# -*- encoding: utf-8 -*-
require 'rdf' # @see https:/rubygems.org/gems/rdf

module SXP; class Reader
  ##
  # A SPARQL Syntax Expressions (SSE) parser.
  #
  # Requires [RDF.rb](https:/rubygems.org/gems/rdf/).
  #
  # @see https:/openjena.org/wiki/SSE
  class SPARQL < Extended
    # Alias for rdf:type
    A         = /^a$/
    # Base token, causes next URI to be treated as the `base_uri` for further URI expansion
    BASE      = /^base$/i
    # Prefix token, causes following prefix and URI pairs to be used for transforming
    # {PNAME} tokens into URIs.
    PREFIX    = /^prefix$/i
    NIL       = /^nil$/i
    FALSE     = /^false$/i
    TRUE      = /^true$/i
    EXPONENT  = /[eE][+-]?[0-9]+/
    DECIMAL   = /^[+-]?(\d*)?\.\d*$/
    DOUBLE    = /^[+-]?(\d*)?\.\d*#{EXPONENT}$/
    # BNode with identifier
    BNODE_ID  = /^_:([^\s]*)/ # FIXME
    # Anonymous BNode
    BNODE_NEW = /^_:$/
    # Distinguished variable
    VAR_ID    = /^\?(.*)/
    # Non-distinguished variable
    ND_VAR   = /^\?(?:[\?\.])(.*)/
    # Distinguished existential variable
    EVAR_ID    = /^\$(.*)/
    # Non-distinguished existential variable
    ND_EVAR   = /^\$(?:[\$\.])(.*)/
    # A QName, subject to expansion to URIs using {PREFIX}
    PNAME     = /([^:]*):(.*)/
    
    RDF_TYPE  = (a = RDF.type.dup; a.lexical = 'a'; a).freeze

    ##
    # Base URI as specified or when parsing parsing a BASE token using the immediately following
    # token, which must be a URI.
    attr_accessor :base_uri

    ##
    # Prefixes defined while parsing
    # @return [Hash{Object => RDF::URI}]
    attr_accessor :prefixes

    ##
    # Defines the given named URI prefix for this parser.
    #
    # @example Defining a URI prefix
    #   parser.prefix :dc, RDF::URI('http://purl.org/dc/terms/')
    #
    # @example Returning a URI prefix
    #   parser.prefix(:dc)    #=> RDF::URI('http://purl.org/dc/terms/')
    #
    # @param  [Symbol, #to_s]   name
    # @param  [RDF::URI, #to_s] uri
    # @return [RDF::URI]
    def prefix(name, uri = nil)
      name = name.to_s.empty? ? nil : (name.respond_to?(:to_sym) ? name.to_sym : name.to_s.to_sym)
      uri.nil? ? @prefixes[name] : @prefixes[name] = uri
    end

    ##
    # Initializes the reader.
    #
    # @param  [IO, StringIO, String]   input
    # @param  [Hash{Symbol => Object}] options
    def initialize(input, **options, &block)
      super { @prefixes = {}; @bnodes = {}; @list_depth = 0 }

      if block_given?
        case block.arity
          when 1 then block.call(self)
          else self.instance_eval(&block)
        end
      end
    end

    ##
    # Reads SSE Tokens, including `RDF::Literal`, `RDF::URI` and `RDF::Node`.
    #
    # Performs forward reference for prefix and base URI representations and saves in
    # {#base_uri} and {#prefixes} accessors.
    #
    # Transforms tokens matching a {PNAME} pattern into `RDF::URI` instances if a match is
    # found with a previously identified {PREFIX}.
    # @return [Object]
    def read_token
      case peek_char
      when ?" then [:atom, read_rdf_literal] # "
      when ?' then [:atom, read_rdf_literal] # '
      when ?< then [:atom, read_rdf_uri]
      else
        tok = super
        
        # If we just parsed "PREFIX", and this is an opening list, then
        # record list depth and process following as token, URI pairs
        #
        # Once we've closed the list, go out of prefix mode
        if tok.is_a?(Array) && tok[0] == :list
          if '(['.include?(tok[1])
            @list_depth += 1
          else
            @list_depth -= 1
            @prefix_depth = nil if @prefix_depth && @list_depth < @prefix_depth
          end
        end

        if tok.is_a?(Array) && tok[0] == :atom && tok[1].is_a?(Symbol)
          value = tok[1].to_s

          # We previously parsed a PREFIX, this will be the map value
          @parsed_prefix = value.chop if @prefix_depth && @prefix_depth > 0
          
