expressir 2.4.1 → 2.4.2

data/lib/expressir/express/remark_attacher.rb

@@ -35,6 +35,10 @@ module Expressir
       # via `child_attributes :foo, :bar, ...`. See TODO.bugs/15.
       EXPRESSION_CHILDREN = Model::ModelElement.child_attributes_registry
 
+      # `WHERE <label> :` clause headers; the captured label maps to the
+      # 1-based line number(s) it appears on.
+      WHERE_CLAUSE_PATTERN = /\A\s*WHERE\s+(\w+)\s*:/i
+
       def initialize(source)
         @source = source
         @attached_spans = Set.new
@@ -71,6 +75,8 @@ module Expressir
         @line_map = nil
         @owner_map = nil
         @active_scope_map = nil
+        @source_lines_for_where_clause = nil
+        @where_clause_line_index = nil
       end
 
       private
@@ -658,27 +664,39 @@ module Expressir
         where_rules = get_collection(scope, :where_rules)
         return nil unless where_rules&.any?
 
-        lines = source_lines_for_where_clause
+        where_clause_lines = where_clause_line_index
 
         where_rules.each do |wr|
           next unless wr.id
 
-          lines.each_with_index do |line, idx|
-            line_num = idx + 1
+          where_clause_lines.fetch(wr.id, []).each do |line_num|
             next unless line_num < remark_line
 
-            if (line =~ /^\s*WHERE\s+#{Regexp.escape(wr.id)}\s*:/i) && remark_line.between?(line_num, line_num + 5)
-              return create_remark_item(wr, tag)
-            end
+            return create_remark_item(wr, tag) if remark_line.between?(line_num, line_num + 5)
           end
         end
 
         nil
       end
 
+      # Single scan over the source lines mapping each `WHERE <id>:` label to
+      # its 1-based line number, so per-remark lookups stop re-testing every
+      # line against every WHERE rule's regex.
+      def where_clause_line_index
+        @where_clause_line_index ||= begin
+          index = Hash.new { |h, k| h[k] = [] }
+          source_lines_for_where_clause.each_with_index do |line, idx|
+            if (match = line.match(WHERE_CLAUSE_PATTERN))
+              index[match[1]] << (idx + 1)
+            end
+          end
+          index
+        end
+      end
+
       def source_lines_for_where_clause
         # @source is set for the duration of `attach`; freed at the end.
-        @source.lines
+        @source_lines_for_where_clause ||= @source.lines
       end
 
       def find_node_in_statement(stmt, tag)

data/lib/expressir/express/schema_block_scanner.rb

@@ -6,8 +6,9 @@ module Expressir
     #
     # A hand-rolled state machine that skips comments (`(* ... *)`) and
     # string literals, tracks SCHEMA/END_SCHEMA depth, and returns one
-    # block per schema declaration. Used by `Parser.from_exp_streaming`
-    # to parse each schema independently with a fresh memory arena.
+    # block per schema declaration. Retained for the future streaming
+    # parse path (see TODO.max-perf/02) that parses each schema
+    # independently with a fresh memory arena.
     #
     # Extracted from the outer Parser class (TODO.bugs/09) so block-
     # scanning logic lives in one focused module behind one interface.

data/lib/expressir/express/scope_resolver.rb

@@ -50,8 +50,13 @@ module Expressir
         @model = model
         @nodes_with_positions = nodes_with_positions
         @scope_map = nil
+        @source_lines = nil
+        @position_buckets = nil
       end
 
+      # Line-band width for the position fallback index.
+      BUCKET_LINES = 1024
+
       # Returns the innermost Model::ScopeContainer whose source span contains
       # the given 1-based remark line, or nil if none is found.
       def containing_scope_for(remark_line)
@@ -69,7 +74,7 @@ module Expressir
         type_state = { line: nil, name: nil }
         rule_state = { line: nil, name: nil }
 
-        @source.lines.each_with_index do |line, idx|
+        source_lines.each_with_index do |line, idx|
           line_num = idx + 1
 
           case line
@@ -131,8 +136,15 @@ module Expressir
         @scope_map ||= build_scope_map
       end
 
+      # Source split into lines, computed once per resolver: both the
+      # scope-map build and find_by_source_text need it, and re-splitting the
+      # whole source per lookup dominates remark attachment on large files.
+      def source_lines
+        @source_lines ||= @source.lines
+      end
+
       def build_scope_map
-        lines = @source.lines
+        lines = source_lines
         map = {}
         return map if lines.empty?
 
