dcc 0.2.0 → 0.2.2

data/lib/dcc/extract/formula/parser.rb

@@ -0,0 +1,305 @@
+# frozen_string_literal: true
+
+require "bigdecimal"
+
+module Dcc
+  module Extract
+    module Formula
+      # Turns an `Mml::V3::Math` model into a `Formula::Ast`.
+      #
+      # mml splits an element's children into one collection per child type
+      # (`apply_value`, `ci_value`, …), which loses order across types.
+      # `element_order` records the original child sequence, so we walk that
+      # and pop from the matching collection with a per-name cursor.
+      module Parser
+        OPERATORS = {
+          "plus" => :+,
+          "minus" => :-,
+          "times" => :*,
+          "divide" => :/,
+          "power" => :**,
+          "root" => :root,
+          "exp" => :exp,
+          "ln" => :ln,
+          "log" => :log,
+          "sin" => :sin,
+          "cos" => :cos,
+          "tan" => :tan,
+          "abs" => :abs,
+        }.freeze
+
+        CONSTANTS = { "pi" => :pi, "exponentiale" => :e }.freeze
+
+        # `<cn>` encodings whose value is a single decimal literal. Every
+        # other type spreads the number across `<sep/>`-separated parts.
+        PLAIN_CN_TYPES = %w[real integer double].freeze
+
+        class << self
+          # @param math [Mml::V3::Math]
+          # @param bindings [Hash{String => Quantity}]
+          # @return [Dcc::Extract::Formula::Ast]
+          def call(math, bindings: {})
+            name, model = only(children(math))
+            return parse_declaration(model, bindings) if name == "declare"
+
+            Ast.new(name: nil, bound_variables: [],
+                    body: node(name, model, []), bindings: bindings)
+          end
+
+          private
+
+          def parse_declaration(declaration, bindings)
+            nodes = children(declaration)
+            check_declaration(nodes)
+            lambda_children = children(fetch(nodes, "lambda"))
+            bound = bound_variable_names(lambda_children)
+            body = expression(without_bvars(lambda_children), bound)
+            Ast.new(name: text_of(nodes, "ci"), bound_variables: bound,
+                    body: body, bindings: bindings)
+          end
+
+          # `fetch` and `text_of` take the first match, so a second
+          # <lambda> or a second name <ci> would vanish the same way a
+          # second top-level expression used to.
+          def check_declaration(nodes)
+            names = nodes.map(&:first)
+            return if names == %w[ci lambda]
+
+            raise ::Dcc::ExtractionError,
+                  "<declare> must hold one <ci> then one <lambda>, " \
+                  "found #{names.empty? ? 'nothing' : names.join(', ')}"
+          end
+
+          def without_bvars(nodes)
+            nodes.reject { |pair| pair.first == "bvar" }
+          end
+
+          def bound_variable_names(nodes)
+            nodes.select { |pair| pair.first == "bvar" }
+              .map { |pair| bound_name(children(pair.last)) }
+          end
+
+          # A <bvar> names one variable. MathML also allows a <degree>
+          # qualifier there, but that spells a derivative and this
+          # evaluator has none to offer, so it is refused rather than
+          # ignored — and the message names what was actually there.
+          def bound_name(contents)
+            names = contents.map(&:first)
+            return text(contents.first.last) if names == %w[ci]
+
+            raise ::Dcc::ExtractionError,
+                  "<bvar> must contain exactly one <ci>, " \
+                  "found #{names.empty? ? 'nothing' : names.join(', ')}"
+          end
+
+          # @param nodes [Array<(String, Object)>] ordered
+          #   [element name, model] pairs.
+          # @return [Object] the single expression node.
+          def expression(nodes, bound = [])
+            name, model = only(nodes)
+            node(name, model, bound)
+          end
+
+          # `<math>` and a `<lambda>` body each hold exactly one
+          # expression. Taking `.first` of several dropped the rest
+          # without a word — the same silent wrong answer this file
+          # refuses everywhere else. (`<bvar>` has its own check, in
+          # `bound_name`, because a qualifier there is not an
+          # expression.)
+          def only(nodes)
+            case nodes.size
+            when 1 then nodes.first
+            when 0
+              raise ::Dcc::ExtractionError,
+                    "dcc:formula contains no formula expression"
+            else
+              raise ::Dcc::ExtractionError,
+                    "dcc:formula contains #{nodes.size} expressions " \
+                    "where exactly one is expected"
+            end
+          end
+
+          def node(name, model, bound)
+            case name
+            when "apply" then apply(model, bound)
+            when "cn" then number_node(model)
+            when "ci" then variable(text(model), bound)
+            else constant_node(name)
+            end
+          end
+
+          # A structured `<cn>` splits its number across text segments
+          # separated by `<sep/>`, and `text` concatenates them, so a
+          # rational `1<sep/>2` would silently read as 12 rather than one
+          # half. A wrong answer is worse than a refusal, so encodings we
+          # do not implement are rejected outright.
+          #
+          # No explicit coercion here — Ast::Number's constructor runs the
+          # text through Quantity.decimal.
+          def number_node(model)
+            check_plain_number(model)
+            literal = text(model)
+            number = Ast::Number.new(value: literal)
+            check_integral(model.type, number.value, literal)
+            number
+          end
+
+          # `<cn type="integer">1.5</cn>` declares one thing and writes
