module Puppet::Pops
module Loader
# =ModuleLoaders
# A ModuleLoader loads items from a single module.
# The ModuleLoaders (ruby) module contains various such loaders. There is currently one concrete
# implementation, ModuleLoaders::FileBased that loads content from the file system.
# Other implementations can be created - if they are based on name to path mapping where the path
# is relative to a root path, they can derive the base behavior from the ModuleLoaders::AbstractPathBasedModuleLoader class.
#
# Examples of such extensions could be a zip/jar/compressed file base loader.
#
# Notably, a ModuleLoader does not configure itself - it is given the information it needs (the root, its name etc.)
# Logic higher up in the loader hierarchy of things makes decisions based on the "shape of modules", and "available
# modules" to determine which module loader to use for each individual module. (There could be differences in
# internal layout etc.)
#
# A module loader is also not aware of the mapping of name to relative paths.
#
# @api private
#
module ModuleLoaders

  # Wildcard module name for module loaders, makes loading possible from any namespace.
  NAMESPACE_WILDCARD = '*'.freeze

  # This is exactly the same as the #system_loader_from method, but the argument for path is changed to
  # location where pluginsync stores functions. It also accepts definitions in any namespace since pluginsync
  # places all of them in the same directory.
  #
  def self.cached_loader_from(parent_loader, loaders)
    LibRootedFileBased.new(parent_loader,
      loaders,
      NAMESPACE_WILDCARD,
      Puppet[:libdir],
      'cached_puppet_lib',
      [:func_4x, :func_3x, :datatype]
    )
  end

  def self.system_loader_from(parent_loader, loaders)
    # Puppet system may be installed in a fixed location via RPM, installed as a Gem, via source etc.
    # The only way to find this across the different ways puppet can be installed is
    # to search up the path from this source file's __FILE__ location until it finds the base of
    # puppet.
    #
    puppet_lib = File.realpath(File.join(File.dirname(__FILE__), '../../..'))
    LibRootedFileBased.new(parent_loader,
                                                       loaders,
                                                       nil,
                                                       puppet_lib,   # may or may not have a 'lib' above 'puppet'
                                                       'puppet_system',
                                                        [:func_4x, :func_3x, :datatype]   # only load ruby functions and types from "puppet"
                                                       )
  end

  def self.environment_loader_from(parent_loader, loaders, env_path)
    if env_path.nil? || env_path.empty?
      EmptyLoader.new(parent_loader, ENVIRONMENT)
    else
      FileBased.new(parent_loader,
        loaders,
        ENVIRONMENT,
        env_path,
        ENVIRONMENT
      )
    end
  end

  def self.module_loader_from(parent_loader, loaders, module_name, module_path)
    ModuleLoaders::FileBased.new(parent_loader,
                                                       loaders,
                                                       module_name,
                                                       module_path,
                                                       module_name
                                                       )
  end

  def self.pcore_resource_type_loader_from(parent_loader, loaders, environment_path)
    ModuleLoaders::FileBased.new(parent_loader,
      loaders,
      nil,
      environment_path,
      'pcore_resource_types'
    )
  end

  class EmptyLoader < BaseLoader
    def find(typed_name)
      return nil
    end

    def private_loader
      @private_loader ||= self
    end

    def private_loader=(loader)
      @private_loader = loader
    end
  end

  class AbstractPathBasedModuleLoader < BaseLoader

    # The name of the module, or nil, if this is a global "component", or "any module" if set to the `NAMESPACE_WILDCARD` (*)
    attr_reader :module_name


    # The path to the location of the module/component - semantics determined by subclass
    attr_reader :path

    # A map of type to smart-paths that help with minimizing the number of paths to scan
    attr_reader :smart_paths

    # A Module Loader has a private loader, it is lazily obtained on request to provide the visibility
    # for entities contained in the module. Since a ModuleLoader also represents an environment and it is
    # created a different way, this loader can be set explicitly by the loaders bootstrap logic.
    #
    # @api private
    attr_accessor :private_loader