@@ -160,10 +172,27 @@ module Expressir
 
       # --- Strategy 2: position-based fallback against the node index ---
 
+      # Nodes bucketed by 1024-line bands: a node spanning [line, end_line]
+      # is registered in every band it overlaps, so a remark-line lookup only
+      # scans nodes that can possibly contain it. Bucket order preserves the
+      # original index order, so select+reverse_each semantics are unchanged.
+      def position_buckets
+        @position_buckets ||= begin
+          buckets = Hash.new { |h, k| h[k] = [] }
+          @nodes_with_positions.each do |n|
+            next unless n[:line] && n[:end_line]
+
+            ((n[:line] / BUCKET_LINES)..(n[:end_line] / BUCKET_LINES)).each do |b|
+              buckets[b] << n
+            end
+          end
+          buckets
+        end
+      end
+
       def find_by_position(remark_line)
-        containing = @nodes_with_positions.select do |n|
-          n[:line] && n[:end_line] &&
-            remark_line >= n[:line] && remark_line <= n[:end_line] &&
+        containing = position_buckets[remark_line / BUCKET_LINES].select do |n|
+          remark_line.between?(n[:line], n[:end_line]) &&
             !n[:node].is_a?(Model::Repository) && !n[:node].is_a?(Model::Cache)
         end
 

data/lib/expressir/express.rb

@@ -2,6 +2,7 @@
 
 module Expressir
   module Express
+    autoload :AstKeyConverter, "#{__dir__}/express/ast_key_converter"
     autoload :Builder, "#{__dir__}/express/builder"
     autoload :BuilderContext, "#{__dir__}/express/builder_context"
     autoload :BuilderRegistry, "#{__dir__}/express/builder_registry"
@@ -17,6 +18,7 @@ module Expressir
     autoload :LineMap, "#{__dir__}/express/line_map"
     autoload :ModelVisitor, "#{__dir__}/express/model_visitor"
     autoload :NodePositionIndex, "#{__dir__}/express/node_position_index"
+    autoload :ParallelFiles, "#{__dir__}/express/parallel_files"
     autoload :Parser, "#{__dir__}/express/parser"
     autoload :PrettyFormatter, "#{__dir__}/express/pretty_formatter"
     autoload :RemarkAttacher, "#{__dir__}/express/remark_attacher"

data/lib/expressir/model/model_element.rb

@@ -155,7 +155,12 @@ module Expressir
       end
 
       def source
-        Expressir::Express::SourceFormatter.format(self)
+        # Formatting is not free and callers probe `source` repeatedly while
+        # walking the tree (position index, remark attachment), so the text is
+        # computed once per node. Deliberately NOT @source: that ivar holds
+        # the raw source span lutaml-model's `source=` stores when
+        # include_source is on, and readers must keep seeing formatted text.
+        @formatted_source ||= Expressir::Express::SourceFormatter.format(self) # rubocop:disable Naming/MemoizedInstanceVariableName
       end
 
       # @param [Hash] options

data/lib/expressir/model/repository.rb

@@ -291,14 +291,27 @@ reference_index: nil)
       # Build repository from list of schema files
       # @param file_paths [Array<String>] Schema file paths
       # @param base_dir [String, nil] Base directory for path resolution
+      # @param max_processes [Integer, nil] parse in a fork worker pool when
+      #   set above 1 on a fork-capable platform (advisory; degrades to
+      #   sequential otherwise). Unlike Express::Parser.from_files, a file
+      #   that fails to parse always raises.
       # @return [Repository] Built repository with all schemas
-      def self.from_files(file_paths, base_dir: nil)
+      def self.from_files(file_paths, base_dir: nil, max_processes: nil)
         repo = new(base_dir: base_dir)
 