+          # another. Reading it as 1.5 is the same contradiction
+          # `base="8"` is refused for.
+          #
+          # Returns the node explicitly rather than ending on the check:
+          # a check's return value in an operand slot would put nil into
+          # the tree and leak a raw NoMethodError out of `abs`.
+          # Quotes the source lexeme, not `value.to_s("F")`: expanding
+          # the decimal turns `1e-100000` — nine bytes of document —
+          # into a hundred-kilobyte exception message.
+          def check_integral(type, value, literal)
+            return unless type == "integer"
+            return if value.frac.zero?
+
+            raise ::Dcc::ExtractionError,
+                  %(<cn type="integer"> holds a non-integer value: #{literal})
+          end
+
+          def check_plain_number(model)
+            check_base(model)
+            unless model.sep_value.empty?
+              raise ::Dcc::ExtractionError,
+                    "Structured <cn> with <sep/> is not supported"
+            end
+            return if model.type.nil? || PLAIN_CN_TYPES.include?(model.type)
+
+            raise ::Dcc::ExtractionError,
+                  "Unsupported <cn> type: #{model.type}"
+          end
+
+          # `<cn base="8">17</cn>` is octal — fifteen, not seventeen.
+          # Reading any other base as decimal is the same silent wrong
+          # number the `<sep/>` check refuses, so only base 10 is accepted.
+          def check_base(model)
+            return if model.base.nil? || model.base == "10"
+
+            raise ::Dcc::ExtractionError,
+                  "Unsupported <cn> base: #{model.base}"
+          end
+
+          def constant_node(name)
+            constant = CONSTANTS.fetch(name) do
+              raise ::Dcc::ExtractionError,
+                    "Unsupported MathML element: #{name}"
+            end
+            Ast::Constant.new(name: constant)
+          end
+
+          def apply(model, bound)
+            (operator_name,), *operands = children(model)
+            operator = OPERATORS.fetch(operator_name) do
+              raise ::Dcc::ExtractionError,
+                    "Unsupported MathML operator: #{operator_name}"
+            end
+            Ast::Apply.new(
+              operator: operator,
+              operands: operand_nodes(operator, operands, bound),
+            )
+          end
+
+          def operand_nodes(operator, operands, bound)
+            return root_operands(operands, bound) if operator == :root
+
+            operands.map { |name, child| node(name, child, bound) }
+          end
+
+          # MathML puts an optional `<degree>` wrapper before the radicand.
+          # Unwrap it and store operands as [value, degree]. Cardinality is
+          # enforced here: silently taking `.first` of several radicands or
+          # several degrees would corrupt the evaluator's operand ordering.
+          def root_operands(operands, bound)
+            degrees, values =
+              operands.partition { |pair| pair.first == "degree" }
+            check_root_arity(values, degrees)
+
+            name, child = values.first
+            nodes = [node(name, child, bound)]
+            return nodes if degrees.empty?
+
+            nodes + [degree_node(degrees.first.last, bound)]
+          end
+
+          def check_root_arity(values, degrees)
+            unless values.size == 1
+              raise ::Dcc::ExtractionError,
+                    "<root> must have exactly one radicand, " \
+                    "found #{values.size}"
+            end
+            return unless degrees.size > 1
+
+            raise ::Dcc::ExtractionError,
+                  "<root> must have at most one <degree>, " \
+                  "found #{degrees.size}"
+          end
+
+          def degree_node(model, bound)
+            contents = children(model)
+            unless contents.size == 1
+              raise ::Dcc::ExtractionError,
+                    "<degree> must contain exactly one expression, " \
+                    "found #{contents.size}"
+            end
+
+            name, child = contents.first
+            node(name, child, bound)
+          end
+
+          def variable(name, bound)
+            return Ast::BoundVariable.new(name: name) if bound.include?(name)
+
+            Ast::Variable.new(name: name)
+          end
+
+          # Rebuild the original child sequence from `element_order`.
+          # @return [Array<(String, Object)>]
+          def children(model)
+            cursors = ::Hash.new(0)
+            elements = model.element_order.select do |e|
+              e.node_type == :element
+            end
+            elements.map do |element|
+              name = element.name
+              child = collection(model, name)[cursors[name]]
+              cursors[name] += 1
+              [name, child]
+            end
+          end
+
+          # `element_order` can name any MathML element; only the ones mml
+          # models have a `*_value` reader. This is a boundary check against
+          # the gem, not defensive padding.
+          def collection(model, name)
+            reader = :"#{name.tr('-', '_')}_value"
+            unless model.respond_to?(reader)
+              raise ::Dcc::ExtractionError,
+                    "Unsupported MathML element: #{name}"
+            end
+
+            model.public_send(reader)
+          end
+
+          def fetch(nodes, name)
+            nodes.assoc(name)&.last ||
+              raise(::Dcc::ExtractionError,
+                    "Expected a <#{name}> element in dcc:formula")
+          end
+
+          def text_of(nodes, name)
+            text(fetch(nodes, name))
+          end
+
+          def text(model)
+            Array(model.value).join.strip
+          end
+        end
+      end
+    end
+  end
+end