    # Initialize a kind of ModuleLoader for one module
    # @param parent_loader [Loader] loader with higher priority
    # @param loaders [Loaders] the container for this loader
    # @param module_name [String] the name of the module (non qualified name), may be nil for a global "component"
    # @param path [String] the path to the root of the module (semantics defined by subclass)
    # @param loader_name [String] a name that is used for human identification (useful when module_name is nil)
    #
    def initialize(parent_loader, loaders, module_name, path, loader_name, loadables)
      super parent_loader, loader_name

      raise ArgumentError, 'path based loader cannot be instantiated without a path' if path.nil? || path.empty?

      @module_name = module_name
      @path = path
      @smart_paths = LoaderPaths::SmartPaths.new(self)
      @loaders = loaders
      @loadables = loadables
      unless (loadables - LOADABLE_KINDS).empty?
        #TRANSLATORS 'loadables' is a variable containing loadable modules and should not be translated
        raise ArgumentError, _('given loadables are not of supported loadable kind')
      end
      loaders.add_loader_by_name(self)
    end

    def loadables
      @loadables
    end

    def discover(type, error_collector = nil, name_authority = Pcore::RUNTIME_NAME_AUTHORITY, &block)
      global = global?
      if name_authority == Pcore::RUNTIME_NAME_AUTHORITY
        smart_paths.effective_paths(type).each do |sp|
          relative_paths(sp).each do |rp|
            tp = sp.typed_name(type, name_authority, rp, global ? nil : @module_name)
            next unless sp.valid_name?(tp)
            begin
              load_typed(tp) unless block_given? && !block.yield(tp)
            rescue StandardError => e
              if error_collector.nil?
                Puppet.warn_once(:unloadable_entity, tp.to_s, e.message)
              else
                err = Puppet::DataTypes::Error.new(
                  Issues::LOADER_FAILURE.format(:type => type),
                  'PUPPET_LOADER_FAILURE',
                  { 'original_error' => e.message },
                  Issues::LOADER_FAILURE.issue_code)
                error_collector << err unless error_collector.include?(err)
              end
            end
          end
        end
      end
      super
    end

    # Finds typed/named entity in this module
    # @param typed_name [TypedName] the type/name to find
    # @return [Loader::NamedEntry, nil found/created entry, or nil if not found
    #
    def find(typed_name)
      # This loader is tailored to only find entries in the current runtime
      return nil unless typed_name.name_authority == Pcore::RUNTIME_NAME_AUTHORITY

      # Assume it is a global name, and that all parts of the name should be used when looking up
      name_parts = typed_name.name_parts

      # Certain types and names can be disqualified up front
      if name_parts.size > 1
        # The name is in a name space.

        # Then entity cannot possible be in this module unless the name starts with the module name.
        # Note:
        # * If "module" represents a "global component", the module_name is nil and cannot match which is
        #   ok since such a "module" cannot have namespaced content).
        # * If this loader is allowed to have namespaced content, the module_name can be set to NAMESPACE_WILDCARD `*`
        #
        return nil unless name_parts[0] == module_name || module_name == NAMESPACE_WILDCARD
      else
        # The name is in the global name space.

        case typed_name.type
        when :function, :resource_type, :resource_type_pp
          # Can be defined in module using a global name. No action required

        when :plan
          if !global?
            # Global name must be the name of the module
            return nil unless name_parts[0] == module_name

            # Look for the special 'init' plan.
            origin, smart_path = find_existing_path(init_plan_name)
            return smart_path.nil? ? nil : instantiate(smart_path, typed_name, origin)
          end

        when :task
          if !global?
            # Global name must be the name of the module
            return nil unless name_parts[0] == module_name