-        file_paths.each do |path|
-          parsed = Expressir::Express::Parser.from_file(path)
-          next unless parsed
-
+        pool = Expressir::Express::ParallelFiles
+        parsed_files = if pool.sequential?(file_paths, max_processes)
+                         file_paths.map do |path|
+                           Expressir::Express::Parser.from_file(path)
+                         end
+                       else
+                         pool.run(file_paths,
+                                  parse: ->(path) { Expressir::Express::Parser.from_file(path) },
+                                  max_processes: max_processes,
+                                  strict: true)
+                       end
+
+        parsed_files.each do |parsed|
           repo.files << parsed if parsed.is_a?(ExpFile)
         end
 

data/lib/expressir/version.rb

@@ -3,6 +3,6 @@ module Expressir
   # autoload it (Ruby autoload only works for module/class constants, not
   # for string constants like `Expressir::VERSION`).
   module Version
-    VERSION = "2.4.1".freeze
+    VERSION = "2.4.2".freeze
   end
 end

data/lib/expressir.rb

@@ -21,6 +21,24 @@ else
 end
 
 module Expressir
+  # yeptris mingw builds crash once loaded (`uninitialized constant
+  # Yeptris::FFI::NODE_SCALAR`, leptris/yeptris#318) and the gem rebinds
+  # core YAML on require, so merely having it in the bundle breaks every
+  # Hash#to_yaml on Windows. Pinning the portable adapters stops
+  # lutaml-model's bundle-based autodetection from loading yeptris there;
+  # other platforms keep the native engine. Revisit when yeptris ships a
+  # working mingw build.
+  def self.select_serialization_engines(windows: Gem.win_platform?)
+    return unless windows
+
+    Lutaml::Model::Config.configure do |config|
+      config.yaml_adapter_type = :standard
+      config.json_adapter_type = :standard
+    end
+  end
+
+  select_serialization_engines
+
   # Namespace modules - autoload loads the namespace file which contains
   # autoload definitions for classes within that namespace
   autoload :Config, "#{__dir__}/expressir/config"

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: expressir
 version: !ruby/object:Gem::Version
-  version: 2.4.1
+  version: 2.4.2
 platform: ruby
 authors:
 - Ribose Inc.
 autorequire:
 bindir: exe
 cert_chain: []
-date: 2026-09-17 00:00:00.000000000 Z
+date: 2026-09-18 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: base64
@@ -174,16 +174,16 @@ dependencies:
   name: rubyzip
   requirement: !ruby/object:Gem::Requirement
     requirements:
-    - - "~>"
+    - - ">="
       - !ruby/object:Gem::Version
-        version: '3.4'
+        version: '0'
   type: :runtime
   prerelease: false
   version_requirements: !ruby/object:Gem::Requirement
     requirements:
-    - - "~>"
+    - - ">="
       - !ruby/object:Gem::Version
-        version: '3.4'
+        version: '0'
 - !ruby/object:Gem::Dependency
   name: table_tennis
   requirement: !ruby/object:Gem::Requirement
@@ -282,6 +282,20 @@ files:
 - TODO.bugs/29-validate-ascii-god-class.md
 - TODO.bugs/30-unicode-map-extraction.md
 - TODO.bugs/README.md
+- TODO.max-perf/01-restore-ci-green.md
+- TODO.max-perf/02-streaming-parse-path.md
+- TODO.max-perf/03-cli-parallel-opt-in.md
+- TODO.max-perf/04-benchmark-harness.md
+- TODO.max-perf/05-parallel-fidelity-specs.md
+- TODO.max-perf/06-builder-cpu-audit.md
+- TODO.max-perf/07-upstream-parsanol-roadmap.md
+- TODO.max-perf/08-builder-build-perf.md
+- TODO.max-perf/09-grammar-cold-start.md
+- TODO.max-perf/10-parser-facade-hygiene.md
+- TODO.max-perf/11-ci-green-closeout.md
+- TODO.max-perf/12-require-boot-profile.md
+- TODO.max-perf/13-key-conversion-specs.md
+- TODO.max-perf/14-builder-call-handler-audit.md
 - benchmark/srl_benchmark.rb
 - benchmark/srl_native_benchmark.rb
 - benchmark/srl_ruby_benchmark.rb
@@ -397,6 +411,7 @@ files:
 - lib/expressir/eengine/modified_object.rb
 - lib/expressir/errors.rb
 - lib/expressir/express.rb
+- lib/expressir/express/ast_key_converter.rb
 - lib/expressir/express/builder.rb
 - lib/expressir/express/builder_context.rb
 - lib/expressir/express/builder_registry.rb
@@ -453,6 +468,7 @@ files:
 - lib/expressir/express/line_map.rb
 - lib/expressir/express/model_visitor.rb
 - lib/expressir/express/node_position_index.rb
+- lib/expressir/express/parallel_files.rb
 - lib/expressir/express/parser.rb
 - lib/expressir/express/pretty_formatter.rb
 - lib/expressir/express/remark_attacher.rb