data/lib/dcc/extract/formula/quantity.rb

@@ -0,0 +1,89 @@
+# frozen_string_literal: true
+
+require "bigdecimal"
+
+module Dcc
+  module Extract
+    module Formula
+      # A named numeric binding taken from a `dcc:quantity`. `value` is a
+      # single `BigDecimal` for `si:real` and an array for the `si:*XMLList`
+      # forms. `uncertainty` is part of the shape callers see, but nothing
+      # populates it yet — extraction always passes nil.
+      #
+      # `Quantity.decimal` is the one place untrusted input becomes a
+      # BigDecimal. The constructor runs every value through it, so a
+      # Quantity cannot exist holding a non-finite or non-numeric value.
+      # That invariant is what lets `Evaluator#quantify` trust its contents.
+      Quantity = ::Data.define(:value, :unit, :uncertainty) do
+        # The single coercion boundary for the whole formula pipeline: the
+        # parser's `<cn>` text, the extractor's `si:value` (declared merely
+        # `:string`, so it can hold anything), and the evaluator's overrides
+        # all land here.
+        #
+        # @param input [Object] anything that might be a number.
+        # @return [BigDecimal] finite, always.
+        # @raise [Dcc::ExtractionError] if it is not a finite number.
+        def self.decimal(input)
+          number = input.is_a?(::BigDecimal) ? input : from_text(input)
+          unless number.finite?
+            raise ::Dcc::ExtractionError,
+                  "Formula value is not a finite number: #{input.inspect}"
+          end
+
+          number
+        rescue ArgumentError
+          raise ::Dcc::ExtractionError,
+                "Formula value is not a number: #{input.inspect}"
+        end
+
+        # `BigDecimal("1_000")` quietly reads 1000, but no document writes a
+        # number that way. Malformed text, not a value ten times its face.
+        def self.from_text(input)
+          text = input.to_s
+          raise ::ArgumentError if text.include?("_")
+
+          BigDecimal(text)
+        end
+        private_class_method :from_text
+
+        # Lists are narrowed with `Array.try_convert`, not `is_a?(Array)`:
+        # the D-SI branch hands back a `Dcc::Type::DecimalXmlList::Values`,
+        # which is array-like but is not an Array. `try_convert` also
+        # rejects an object whose `to_ary` breaks the protocol, which a
+        # bare `respond_to?(:to_ary)` would pass through to a NoMethodError.
+        #
+        # `Data` freezes the struct but not what it holds, so the coerced
+        # list is frozen too.
+        def initialize(value:, unit: nil, uncertainty: nil)
+          super(value: Quantity.coerce(value), unit: unit,
+                uncertainty: uncertainty)
+        end
+
+        def self.coerce(value)
+          return decimal(value) unless value.respond_to?(:to_ary)
+
+          list = ::Array.try_convert(value)
+          raise ::Dcc::ExtractionError, broken_list(value) if list.nil?
+
+          list.map { |item| decimal(item) }.freeze
+        rescue ::TypeError
+          raise ::Dcc::ExtractionError, broken_list(value)
+        end
+
+        def self.broken_list(value)
+          "Formula value is not a usable list: #{value.class}"
+        end
+
+        # @return [Array<BigDecimal>]
+        def values
+          list? ? value : [value]
+        end
+
+        # @return [Boolean]
+        def list?
+          value.is_a?(::Array)
+        end
+      end
+    end
+  end
+end

