tree_haver 3.2.2 → 3.2.4

data/lib/tree_haver/backend_api.rb

@@ -0,0 +1,349 @@
+# frozen_string_literal: true
+
+module TreeHaver
+  # Backend API contract definitions and validation
+  #
+  # This module defines the expected API surface for TreeHaver backends.
+  # Each backend must provide Parser, Language, Tree, and Node classes/objects
+  # that conform to these interfaces.
+  #
+  # == Architecture
+  #
+  # TreeHaver backends fall into two categories:
+  #
+  # 1. **Raw backends** (MRI, FFI, Rust) - Return raw tree-sitter objects
+  #    (e.g., ::TreeSitter::Node). TreeHaver::Node wraps these and provides
+  #    a unified API via method delegation.
+  #
+  # 2. **Wrapper backends** (Java, Citrus, Prism, Psych, Commonmarker, Markly) -
+  #    Return their own wrapper objects that must implement the expected API
+  #    directly, since TreeHaver::Node will delegate to them.
+  #
+  # == Usage
+  #
+  #   # Validate a backend's API compliance
+  #   TreeHaver::BackendAPI.validate!(backend_module)
+  #
+  #   # Check specific class compliance
+  #   TreeHaver::BackendAPI.validate_node!(node_instance)
+  #
+  module BackendAPI
+    # Required methods for Language class/instances
+    #
+    # All backends MUST implement `from_library` for API consistency.
+    # Language-specific backends (Psych, Prism, Commonmarker, Markly) should
+    # implement `from_library` to accept (and ignore) path/symbol parameters,
+    # returning their single supported language.
+    #
+    # This ensures `TreeHaver.parser_for(:yaml)` works regardless of backend -
+    # tree-sitter backends load the YAML grammar, while Psych returns its
+    # built-in YAML support.
+    #
+    # Convenience methods (yaml, ruby, markdown) are OPTIONAL and only make
+    # sense on backends that only support one language family.
+    LANGUAGE_CLASS_METHODS = %i[
+      from_library
+    ].freeze
+
+    # Optional convenience methods for language-specific backends
+    # These are NOT required - they're just shortcuts for single-language backends
+    LANGUAGE_OPTIONAL_CLASS_METHODS = %i[
+      yaml
+      ruby
+      markdown
+    ].freeze
+
+    LANGUAGE_INSTANCE_METHODS = %i[
+      backend
+    ].freeze
+
+    # Required methods for Parser class/instances
+    PARSER_CLASS_METHODS = %i[
+      new
+    ].freeze
+
+    PARSER_INSTANCE_METHODS = %i[
+      language=
+      parse
+    ].freeze
+
+    # Optional Parser methods (for incremental parsing)
+    PARSER_OPTIONAL_METHODS = %i[
+      parse_string
+    ].freeze
+
+    # Required methods for Tree instances
+    # Note: Tree is returned by Parser#parse, not instantiated directly
+    TREE_INSTANCE_METHODS = %i[
+      root_node
+    ].freeze
+
+    # Optional Tree methods (for incremental parsing)
+    TREE_OPTIONAL_METHODS = %i[
+      edit
+    ].freeze
+
+    # Required methods for Node instances returned by wrapper backends
+    # These are the methods TreeHaver::Node delegates to inner_node
+    #
+    # Raw backends (MRI, FFI, Rust) return tree-sitter native nodes which
+    # have their own API. TreeHaver::Node handles the translation.
+    #
+    # Wrapper backends (Java, Citrus, etc.) must implement these methods
+    # on their Node class since TreeHaver::Node delegates to them.
+    NODE_INSTANCE_METHODS = %i[
+      type
+      child_count
+      child
+      start_byte
+      end_byte
+    ].freeze
+
+    # Optional Node methods - should return nil if not supported
+    NODE_OPTIONAL_METHODS = %i[
+      parent
+      next_sibling
+      prev_sibling
+      named?
+      has_error?
+      missing?
+      text
+      child_by_field_name
+      start_point
+      end_point
+    ].freeze
+
+    # Methods that have common aliases across backends
+    NODE_ALIASES = {
+      type: %i[kind],
+      named?: %i[is_named? is_named],
+      has_error?: %i[has_error],
+      missing?: %i[is_missing? is_missing],
+      next_sibling: %i[next_named_sibling],
+      prev_sibling: %i[previous_sibling previous_named_sibling prev_named_sibling],
+    }.freeze
+
+    class << self
+      # Validate a backend module for API compliance
+      #
+      # @param backend_module [Module] The backend module (e.g., TreeHaver::Backends::Java)
+      # @param strict [Boolean] If true, raise on missing optional methods
+      # @return [Hash] Validation results with :valid, :errors, :warnings keys
+      def validate(backend_module, strict: false)
+        results = {
+          valid: true,
+          errors: [],
+          warnings: [],
+          capabilities: {},
+        }
+
+        # Check module-level methods
+        validate_module_methods(backend_module, results)
+
+        # Check Language class
+        if backend_module.const_defined?(:Language)
+          validate_language(backend_module::Language, results)
+        else
+          results[:errors] << "Missing Language class"
+          results[:valid] = false
+        end
+
+        # Check Parser class
+        if backend_module.const_defined?(:Parser)
+          validate_parser(backend_module::Parser, results)
+        else
+          results[:errors] << "Missing Parser class"
+          results[:valid] = false
+        end
+
+        # Check Tree class if present (some backends return raw trees)
+        if backend_module.const_defined?(:Tree)
+          validate_tree(backend_module::Tree, results)
+        else
+          results[:warnings] << "No Tree class (backend returns raw trees)"
+        end
+
+        # Check Node class if present (wrapper backends)
+        if backend_module.const_defined?(:Node)
+          validate_node_class(backend_module::Node, results, strict: strict)
+        else
+          results[:warnings] << "No Node class (backend returns raw nodes, TreeHaver::Node will wrap)"
+        end
+
+        # Fail on warnings in strict mode
+        if strict && results[:warnings].any?
+          results[:valid] = false
+        end
