metanorma-plugin-lutaml 0.7.61 → 0.7.63

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: e10152f9146a2100063d02e244208ed4acc105b8a105b6a15604a14e8522e9ab
-  data.tar.gz: 8c453ee0ba82bc78d20c78e7bf31d34c048874301a12bbd13a63810791db3a8e
+  metadata.gz: 9087a3bdc384a47318affb8b8fc243d2bd7817d3c726b3b5cf27f9c229068a4a
+  data.tar.gz: 7685b187bbf7d47e771b2276d42f62b3a4ed849abdec9ddd2154ff48f393864d
 SHA512:
-  metadata.gz: aa862500cb613b0cfab7ce3816f197a23cab52a44d778fc6cc1945b0e7e48c3a466cec55a1d7601e439256deb9b327e15e840368174896371dadbd457aeb3fb3
-  data.tar.gz: 5cdcea37e490d21fcf555a952c836350570ab8e60499f2a1b98ed54a754af497832a2b749c052a2048a86c0be689f1c2be173e5f0ad74612d1363a4e6ada2cba
+  metadata.gz: cc8c7e181551079b69bb24d41c6bacea72bfa7e64cb6bce41ea3c6bc676f3883f7372a40d97d032e9dbafa0e2fab92c4e3783f280b579eeb15dd1aa675f68067
+  data.tar.gz: 90a1af9e236cf1a855b4042e459951da82d477e9180f0e3192819ef60c9e157e0236e9fc96d2343e30fe70e053154b88ff53f436f0b79b59262d54b86bef513a

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

@@ -17,43 +17,25 @@ module Metanorma
           )
           resolved = document.path_resolver
             .system_path(file_path, docfile_directory)
+          # The decorator expands emitted paths against this base: relative
+          # prefixes alone get re-anchored by asciidoctor's imagesdir
+          # handling (images/../../schemas collapses a level), so the
+          # decoration must produce paths that resolve from the docfile.
+          options["relative_path_base"] =
+            File.expand_path(docfile_directory)
           # The prefix lands inside emitted image macros, so it must not
           # carry the absolute build path: two builds of the same tree in
           # two directories produced different image/@src values
           # (metanorma/metanorma-plugin-lutaml#325). A path relative to the
           # docfile resolves to the same schema file at render time and is
-          # byte-identical across build directories; the decorator joins it
-          # as written rather than expanding it.
-          #
-          # The render resolves image targets against the imagesdir joined
-          # onto the docfile directory, not the bare docfile: a target
-          # relative to the bare docfile is re-anchored through the
-          # imagesdir prefix and every decorated diagram misses by its
-          # depth (metanorma/iso-10303 smol-all FATAL: emitted
-          # `../schemas/…`, resolved as `images/../schemas/…`). Compute the
-          # prefix against the directory the render resolves from.
+          # byte-identical across build directories.
           # The docfile directory may be relative ("."): relative_path_from
           # refuses mixed prefixes, so both sides expand first
           Pathname.new(File.expand_path(resolved))
-            .relative_path_from(
-              Pathname.new(image_resolve_base(document, docfile_directory)),
-            )
+            .relative_path_from(Pathname.new(File.expand_path(docfile_directory)))
             .to_s
         end
 
-        # The directory image targets resolve against: the docfile
-        # directory with the imagesdir joined on. A web or absolute
-        # imagesdir bypasses the join; the bare docfile directory is the
-        # closest approximation.
-        def self.image_resolve_base(document, docfile_directory)
-          base = File.expand_path(docfile_directory)
-          imagesdir = document.attributes["imagesdir"]
-          return base if imagesdir.nil? || imagesdir.empty?
-          return base if imagesdir.match?(%r{\A(?:[a-z]+:|/)})
-
-          File.expand_path(imagesdir, base)
-        end
-
         def self.for(model)
           case model
           when Expressir::Model::Cache

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

@@ -68,15 +68,19 @@ module Metanorma
           return line if path_prefix && line.include?(path_prefix)
 
           line.gsub(RELATIVE_PREFIX_MACRO_REGEXP) do |_match|
-            # The prefix is already docfile-relative (see
-            # ExpressRemarkAdapter.relative_path_prefix), so joining it
-            # keeps the emitted path relative to the docfile too. Expanding
-            # to an absolute path embedded the build directory in the
-            # output, making semantic XML, presentation XML and HTML differ
-            # between two builds of the same tree (#325). Resolution is the
-            # renderer's job: a relative target joined with imagesdir still
-            # resolves against the docfile at render time.
-            "#{$1}#{$2}#{File.join(path_prefix, $3.strip)}#{$4}"
+            # Expand against the docfile directory (passed by the adapter
+            # as relative_path_base), NOT the process CWD: a CWD expansion
+            # produced absolute image paths under whatever directory the
+            # build was launched from (e.g. <repo>/../../schemas/...).
+            # The absolute result is deliberate: relative targets are
+            # re-anchored by asciidoctor's imagesdir handling
+            # (images/../../schemas collapses a level), which made every
+            # decorated diagram unresolvable at render time.
+            full_path = File.expand_path(
+              File.join(path_prefix, $3.strip),
+              options["relative_path_base"] || Dir.pwd,
+            )
+            "#{$1}#{$2}#{full_path}#{$4}"
           end
         end
       end

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 = "0.7.61".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: 0.7.61
+  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