data/lib/dcc/extract/formula.rb

@@ -0,0 +1,110 @@
+# frozen_string_literal: true
+
+require "bigdecimal"
+
+module Dcc
+  module Extract
+    # `Dcc::Extract::Formula` pulls content-MathML formulae out of a parsed
+    # DCC. Each `dcc:formula` that carries a `dcc:mathml` becomes one `Ast`,
+    # bound to the `dcc:quantity` values sitting in the same `dcc:data`
+    # block — that is where PTB puts a formula's parameters and inputs.
+    module Formula
+      autoload :Ast, "dcc/extract/formula/ast"
+      autoload :Evaluator, "dcc/extract/formula/evaluator"
+      autoload :Parser, "dcc/extract/formula/parser"
+      autoload :Quantity, "dcc/extract/formula/quantity"
+
+      class << self
+        # @param dcc [Lutaml::Model::Serializable] a parsed DCC.
+        # @return [Array<Dcc::Extract::Formula::Ast>] in document order.
+        def call(dcc)
+          data_blocks(dcc).flat_map { |data| formulae_in(data) }
+        end
+
+        private
+
+        def formulae_in(data)
+          bindings = bindings_in(data)
+          data.formula.filter_map do |formula|
+            math = formula.mathml&.math
+            Parser.call(math, bindings: bindings) if math
+          end
+        end
+
+        # Parameters live in `dcc:data/dcc:list/dcc:quantity`, but the XSD
+        # also allows a bare `dcc:quantity` directly under `dcc:data`.
+        def bindings_in(data)
+          quantities_in(data).each_with_object({}) do |quantity, bindings|
+            next unless quantity.id
+
+            value = quantity_value(quantity)
+            bindings[quantity.id] = value if value
+          end
+        end
+
+        # `dcc:listType` is recursive, and both it and `dcc:dataType`
+        # expose `quantity` and `list`. Walking only the first level of
+        # lists loses every binding nested below it.
+        def quantities_in(container)
+          container.quantity +
+            container.list.flat_map { |list| quantities_in(list) }
+        end
+
+        # `si:value` is declared `attribute :value, :string` in
+        # `lib/dcc/si/base/real.rb` — completely unvalidated, so a real
+        # document can carry "banana", "Infinity", or nothing at all. Hand
+        # it to Quantity untouched and let the one coercion boundary decide;
+        # calling `BigDecimal` here would leak a raw ArgumentError out of a
+        # public extraction call.
+        #
+        # Both `real` and `real_list_xml_list` are collection attributes on
+        # `Dcc::V3::Quantity`, so both need `.first`.
+        def quantity_value(quantity)
+          real = quantity.real.first
+          return scalar(real) if real
+
+          list = quantity.real_list_xml_list.first
+          list_values(list) if list
+        end
+
+        def scalar(real)
+          Quantity.new(value: real.value, unit: real.unit, uncertainty: nil)
+        end
+
+        def list_values(list)
+          Quantity.new(
+            value: list.value_xml_list,
+            unit: list.unit_xml_list,
+            uncertainty: nil,
+          )
+        end
+
+        # Same shape as `Dcc::Extract::File.each`: walk the model graph and
+        # collect every `dcc:data` node in document order.
+        def data_blocks(dcc)
+          blocks = []
+          walk(dcc, ::Set.new, blocks)
+          blocks
+        end
+
+        # `Set#add?` returns nil when already present, so it doubles as the
+        # cycle guard for the self-referential list types.
+        def walk(node, visited, blocks)
+          return unless node.is_a?(::Lutaml::Model::Serializable)
+          return unless visited.add?(node.object_id)
+
+          blocks << node if data?(node)
+          node.class.attributes.each_key do |attr_name|
+            Array(node.public_send(attr_name)).each do |child|
+              walk(child, visited, blocks)
+            end
+          end
+        end
+
+        def data?(node)
+          node.class.name.end_with?("::Data")
+        end
+      end
+    end
+  end
+end