            # Look for the special 'init' Task
            origin, smart_path = find_existing_path(init_task_name)
            return smart_path.nil? ? nil : instantiate(smart_path, typed_name, origin)
          end

        when :type
          if !global?
            # Global name must be the name of the module
            unless name_parts[0] == module_name || module_name == NAMESPACE_WILDCARD
              # Check for ruby defined data type in global namespace before giving up
              origin, smart_path = find_existing_path(typed_name)
              return smart_path.is_a?(LoaderPaths::DataTypePath) ? instantiate(smart_path, typed_name, origin) : nil
            end

            # Look for the special 'init_typeset' TypeSet
            origin, smart_path = find_existing_path(init_typeset_name)
            return nil if smart_path.nil?

            value = smart_path.instantiator.create(self, typed_name, origin, get_contents(origin))
            if value.is_a?(Types::PTypeSetType)
              # cache the entry and return it
              return set_entry(typed_name, value, origin)
            end

            # TRANSLATORS 'TypeSet' should not be translated
            raise ArgumentError, _("The code loaded from %{origin} does not define the TypeSet '%{module_name}'") %
                { origin: origin, module_name: name_parts[0].capitalize }
          end
        else
          # anything else cannot possibly be in this module
          # TODO: should not be allowed anyway... may have to revisit this decision
          return nil
        end
      end

      # Get the paths that actually exist in this module (they are lazily processed once and cached).
      # The result is an array (that may be empty).
      # Find the file to instantiate, and instantiate the entity if file is found
      origin, smart_path = find_existing_path(typed_name)
      return instantiate(smart_path, typed_name, origin) unless smart_path.nil?

      return nil unless typed_name.type == :type && typed_name.qualified?

      # Search for TypeSet using parent name
      ts_name = typed_name.parent
      while ts_name
        # Do not traverse parents here. This search must be confined to this loader
        tse = get_entry(ts_name)
        tse = find(ts_name) if tse.nil? || tse.value.nil?
        if tse && (ts = tse.value).is_a?(Types::PTypeSetType)
          # The TypeSet might be unresolved at this point. If so, it must be resolved using
          # this loader. That in turn, adds all contained types to this loader.
          ts.resolve(self)
          te = get_entry(typed_name)
          return te unless te.nil?
        end
        ts_name = ts_name.parent
      end
      nil
    end

    def instantiate(smart_path, typed_name, origin)
      if origin.is_a?(Array)
        value = smart_path.instantiator.create(self, typed_name, origin)
      else
        value = smart_path.instantiator.create(self, typed_name, origin, get_contents(origin))
      end
      # cache the entry and return it
      set_entry(typed_name, value, origin)
    end

    # Abstract method that subclasses override that checks if it is meaningful to search using a generic smart path.
    # This optimization is performed to not be tricked into searching an empty directory over and over again.
    # The implementation may perform a deep search for file content other than directories and cache this in
    # and index. It is guaranteed that a call to meaningful_to_search? takes place before checking any other
    # path with relative_path_exists?.
    #
    # This optimization exists because many modules have been created from a template and they have
    # empty directories for functions, types, etc. (It is also the place to create a cached index of the content).
    #
    # @param smart_path [String] a path relative to the module's root
    # @return [Boolean] true if there is content in the directory appointed by the relative path
    #
    def meaningful_to_search?(smart_path)
      raise NotImplementedError.new
    end

    # Abstract method that subclasses override to answer if the given relative path exists, and if so returns that path
    #
    # @param resolved_path [String] a path resolved by a smart path against the loader's root (if it has one)
    # @return [String, nil] the found path or nil if no such path was found
    #
    def existing_path(resolved_path)
      raise NotImplementedError.new
    end

    # Abstract method that subclasses override to return an array of paths that may be associated with the resolved path.
    #
    # @param resolved_path [String] a path, without extension, resolved by a smart path against the loader's root (if it has one)
    # @return [Array<String>]
    #
    def candidate_paths(resolved_path)
      raise NotImplementedError.new
    end

