metanorma-plugin-lutaml 0.7.62 → 0.7.63

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 102085caacd4951ba78acd47b10eecec2fd50c6646270bd15c70444d4a54bb3e
-  data.tar.gz: 42939df5ca71fe3e838277864054f215d5b3432310792d75384983eb4a35b7cb
+  metadata.gz: 9087a3bdc384a47318affb8b8fc243d2bd7817d3c726b3b5cf27f9c229068a4a
+  data.tar.gz: 7685b187bbf7d47e771b2276d42f62b3a4ed849abdec9ddd2154ff48f393864d
 SHA512:
-  metadata.gz: 550a516ed78801a0a5359a114edac8258bc1c94e0f72384a2d0f338ab38c8476f729de6fb2d4876c21729569badbdfdce17b13e7a6bd0d4164a1a860b319c388
-  data.tar.gz: a641b9efc125a5fce7432c43dd1e49fd673abef9f8991e73a3e182f5a44e7a987e40a741164e8be75f9198497d9063e29d44972a1a500f9cc1c90573d7fe9472
+  metadata.gz: cc8c7e181551079b69bb24d41c6bacea72bfa7e64cb6bce41ea3c6bc676f3883f7372a40d97d032e9dbafa0e2fab92c4e3783f280b579eeb15dd1aa675f68067
+  data.tar.gz: 90a1af9e236cf1a855b4042e459951da82d477e9180f0e3192819ef60c9e157e0236e9fc96d2343e30fe70e053154b88ff53f436f0b79b59262d54b86bef513a

data/lib/metanorma/plugin/lutaml/express_repository_cache.rb

@@ -0,0 +1,90 @@
+# frozen_string_literal: true
+
+require "digest"
+require "expressir"
+
+module Metanorma
+  module Plugin
+    module Lutaml
+      # Process-wide, content-keyed store of parsed Expressir repositories.
+      #
+      # A Metanorma collection compiles many documents against the same
+      # schema set, and every document compile used to re-parse the whole
+      # EXPRESS corpus (measured on ISO 10303-104: 125 schemas, 8.29 MB,
+      # ~12.5 s wall and ~1.1 GB of allocations per document). This cache
+      # parses each distinct source once per process: the key is derived
+      # from the schema sources' content digests, so an edited source
+      # produces a new key and re-parses instead of serving a stale
+      # repository.
+      #
+      # Deliberately NOT registered in CacheRegistry: the metanorma
+      # toolchain flushes CacheRegistry after every document compile to
+      # release the render-time XMI/XSD graphs, and a flush here would
+      # restore the per-document re-parse this cache exists to remove
+      # (metanorma/metanorma-plugin-lutaml#341). The repositories are
+      # collection-lifetime state; the LRU bound below caps how many
+      # corpora stay resident.
+      class ExpressRepositoryCache
+        # Key digest version: bump when the key composition changes so
+        # keys from different compositions can never collide.
+        KEY_VERSION = "metanorma-plugin-lutaml/express-repository-cache/v1"
+
+        MAX_SIZE = 4
+
+        class << self
+          # Content key for a set of EXPRESS source files: the key version,
+          # the Expressir version, then every file's path and content
+          # digest, in the order given (the order shapes the repository,
+          # and the parsed model embeds file paths, so both belong in the
+          # key). A changed source yields a new key.
+          def source_key(paths)
+            digest = Digest::SHA256.new
+            digest.update(KEY_VERSION)
+            digest.update(Expressir::Version::VERSION.to_s)
+            paths.each do |path|
+              digest.update(path)
+              digest.update(Digest::SHA256.file(path).digest)
+            end
+            digest.hexdigest
+          end
+
+          # Identity key for one prebuilt file (a compiled-set artifact or
+          # an Expressir cache file): expanded path plus mtime and size.
+          # The file is not read for the key — compiled sets hydrate lazily
+          # and re-reading a whole artifact per document would defeat the
+          # sharing; path + mtime + size detects a rebuilt file.
+          def file_key(path)
+            [
+              KEY_VERSION,
+              File.expand_path(path),
+              File.mtime(path).to_f,
+              File.size(path),
+            ]
+          end
+
+          def fetch_or_store(key, &block)
+            store.fetch_or_store(key, &block)
+          end
+
+          def fetch(key)
+            store.fetch(key)
+          end
+
+          def clear
+            store.clear
+          end
+
+          def size
+            store.size
+          end
+
+          private
+
+          def store
+            @store ||= CacheStore.new(max_size: MAX_SIZE)
+          end
+        end
+      end
+    end
+  end
+end

data/lib/metanorma/plugin/lutaml/lutaml_preprocessor.rb

@@ -27,10 +27,16 @@ module Metanorma
         def load_lutaml_file(document, file_path, _options)
           full_path = Utils.relative_file_path(document, file_path)
 