data/lib/dcc/extract/ring.rb

@@ -25,4 +25,4 @@ module Dcc
       end
     end
   end
-end
+end

data/lib/dcc/extract.rb

@@ -6,4 +6,4 @@ module Dcc
     autoload :Ring, "dcc/extract/ring"
     autoload :Formula, "dcc/extract/formula"
   end
-end
+end

data/lib/dcc/i18n/text_lookup.rb

@@ -18,7 +18,9 @@ module Dcc
         # @param lang [String, nil] requested ISO 639-1 code.
         # @return [String, nil]
         def call(text_obj, dcc: nil, lang: nil)
-          return nil unless text_obj && Dcc::TypeGuards.has_attribute?(text_obj, :content)
+          return nil unless text_obj && Dcc::TypeGuards.has_attribute?(
+            text_obj, :content
+          )
 
           contents = Array(text_obj.content)
           return nil if contents.empty?
@@ -49,10 +51,12 @@ module Dcc
         end
 
         def declared_languages(dcc)
-          return [] unless dcc && Dcc::TypeGuards.has_attribute?(dcc, :administrative_data)
+          return [] unless dcc && Dcc::TypeGuards.has_attribute?(dcc,
+                                                                 :administrative_data)
 
           admin = dcc.administrative_data
-          return [] unless admin && Dcc::TypeGuards.has_attribute?(admin, :core_data)
+          return [] unless admin && Dcc::TypeGuards.has_attribute?(admin,
+                                                                   :core_data)
 
           core = admin.core_data
           return [] unless core
@@ -64,4 +68,4 @@ module Dcc
       end
     end
   end
-end
+end

data/lib/dcc/i18n.rb

@@ -9,4 +9,4 @@ module Dcc
   module I18n
     autoload :TextLookup, "dcc/i18n/text_lookup"
   end
-end
+end

data/lib/dcc/inspect/report.rb

@@ -23,7 +23,7 @@ module Dcc
       def to_s
         lines = []
         lines << "DCC Inspection Report"
-        lines << "=" * 50
+        lines << ("=" * 50)
         lines << "Schema version:     #{schema_version}"
         lines << "Unique identifier:  #{unique_identifier}"
         lines << "Country:            #{country_code}"
@@ -56,7 +56,7 @@ module Dcc
         # @param dcc [Lutaml::Model::Serializable]
         # @return [Dcc::Inspect::Report]
         def call(dcc)
-          admin = admin_data(dcc)
+          admin_data(dcc)
           mr_list = measurement_results(dcc)
           new(
             schema_version: dcc.schema_version.to_s,
@@ -76,42 +76,75 @@ module Dcc
         private
 
         def admin_data(dcc)
-          Dcc::TypeGuards.has_attribute?(dcc, :administrative_data) ? dcc.administrative_data : nil
+          if Dcc::TypeGuards.has_attribute?(dcc,
+                                            :administrative_data)
+            dcc.administrative_data
+          end
         end
 
         def core_data(dcc)
           ad = admin_data(dcc)
-          ad && Dcc::TypeGuards.has_attribute?(ad, :core_data) ? ad.core_data : nil
+          if ad && Dcc::TypeGuards.has_attribute?(ad,
+                                                  :core_data)
+            ad.core_data
+          end
         end
 
         def items(dcc)
           ad = admin_data(dcc)
-          return [] unless ad && Dcc::TypeGuards.has_attribute?(ad, :items) && ad.items && Dcc::TypeGuards.has_attribute?(ad.items, :item)
+          return [] unless ad && Dcc::TypeGuards.has_attribute?(ad,
+                                                                :items) && ad.items && Dcc::TypeGuards.has_attribute?(
+                                                                  ad.items, :item
+                                                                )
 