+
+        results
+      end
+
+      # Validate and raise on failure
+      #
+      # @param backend_module [Module] The backend module to validate
+      # @param strict [Boolean] If true, treat warnings as errors
+      # @raise [TreeHaver::Error] if validation fails
+      # @return [Hash] Validation results if valid
+      def validate!(backend_module, strict: false)
+        results = validate(backend_module, strict: strict)
+        unless results[:valid]
+          raise TreeHaver::Error,
+            "Backend #{backend_module.name} API validation failed:\n  " \
+              "Errors: #{results[:errors].join(", ")}\n  " \
+              "Warnings: #{results[:warnings].join(", ")}"
+        end
+        results
+      end
+
+      # Validate a Node instance for API compliance
+      #
+      # @param node [Object] A node instance to validate
+      # @return [Hash] Validation results
+      def validate_node_instance(node)
+        results = {
+          valid: true,
+          errors: [],
+          warnings: [],
+          supported_methods: [],
+          unsupported_methods: [],
+        }
+
+        # Check required methods
+        NODE_INSTANCE_METHODS.each do |method|
+          if responds_to_with_aliases?(node, method)
+            results[:supported_methods] << method
+          else
+            results[:errors] << "Missing required method: #{method}"
+            results[:valid] = false
+          end
+        end
+
+        # Check optional methods
+        NODE_OPTIONAL_METHODS.each do |method|
+          if responds_to_with_aliases?(node, method)
+            results[:supported_methods] << method
+          else
+            results[:unsupported_methods] << method
+            results[:warnings] << "Missing optional method: #{method}"
+          end
+        end
+
+        results
+      end
+
+      private
+
+      def validate_module_methods(mod, results)
+        unless mod.respond_to?(:available?)
+          results[:errors] << "Missing module method: available?"
+          results[:valid] = false
+        end
+
+        unless mod.respond_to?(:capabilities)
+          results[:warnings] << "Missing module method: capabilities"
+        end
+      end
+
+      def validate_language(klass, results)
+        # from_library is REQUIRED for all backends
+        # Language-specific backends should implement it to ignore path/symbol
+        # and return their single language (for API consistency)
+        unless klass.respond_to?(:from_library)
+          results[:errors] << "Language missing required class method: from_library"
+          results[:valid] = false
+        end
+
+        # Check for optional convenience methods
+        optional_methods = LANGUAGE_OPTIONAL_CLASS_METHODS.select { |m| klass.respond_to?(m) }
+        if optional_methods.any?
+          results[:capabilities][:language_shortcuts] = optional_methods
+        end
+
+        results[:capabilities][:language] = {
+          class_methods: LANGUAGE_CLASS_METHODS.select { |m| klass.respond_to?(m) } +
+            optional_methods,
+        }
+      end
+
+      def validate_parser(klass, results)
+        PARSER_CLASS_METHODS.each do |method|
+          unless klass.respond_to?(method)
+            results[:errors] << "Parser missing class method: #{method}"
+            results[:valid] = false
+          end
+        end
+
+        # Check instance methods by inspecting the class
+        PARSER_INSTANCE_METHODS.each do |method|
+          unless klass.instance_methods.include?(method) || klass.private_instance_methods.include?(method)
+            results[:errors] << "Parser missing instance method: #{method}"
+            results[:valid] = false
+          end
+        end
+
+        PARSER_OPTIONAL_METHODS.each do |method|
+          unless klass.instance_methods.include?(method)
+            results[:warnings] << "Parser missing optional method: #{method}"
+          end
+        end
+      end
+
+      def validate_tree(klass, results)
+        TREE_INSTANCE_METHODS.each do |method|
+          unless klass.instance_methods.include?(method)
+            results[:errors] << "Tree missing instance method: #{method}"
+            results[:valid] = false
+          end
+        end
+
+        TREE_OPTIONAL_METHODS.each do |method|
+          unless klass.instance_methods.include?(method)
+            results[:warnings] << "Tree missing optional method: #{method}"
+          end
+        end
+      end
+
+      def validate_node_class(klass, results, strict: false)
+        NODE_INSTANCE_METHODS.each do |method|
+          unless has_method_or_alias?(klass, method)
+            results[:errors] << "Node missing required method: #{method}"
+            results[:valid] = false
+          end
+        end
+
+        NODE_OPTIONAL_METHODS.each do |method|
+          unless has_method_or_alias?(klass, method)
+            msg = "Node missing optional method: #{method}"
+            if strict
+              results[:errors] << msg
+              results[:valid] = false
+            else
+              results[:warnings] << msg
+            end
+          end
+        end
+
+        results[:capabilities][:node] = {
+          required: NODE_INSTANCE_METHODS.select { |m| has_method_or_alias?(klass, m) },
+          optional: NODE_OPTIONAL_METHODS.select { |m| has_method_or_alias?(klass, m) },
+        }
+      end
+
+      def has_method_or_alias?(klass, method)
+        return true if klass.instance_methods.include?(method)
+
+        # Check aliases
+        aliases = NODE_ALIASES[method] || []
+        aliases.any? { |alt| klass.instance_methods.include?(alt) }
+      end
+
+      def responds_to_with_aliases?(obj, method)
+        return true if obj.respond_to?(method)
+
+        # Check aliases
+        aliases = NODE_ALIASES[method] || []
+        aliases.any? { |alt| obj.respond_to?(alt) }
+      end
+    end
+  end
+end

