require 'open3'
require 'securerandom'
require 'tmpdir'
require 'iiif_print/split_pdfs/pdf_image_extraction_service'

module IiifPrint
  module SplitPdfs
    class PagesIntoImagesService
      include Enumerable

      def initialize(path)
        @baseid = SecureRandom.uuid
        @pdfpath = path
        @info = nil
        @entries = nil
        @tmpdir = nil
        @size = nil
        @pagecount = nil
        @pdftext = nil
        @compression = 'lzw'
      end

      # return
      def pdfinfo
        @info = IiifPrint::SplitPdfs::PdfImageExtractionService.new(@pdfpath) if @info.nil?
        @info
      end

      # TODO: put this test somewhere to prevent invalid pdfs from crashing the image service.
      def invalid_pdf?
        return true if pdfinfo.color.include?(nil) || pdfinfo.width.nil? || pdfinfo.height.nil? || pdfinfo.entries.length.zero?
        false
      end

      def tmpdir
        @tmpdir = Dir.mktmpdir if @tmpdir.nil?
        @tmpdir
      end

      def colordevice(channels, bpc)
        bits = bpc * channels
        # will be either 8bpc/16bpd color TIFF,
        #   with any CMYK source transformed to 8bpc RBG
        bits = 24 unless [24, 48].include? bits
        "tiff#{bits}nc"
      end

      def gsdevice
        color, channels, bpc = pdfinfo.color
        device = nil
        # CCITT Group 4 Black and White, if applicable:
        if color == 'gray' && bpc == 1
          device = 'tiffg4'
          @compression = 'g4'
        end
        # 8 Bit Grayscale, if applicable:
        device = 'tiffgray' if color == 'gray' && bpc > 1
        # otherwise color:
        device = colordevice(channels, bpc) if device.nil?
        device
      end

      # TODO: this method came from newspaper gem but appears to be unused. Is it needed anywhere?
      # def gstext
      #   cmd = "gs -q -dNOPAUSE -dBATCH -sDEVICE=txtwrite " \
      #         "-sOutputFile=- -f #{@pdfpath}"
      #   Open3.popen3(cmd) do |_stdin, stdout, _stderr, _wait_thr|
      #     @pdftext = stdout.read
      #   end
      #   @pdftext
      # end

      def pagecount
        cmd = "pdfinfo #{@pdfpath}"
        Open3.popen3(cmd) do |_stdin, stdout, _stderr, _wait_thr|
          output = stdout.read.split("\n")
          # rubocop:disable Performance/Detect
          pages_e = output.select { |e| e.start_with?('Pages:') }[0]
          # rubocop:enable Performance/Detect
          @pagecount = pages_e.split[-1].to_i
        end
        @pagecount
      end

      def looks_scanned
        max_image_px = pdfinfo.width * pdfinfo.height
        single_image_per_page = pdfinfo.entries.length == pagecount
        # single 10mp+ image per page?
        single_image_per_page && max_image_px > 1024 * 1024 * 10
      end

      def ppi
        unless looks_scanned
          # 400 dpi for something that does not look like scanned media:
          return 400
        end
        # For scanned media, defer to detected image PPI:
        pdfinfo.ppi
      end

      # ghostscript convert all pages to TIFF
      def gsconvert
        output_base = File.join(tmpdir, "#{@baseid}-page%d.tiff")
        cmd = "gs -dNOPAUSE -dBATCH -sDEVICE=#{gsdevice} " \
              "-dTextAlphaBits=4 -sCompression=#{@compression} " \
              "-sOutputFile=#{output_base} -r#{ppi} -f #{@pdfpath}"
        Open3.popen3(cmd) do |_stdin, stdout, _stderr, _wait_thr|
          output = stdout.read.split("\n")
          # rubocop:disable Performance/Count
          @size = output.select { |e| e.start_with?('Page ') }.length
          # rubocop:enable Performance/Count
        end
        # Return an array of expected filenames
        (1..@size).map { |n| File.join(tmpdir, "#{@baseid}-page#{n}.tiff") }
      end

      # entries for each page
      def entries
        @entries = gsconvert if @entries.nil?
        @entries
      end

      def each
        entries.each do |e|
          yield(e)
        end
      end
    end
  end
end