           Array(ad.items.item)
         end
 
         def measurement_results(dcc)
-          Dcc::TypeGuards.has_attribute?(dcc, :measurement_results) && dcc.measurement_results && Dcc::TypeGuards.has_attribute?(dcc.measurement_results, :measurement_result) ?
-            Array(dcc.measurement_results.measurement_result) : []
+          if Dcc::TypeGuards.has_attribute?(dcc,
+                                            :measurement_results) && dcc.measurement_results && Dcc::TypeGuards.has_attribute?(
+                                              dcc.measurement_results, :measurement_result
+                                            )
+            Array(dcc.measurement_results.measurement_result)
+          else
+            []
+          end
         end
 
         def count_quantities(mr_list)
           mr_list.sum do |mr|
-            next 0 unless Dcc::TypeGuards.has_attribute?(mr, :results) && mr.results && Dcc::TypeGuards.has_attribute?(mr.results, :result)
+            next 0 unless Dcc::TypeGuards.has_attribute?(mr,
+                                                         :results) && mr.results && Dcc::TypeGuards.has_attribute?(
+                                                           mr.results, :result
+                                                         )
 
             Array(mr.results.result).sum do |r|
-              Dcc::TypeGuards.has_attribute?(r, :data) && r.data && Dcc::TypeGuards.has_attribute?(r.data, :quantity) ? Array(r.data.quantity).size : 0
+              if Dcc::TypeGuards.has_attribute?(r,
+                                                :data) && r.data && Dcc::TypeGuards.has_attribute?(r.data,
+                                                                                                   :quantity)
+                Array(r.data.quantity).size
+              else
+                0
+              end
             end
           end
         end
 
         def count_lists(mr_list)
           mr_list.sum do |mr|
-            next 0 unless Dcc::TypeGuards.has_attribute?(mr, :results) && mr.results && Dcc::TypeGuards.has_attribute?(mr.results, :result)
+            next 0 unless Dcc::TypeGuards.has_attribute?(mr,
+                                                         :results) && mr.results && Dcc::TypeGuards.has_attribute?(
+                                                           mr.results, :result
+                                                         )
 
             Array(mr.results.result).sum do |r|
-              Dcc::TypeGuards.has_attribute?(r, :data) && r.data && Dcc::TypeGuards.has_attribute?(r.data, :list) ? Array(r.data.list).size : 0
+              if Dcc::TypeGuards.has_attribute?(r,
+                                                :data) && r.data && Dcc::TypeGuards.has_attribute?(r.data,
+                                                                                                   :list)
+                Array(r.data.list).size
+              else
+                0
+              end
             end
           end
         end
@@ -120,12 +153,16 @@ module Dcc
           cd = core_data(dcc)
           return [] unless cd
 
-          [cd.used_lang_code_iso_639_1, cd.mandatory_lang_code_iso_639_1].flatten.compact.map(&:to_s)
+          [cd.used_lang_code_iso_639_1,
+           cd.mandatory_lang_code_iso_639_1].flatten.compact.map(&:to_s)
         end
 
         def resp_persons(dcc)
           ad = admin_data(dcc)
-          return [] unless ad && Dcc::TypeGuards.has_attribute?(ad, :resp_persons) && ad.resp_persons && Dcc::TypeGuards.has_attribute?(ad.resp_persons, :resp_person)
+          return [] unless ad && Dcc::TypeGuards.has_attribute?(ad,
+                                                                :resp_persons) && ad.resp_persons && Dcc::TypeGuards.has_attribute?(
+                                                                  ad.resp_persons, :resp_person
+                                                                )
 
           Array(ad.resp_persons.resp_person)
         end
@@ -136,4 +173,4 @@ module Dcc
       end
     end
   end
-end
+end

data/lib/dcc/inspect.rb

@@ -6,4 +6,4 @@ module Dcc
   module Inspect
     autoload :Report, "dcc/inspect/report"
   end
-end
+end