form cannot carry, which is what three **pre-existing** defects were: `urn:jcgm:gum.6:2020` **raised** (`"JCGM gum.6:2020"` is not a grammar form); the language segment was dropped (`urn:jcgm:100:2008:en` → `JCGM 100:2008`); and `number, year = parts` discarded every segment past the second, so a supplement decayed into **the standard it amends** — `urn:jcgm:200:2008:corrigendum` → `JCGM 200:2008`, a different document, with no error. The supplement marker (`corrigendum`/`amendment`, written by the generator as the `type_code`) splits the segments: everything before it is the base document, everything after is the supplement's own number and date. Four things the string round-trip used to supply for free are now explicit, and each is load-bearing: (1) **`typed_stage` comes from `Jcgm.locate_stage`, the same registry lookup `Builder#locate_typed_stage` uses — NOT from the attribute default** — because `SingleIdentifier.published_typed_stage` additionally sets `original_abbr`, which would make a URN-reconstructed id **unequal** to a parsed one; the marker is mapped to the abbreviation the registry indexes (`amendment` → `Amd`, since `locate_stage` matches `abbr`, not `type_code`), and the class then comes from `Jcgm.locate_type(typed_stage.type_code)` so one lookup fixes both. **Guide and GumGuide deliberately take the default instead**, because the grammar emits no type token for them and `Builder#build` fills theirs from `published_typed_stage` too — matching the builder means matching *which* of the two paths it took. (2) The meeting number is normalized with **`to_i`**, matching `Identifiers::Meeting.ordinal`, so `011` still reads back as `11` and a missing or non-numeric segment as `0`. (3) The year is **validated against the grammar's `19xx|20xx`** and rejected with `Pubid::UrnParser::Errors::ParseError` (the easc/gost convention) — with no re-parse, nothing else would catch a malformed segment. (4) A language code is mapped back through the inverse of `Builder::Base::LANG_CHAR_MAP` to restore `original_code`, which is what the renderer prints (`en,fr` → `(E/F)`). **KNOWN GAP, generator-side and deliberately unchanged**: a full date renders as its year alone (`JCGM GUM-1:2022-11-28` → `urn:jcgm:gum.1:2022`), so a URN cannot restore month and day; widening it would change already-published URNs. 30 of the 32 corpus ids now round-trip through the URN byte-exactly, and those 2 are the truncated-date pair. (That `original_abbr` asymmetry is **since fixed** — see `lib/pubid/iec/CLAUDE.md` — so the default now agrees too and the lookup mirrors the builder for clarity rather than necessity.) **Deliberately not a typed error**: the earlier reading of this defect was to raise on a nil date; an undated meeting reference is unambiguous and legitimate, and raising would leave relaton's `Bib::ItemData#to_most_recent_reference` with nothing to return. **Do not add the English teens exception to `Identifiers::Meeting.ordinal`** — the real records print `11st`/`12nd`/`13rd`, and three published documents round-trip through that naive rule. Locked by `spec/pubid/jcgm/meeting_partial_spec.rb` plus a bare-form row in `urn_parser_spec.rb`. **Still open, separate call**: `Meeting.new(number: "11").to_s` raises `NoMethodError` (`number` is a `Components::Code`; a bare String is accepted at construction and only explodes at render), and `Meeting#ordinal` renders `"0th"` for a nil `number`. (hand-off: jcgm-meeting-render-nil-date.)

data/lib/pubid/jcgm/builder.rb

@@ -104,3 +104,5 @@ module Pubid
     end
   end
 end
+
+Pubid::Jcgm::Builder.prepend(Pubid::Builder::AllPartsWrap)

data/lib/pubid/jcgm/components/publisher.rb

@@ -6,6 +6,8 @@ module Pubid
   module Jcgm
     module Components
       class Publisher < Lutaml::Model::Serializable
+        include ::Pubid::SubsetMatch
+
         attribute :publisher, :string
 
         def to_s

data/lib/pubid/jcgm.rb

@@ -35,7 +35,7 @@ module Pubid
       parser = Parser.new
       builder = Builder.new
 
-      parsed = parser.parse(identifier)
+      parsed = Pubid::Parg::Backend.parse(:jcgm, identifier)
       builder.build(parsed)
     end
 

data/lib/pubid/jis/builder.rb

@@ -20,6 +20,9 @@ module Pubid
                        build_single_identifier(data)
                      end
         attach_symbol(identifier, data)
+        # "(all parts)" names every part of the document, so it wraps the
+        # document, which holds no mark itself.
+        data[:all_parts] ? identifier.to_all_parts : identifier
       end
 
       private
@@ -37,7 +40,6 @@ module Pubid
           parts: extract_part_strings(data[:parts]),
           year: data[:year]&.to_i,
           language: data[:language]&.to_s,
-          all_parts: (true if data[:all_parts]),
           reaffirmed: (true if data[:reaffirmed]),
         }
 
@@ -114,3 +116,5 @@ module Pubid
     end
   end
 end
+
+Pubid::Jis::Builder.prepend(Pubid::Builder::AllPartsWrap)