-          file = File.new(full_path, encoding: "UTF-8")
           if full_path.end_with?(".exp")
-            ::Expressir::Express::Parser.from_file(file.path)
+            # Shared per process (ExpressRepositoryCache): a collection's
+            # documents re-parse the same schema file otherwise.
+            ExpressRepositoryCache.fetch_or_store(
+              ExpressRepositoryCache.source_key([full_path]),
+            ) do
+              ::Expressir::Express::Parser.from_file(full_path)
+            end
           else
+            file = File.new(full_path, encoding: "UTF-8")
             ::Lutaml::Lml::Parser.parse(file)
           end
         end

data/lib/metanorma/plugin/lutaml/utils.rb

@@ -200,7 +200,11 @@ module Metanorma
         end
 
         def load_express_repo_from_cache(path)
-          ::Expressir::Express::Cache.from_file(path)
+          ExpressRepositoryCache.fetch_or_store(
+            ExpressRepositoryCache.file_key(path),
+          ) do
+            ::Expressir::Express::Cache.from_file(path)
+          end
         end
 
         def save_express_repo_to_cache(path, repository, document)
@@ -225,14 +229,32 @@ module Metanorma
         # Compiled-set artifacts (EXSCS1) hydrate per file on first
         # touch: rendering one schema's page costs one hydration, not
         # the whole repository. The artifact already carries content
-        # digests, so the expressir Cache layer does not apply.
+        # digests, so the expressir Cache layer does not apply. The
+        # repository handle itself is shared per process, so a second
+        # document compile reuses the first one's hydrated files.
         def load_express_from_artifact(path)
-          Expressir::Express::LazyRepository.new(path.to_s)
+          ExpressRepositoryCache.fetch_or_store(
+            ExpressRepositoryCache.file_key(path),
+          ) do
+            Expressir::Express::LazyRepository.new(path.to_s)
+          end
         end
 
         def load_express_from_folder(folder)
-          file_paths = Dir["#{folder}/*.exp"]
-          ::Expressir::Express::Parser.from_files(file_paths)
+          parse_express_files(Dir["#{folder}/*.exp"])
+        end
+
+        # Parse a set of EXPRESS sources once per process: the parsed
+        # repository is shared through ExpressRepositoryCache, keyed by
+        # the sources' content digests, so the documents of a collection
+        # reuse the first compile's parse instead of each re-parsing the
+        # whole schema set (#341).
+        def parse_express_files(file_paths)
+          ExpressRepositoryCache.fetch_or_store(
+            ExpressRepositoryCache.source_key(file_paths),
+          ) do
+            ::Expressir::Express::Parser.from_files(file_paths)
+          end
         end
 
         # TODO: Refactor this using Suma::SchemaConfig
@@ -252,7 +274,7 @@ module Metanorma
             (schema_yaml_base_path + schema_path).cleanpath.to_s
           end
 
-          ::Expressir::Express::Parser.from_files(file_paths)
+          parse_express_files(file_paths)
         end
 
         def parse_document_express_indexes(document, input_lines) # rubocop:disable Metrics/AbcSize,Metrics/CyclomaticComplexity,Metrics/MethodLength,Metrics/PerceivedComplexity

data/lib/metanorma/plugin/lutaml/version.rb

@@ -1,7 +1,7 @@
 module Metanorma
   module Plugin
     module Lutaml
-      VERSION = "0.7.62".freeze
+      VERSION = "0.7.63".freeze
     end
   end
 end

data/lib/metanorma-plugin-lutaml.rb

@@ -24,6 +24,8 @@ module Metanorma
                "metanorma/plugin/lutaml/express_remarks_decorator"
       autoload :ExpressRemarkAdapter,
                "metanorma/plugin/lutaml/express_remark_adapter"
+      autoload :ExpressRepositoryCache,
+               "metanorma/plugin/lutaml/express_repository_cache"
       autoload :FileNotFoundError,
                "metanorma/plugin/lutaml/file_not_found_error"
       autoload :Json2TextPreprocessor,

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: metanorma-plugin-lutaml
 version: !ruby/object:Gem::Version
-  version: 0.7.62
+  version: 0.7.63
 platform: ruby
 authors:
 - Ribose Inc.
 autorequire:
 bindir: exe
 cert_chain: []
-date: 2026-11-03 00:00:00.000000000 Z
+date: 2026-11-05 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: asciidoctor
@@ -464,6 +464,7 @@ files:
 - lib/metanorma/plugin/lutaml/data2_text_preprocessor.rb
 - lib/metanorma/plugin/lutaml/express_remark_adapter.rb
 - lib/metanorma/plugin/lutaml/express_remarks_decorator.rb
+- lib/metanorma/plugin/lutaml/express_repository_cache.rb
 - lib/metanorma/plugin/lutaml/file_not_found_error.rb
 - lib/metanorma/plugin/lutaml/gc_budget.rb
 - lib/metanorma/plugin/lutaml/json2_text_preprocessor.rb