data/lib/tree_haver/backends/citrus.rb

@@ -131,17 +131,46 @@ module TreeHaver
         # Alias eql? to ==
         alias_method :eql?, :==
 
-        # Not applicable for Citrus (tree-sitter-specific)
+        # Load language from library path (API compatibility)
         #
-        # Citrus grammars are Ruby modules, not shared libraries.
-        # This method exists for API compatibility but will raise an error.
+        # Citrus grammars are Ruby modules, not shared libraries. This method
+        # provides API compatibility with tree-sitter backends by looking up
+        # registered Citrus grammars by name.
         #
-        # @raise [TreeHaver::NotAvailable] always raises
+        # For full API consistency, register a Citrus grammar with:
+        #   TreeHaver.register_language(:toml, grammar_module: TomlRB::Document)
+        #
+        # Then this method will find it when called via `TreeHaver.parser_for(:toml)`.
+        #
+        # @param path [String, nil] Ignored for Citrus (used to derive language name)
+        # @param symbol [String, nil] Used to derive language name if path not provided
+        # @param name [String, Symbol, nil] Language name to look up
+        # @return [Language] Citrus language wrapper
+        # @raise [TreeHaver::NotAvailable] if no Citrus grammar is registered for the language
         class << self
-          def from_library(path, symbol: nil, name: nil)
-            raise TreeHaver::NotAvailable,
-              "Citrus backend doesn't use shared libraries. " \
-                "Use Citrus::Language.new(GrammarModule) instead."
+          def from_library(path = nil, symbol: nil, name: nil)
+            # Derive language name from path, symbol, or explicit name
+            lang_name = name&.to_sym ||
+              symbol&.to_s&.sub(/^tree_sitter_/, "")&.to_sym ||
+              path && TreeHaver::LibraryPathUtils.derive_language_name_from_path(path)&.to_sym
+
+            unless lang_name
+              raise TreeHaver::NotAvailable,
+                "Citrus backend requires a language name. " \
+                  "Provide name: parameter or register a grammar with TreeHaver.register_language."
+            end
+
+            # Look up registered Citrus grammar
+            registration = TreeHaver::LanguageRegistry.registered(lang_name, :citrus)
+
+            unless registration
+              raise TreeHaver::NotAvailable,
+                "No Citrus grammar registered for #{lang_name.inspect}. " \
+                  "Register one with: TreeHaver.register_language(:#{lang_name}, grammar_module: YourGrammar)"
+            end
+
+            grammar_module = registration[:grammar_module]
+            new(grammar_module)
           end
 
           alias_method :from_path, :from_library

data/lib/tree_haver/backends/commonmarker.rb

@@ -102,6 +102,30 @@ module TreeHaver
           def markdown(options: {})
             new(:markdown, options: options)
           end
+
+          # Load language from library path (API compatibility)
+          #
+          # Commonmarker only supports Markdown, so path and symbol parameters are ignored.
+          # This method exists for API consistency with tree-sitter backends,
+          # allowing `TreeHaver.parser_for(:markdown)` to work regardless of backend.
+          #
+          # @param _path [String] Ignored - Commonmarker doesn't load external grammars
+          # @param symbol [String, nil] Ignored
+          # @param name [String, nil] Language name hint (defaults to :markdown)
+          # @return [Language] Markdown language
+          # @raise [TreeHaver::NotAvailable] if requested language is not Markdown
+          def from_library(_path = nil, symbol: nil, name: nil)
+            # Derive language name from symbol if provided
+            lang_name = name || symbol&.to_s&.sub(/^tree_sitter_/, "")&.to_sym || :markdown
+
+            unless lang_name == :markdown
+              raise TreeHaver::NotAvailable,
+                "Commonmarker backend only supports Markdown, not #{lang_name}. " \
+                  "Use a tree-sitter backend for #{lang_name} support."
+            end
+
+            markdown
+          end
         end
 
         # Comparison for sorting/equality