    # Abstract method that subclasses override to produce the content of the effective path.
    # It should either succeed and return a String or fail with an exception.
    #
    # @param effective_path [String] a path as resolved by a smart path
    # @return [String] the content of the file
    #
    def get_contents(effective_path)
      raise NotImplementedError.new
    end

    # Abstract method that subclasses override to produce a source reference String used to identify the
    # system resource (resource in the URI sense).
    #
    # @param relative_path [String] a path relative to the module's root
    # @return [String] a reference to the source file (in file system, zip file, or elsewhere).
    #
    def get_source_ref(relative_path)
      raise NotImplementedError.new
    end

    # Answers the question if this loader represents a global component (true for resource type loader and environment loader)
    #
    # @return [Boolean] `true` if this loader represents a global component
    #
    def global?
      module_name.nil? || module_name == NAMESPACE_WILDCARD || module_name == ENVIRONMENT
    end

    # Answers `true` if the loader used by this instance is rooted beneath 'lib'. This is
    # typically true for the the system_loader. It will have a path relative to the parent
    # of 'puppet' instead of the parent of 'lib/puppet' since the 'lib' directory of puppet
    # is renamed during install. This is significant for loaders that load ruby code.
    #
    # @return [Boolean] a boolean answering if the loader is rooted beneath 'lib'.
    def lib_root?
      false
    end

    # Produces the private loader for the module. If this module is not already resolved, this will trigger resolution
    #
    def private_loader
      # The system loader has a nil module_name and it does not have a private_loader as there are no functions
      # that can only by called by puppet runtime - if so, it acts as the private loader directly.
      @private_loader ||= (global? ? self : @loaders.private_loader_for_module(module_name))
    end

    # Return all paths that matches the given smart path. The returned paths are
    # relative to the `#generic_path` of the given smart path.
    #
    # @param smart_path [SmartPath] the path to find relative paths for
    # @return [Array<String>] found paths
    def relative_paths(smart_path)
      raise NotImplementedError.new
    end

    private

    # @return [TypedName] the fake typed name that maps to the init_typeset path for this module
    def init_typeset_name
      @init_typeset_name ||= TypedName.new(:type, "#{module_name}::init_typeset")
    end

    # @return [TypedName] the fake typed name that maps to the path of an init[arbitrary extension]
    #   file that represents a task named after the module
    def init_task_name
      @init_task_name ||= TypedName.new(:task, "#{module_name}::init")
    end

    # @return [TypedName] the fake typed name that maps to the path of an init.pp file that represents
    #   a plan named after the module
    def init_plan_name
      @init_plan_name ||= TypedName.new(:plan, "#{module_name}::init")
    end

    # Find an existing path or paths for the given `typed_name`. Return `nil` if no path is found
    # @param typed_name [TypedName] the `typed_name` to find a path for
    # @return [Array,nil] `nil`or a two element array where the first element is an effective path or array of paths
    #   (depending on the `SmartPath`) and the second element is the `SmartPath` that produced the effective path or
    #   paths. A path is a String
    def find_existing_path(typed_name)
      is_global = global?
      smart_paths.effective_paths(typed_name.type).each do |sp|
        next unless sp.valid_name?(typed_name)
        origin = sp.effective_path(typed_name, is_global ? 0 : 1)
        unless origin.nil?
          if sp.fuzzy_matching?
            # If there are multiple *specific* paths for the file, find
            # whichever ones exist. Otherwise, find all paths that *might* be
            # related to origin
            if origin.is_a?(Array)
              origins = origin.map { |ori| existing_path(ori) }.compact
              return [origins, sp] unless origins.empty?
            else
              origins = candidate_paths(origin)
              return [origins, sp] unless origins.empty?
            end
          else
            existing = existing_path(origin)
            return [origin, sp] unless existing.nil?
          end
        end
      end
      nil
    end
  end