          # If we just saw PREFIX, then this starts the parsing mode
          @prefix_depth = @list_depth + 1 if value =~ PREFIX
          
          # If the token is of the form 'prefix:suffix', create a URI and give it the
          # token as a QName
          if value.to_s =~ PNAME && base = prefix($1)
            suffix = $2
            #STDERR.puts "read_tok lexical: pfx: #{$1.inspect} => #{prefix($1).inspect}, sfx: #{suffix.inspect}"
            suffix = suffix.sub(/^\#/, "") if base.to_s.index("#")
            uri = RDF::URI(base.to_s + suffix)
            #STDERR.puts "read_tok lexical uri: #{uri.inspect}"

            [:atom, uri]
          else
            tok
          end
        else
          tok
        end
      end
    end

    ##
    # Reads literals corresponding to SPARQL/Turtle/Notation-3 syntax
    #
    # @example
    #   "a plain literal"
    #   'another plain literal'
    #   "a literal with a language"@en
    #   "a typed literal"^^<http://example/>
    #   "a typed literal with a PNAME"^^xsd:string
    #
    # @return [RDF::Literal]
    def read_rdf_literal
      value   = read_string
      options = case peek_char
        when ?@
          skip_char # '@'
          {language: read_atom.downcase}
        when ?^
          2.times { skip_char } # '^^'
          {datatype: read_token.last}
        else {}
      end
      RDF::Literal(value, **options)
    end

    ##
    # Reads a URI in SPARQL/Turtle/Notation-3 syntax
    #
    # @example
    #   <http://example/>
    #
    # @return [RDF::URI]
    def read_rdf_uri
      buffer = ""
      skip_char # '<'
      return :< if (char = peek_char).nil? || char.chr !~ ATOM # FIXME: nasty special case for the '< symbol
      return :<= if peek_char.chr.eql?(?=.chr) && read_char    # FIXME: nasty special case for the '<= symbol
      until peek_char == ?>
        buffer << read_char # TODO: unescaping
      end
      skip_char # '>'

      # If we have a base URI, use that when constructing a new URI
      uri = if self.base_uri && RDF::URI(buffer).relative?
        self.base_uri.join(buffer)
      else
        RDF::URI(buffer)
      end
      
      # If we previously parsed a "BASE" element, then this URI is used to set that value
      if @parsed_base
        self.base_uri = uri
        @parsed_base = nil
      end
      
      # If we previously parsed a "PREFIX" element, associate this URI with the prefix
      if @parsed_prefix
        prefix(@parsed_prefix, uri)
        @parsed_prefix = nil
      end
      
      uri
    end

    ##
    # Reads an SSE Atom
    #
    # Atoms parsed including `base`, `prefix`, `true`, `false`, numeric, BNodes and variables.
    #
    # Creates `RDF::Literal`, `RDF::Node`, or `RDF::Query::Variable` instances where appropriate.
    #
    # @return [Object]
    def read_atom
      case buffer = read_literal
        when '.'       then buffer.to_sym
        when A         then RDF_TYPE
        when BASE      then @parsed_base = true; buffer.to_sym
        when NIL       then nil
        when FALSE     then RDF::Literal(false)
        when TRUE      then RDF::Literal(true)
        when DOUBLE    then RDF::Literal::Double.new(buffer)
        when DECIMAL   then RDF::Literal::Decimal.new(buffer)
        when INTEGER   then RDF::Literal::Integer.new(buffer)
        when BNODE_ID  then @bnodes[$1] ||= RDF::Node($1)
        when BNODE_NEW then RDF::Node.new
        when ND_VAR    then variable($1, distinguished: false)
        when VAR_ID    then variable($1, distinguished: true)
        when ND_EVAR   then variable($1, existential: true, distinguished: false)
        when EVAR_ID   then variable($1, existential: true, distinguished: true)
        else buffer.to_sym
      end
    end

    ##
    # @return [void]
    def skip_comments
      until eof?
        case (char = peek_char).chr
          when /\s+/ then skip_char
          when /;/   then skip_line
          when /#/   then skip_line
          else break
        end
      end
    end
    
    ##
    # Return variable allocated to an ID.
    # If no ID is provided, a new variable
    # is allocated. Otherwise, any previous assignment will be used.
    #
    # The variable has a #distinguished? method applied depending on if this
    # is a disinguished or non-distinguished variable. Non-distinguished
    # variables are effectively the same as BNodes.
    # @return [RDF::Query::Variable]
    def variable(id, distinguished: true, existential: false)
      id = nil if id.to_s.empty?
      
      if id
        @vars ||= {}
        @vars[id] ||= begin
          RDF::Query::Variable.new(id, distinguished: distinguished, existential: existential)
        end
      else
        RDF::Query::Variable.new(distinguished: distinguished, existential: existential)
      end
    end
  end # SPARQL
end; end # SXP::Reader