  # @api private
  #
  class FileBased < AbstractPathBasedModuleLoader

    attr_reader :smart_paths
    attr_reader :path_index

    # Create a kind of ModuleLoader for one module (Puppet Module, or module like)
    #
    # @param parent_loader [Loader] typically the loader for the environment or root
    # @param module_name [String] the name of the module (non qualified name), may be nil for "modules" only containing globals
    # @param path [String] the path to the root of the module (semantics defined by subclass)
    # @param loader_name [String] a name that identifies the loader
    #
    def initialize(parent_loader, loaders, module_name, path, loader_name, loadables = LOADABLE_KINDS)
      super
      @path_index = Set.new
    end

    def existing_path(effective_path)
      # Optimized, checks index instead of visiting file system
      @path_index.include?(effective_path) ? effective_path : nil
    end

    def candidate_paths(effective_path)
      basename = File.basename(effective_path, '.*')
      dirname = File.dirname(effective_path)

      files = @path_index.select do |path|
        File.dirname(path) == dirname
      end

      # At least one file has to match what we're loading, or it certainly doesn't exist
      if files.any? { |file| File.basename(file, '.*') == basename }
        files
      else
        []
      end
    end

    def meaningful_to_search?(smart_path)
      ! add_to_index(smart_path).empty?
    end

    def to_s()
      "(ModuleLoader::FileBased '#{loader_name}' '#{module_name}')"
    end

    def add_to_index(smart_path)
      found = Dir.glob(File.join(smart_path.generic_path, '**', "*#{smart_path.extension}"))

      # The reason for not always rejecting directories here is performance (avoid extra stat calls). The
      # false positives (directories with a matching extension) is an error in any case and will be caught
      # later.
      found = found.reject { |file_name| File.directory?(file_name) } if smart_path.extension.empty?

      @path_index.merge(found)
      found
    end

    def get_contents(effective_path)
      Puppet::FileSystem.read(effective_path, :encoding => 'utf-8')
    end

    # Return all paths that matches the given smart path. The returned paths are
    # relative to the `#generic_path` of the given smart path.
    #
    # This method relies on the cache and does not perform any file system access
    #
    # @param smart_path [SmartPath] the path to find relative paths for
    # @return [Array<String>] found paths
    def relative_paths(smart_path)
      root = smart_path.generic_path
      found = []
      @path_index.each do |path|
        found << Pathname(path).relative_path_from(Pathname(root)).to_s if smart_path.valid_path?(path)
      end
      found
    end
  end

  # Specialization used by the system_loader which is limited to see what's beneath 'lib' and hence
  # cannot be rooted in its parent. The 'lib' directory is renamed during install so any attempt
  # to traverse into it from above would fail.
  #
  # @api private
  #
  class LibRootedFileBased < FileBased
    def lib_root?
      true
    end
  end

  # Loads from a gem specified as a URI, gem://gemname/optional/path/in/gem, or just a String gemname.
  # The source reference (shown in errors etc.) is the expanded path of the gem as this is believed to be more
  # helpful - given the location it should be quite obvious which gem it is, without the location, the user would
  # need to go on a hunt for where the file actually is located.
  #
  # TODO: How does this get instantiated? Does the gemname refelect the name of the module (the namespace)
  #   or is that specified a different way? Can a gem be the container of multiple modules?
  #
  # @api private
  #
  class GemBased < FileBased
    include GemSupport

    attr_reader :gem_ref

    # Create a kind of ModuleLoader for one module
    # The parameters are:
    # * parent_loader - typically the loader for the root
    # * module_name - the name of the module (non qualified name)
    # * gem_ref - [URI, String] gem reference to the root of the module (URI, gem://gemname/optional/path/in/gem), or
    #     just the gem's name as a String.
    #
    def initialize(parent_loader, loaders, module_name, gem_ref, loader_name, loadables = LOADABLE_KINDS)
      @gem_ref = gem_ref
      super parent_loader, loaders, module_name, gem_dir(gem_ref), loader_name, loadables
    end

    def to_s()
      "(ModuleLoader::GemBased '#{loader_name}' '#{@gem_ref}' [#{module_name}])"
    end
  end
end
end
end