format_parser 2.3.0 → 2.4.4

data/lib/parsers/iso_base_media_file_format/utils.rb

@@ -0,0 +1,89 @@
+# frozen_string_literal: true
+
+require 'matrix'
+
+module FormatParser
+  module ISOBaseMediaFileFormat
+    module Utils
+
+      IDENTITY_MATRIX = Matrix.identity(3)
+
+      def dimensions(box_tree)
+        moov_box = box_tree.find { |box| box.type == 'moov' }
+        return unless moov_box
+        movie_matrix = moov_box.first_child('mvhd')&.dig(:fields, :matrix) || IDENTITY_MATRIX
+        extreme_coordinates = video_trak_boxes(box_tree).each_with_object({}) do |trak_box, extreme_coordinates|
+          tkhd_box = trak_box.first_child('tkhd')
+          next unless tkhd_box
+          x = tkhd_box.fields[:width]
+          y = tkhd_box.fields[:height]
+          next unless x && y
+          track_matrix = tkhd_box.fields[:matrix] || IDENTITY_MATRIX
+          [[0, 0], [0, y], [x, 0], [x, y]].each do |coordinates|
+            x, y = (Matrix[[*coordinates, 1]] * track_matrix * movie_matrix).to_a[0][0..1]
+            extreme_coordinates[:min_x] = x if !extreme_coordinates[:min_x] || x < extreme_coordinates[:min_x]
+            extreme_coordinates[:max_x] = x if !extreme_coordinates[:max_x] || x > extreme_coordinates[:max_x]
+            extreme_coordinates[:min_y] = y if !extreme_coordinates[:min_y] || y < extreme_coordinates[:min_y]
+            extreme_coordinates[:max_y] = y if !extreme_coordinates[:max_y] || y > extreme_coordinates[:max_y]
+          end
+        end
+        unless extreme_coordinates.empty?
+          [
+            extreme_coordinates[:max_x] - extreme_coordinates[:min_x],
+            extreme_coordinates[:max_y] - extreme_coordinates[:min_y]
+          ]
+        end
+      end
+
+      def duration(box_tree)
+        mvhd_box = box_tree.find { |box| box.type == 'moov' }&.first_child('mvhd')
+        return unless mvhd_box
+        duration = mvhd_box.fields[:duration]
+        timescale = mvhd_box.fields[:timescale]&.to_f
+        duration / timescale if duration && timescale
+      end
+
+      def frame_rate(box_tree)
+        video_trak_boxes(box_tree).each do |trak_box|
+          mdhd_box = trak_box.first_descendent_by_path(%w[mdia mdhd])
+          stts_box = trak_box.first_descendent_by_path(%w[mdia minf stbl stts])
+
+          next unless mdhd_box && stts_box
+
+          timescale = mdhd_box.fields[:timescale]&.to_f
+          sample_delta = stts_box.dig(:fields, :entries, 0, :sample_delta)
+
+          next unless timescale && sample_delta
+
+          return (timescale / sample_delta).truncate(2)
+        end
+        nil
+        # TODO: Properly account for and represent variable frame-rates.
+      end
+
+      def video_codecs(box_tree)
+        video_trak_boxes(box_tree).flat_map do |trak_box|
+          trak_box.all_descendents_by_path(%w[mdia minf stbl stsd]).flat_map { |stsd_box| stsd_box.children.map(&:type) }
+        end.compact.uniq
+      end
+
+      private
+
+      # Find any and all `trak` boxes containing a video media handler.
+      def video_trak_boxes(box_tree)
+        moov_box = box_tree.find { |box| box.type == 'moov' }
+        return [] unless moov_box
+        moov_box.all_children('trak').select do |trak_box|
+          trak_box.all_descendents('hdlr').find do |hdlr_box|
+            hdlr_fields = hdlr_box.fields
+            if hdlr_fields.include?(:component_type) && hdlr_fields.include?(:component_subtype) # MOV
+              hdlr_fields[:component_type] == 'mhlr' && hdlr_fields[:component_subtype] == 'vide'
+            else
+              hdlr_fields[:handler_type] == 'vide'
+            end
+          end
+        end
+      end
+    end
+  end
+end

data/lib/parsers/mov_parser/decoder.rb

@@ -0,0 +1,53 @@
+require 'parsers/iso_base_media_file_format/decoder'
+
+class FormatParser::MOVParser::Decoder < FormatParser::ISOBaseMediaFileFormat::Decoder
+  protected
+
+  def hdlr(size)
+    fields = read_version_and_flags.merge({
+      component_type: read_string(4),
+      component_subtype: read_string(4),
+      component_manufacturer: read_bytes(4),
+      component_flags: read_bytes(4),
+      component_flags_mask: read_bytes(4),
+      component_name: read_string(size - 24)
+    })
+    [fields, nil]
+  end
+
+  def mvhd(_)
+    fields = read_version_and_flags.merge({
+      creation_time: read_int,
+      modification_time: read_int,
+      timescale: read_int,
+      duration: read_int,
+      rate: read_fixed_point(n: 4),
+      volume: read_fixed_point(n: 2, signed: true),
+      matrix: skip_bytes(10) { read_matrix },
+      preview_time: read_int,
+      preview_duration: read_int,
+      poster_time: read_int,
+      selection_time: read_int,
+      selection_duration: read_int,
+      current_time: read_int,
+      next_trak_id: read_int,
+    })
+    [fields, nil]
+  end
+
+  def tkhd(_)
+    fields = read_version_and_flags.merge({
+      creation_time: read_int,
+      modification_time: read_int,
+      track_id: read_int,
+      duration: skip_bytes(4) { read_int },
+      layer: skip_bytes(8) { read_int(n: 2) },
+      alternate_group: read_int(n: 2),
+      volume: read_fixed_point(n: 2, signed: true),
+      matrix: skip_bytes(2) { read_matrix },
+      width: read_fixed_point(n: 4),
+      height: read_fixed_point(n: 4)
+    })
+    [fields, nil]
+  end
+end

data/lib/parsers/mov_parser.rb

@@ -0,0 +1,48 @@
+require 'parsers/iso_base_media_file_format/utils'
+
+class FormatParser::MOVParser
+  include FormatParser::IOUtils
+  include FormatParser::ISOBaseMediaFileFormat::Utils
+  require_relative 'mov_parser/decoder'
+
+  MAGIC_BYTES = 'ftypqt  '
+  MOV_MIME_TYPE = 'video/quicktime'
+
+  def likely_match?(filename)
+    /\.(mov|moov|qt)$/i.match?(filename)
+  end
+
+  def call(io)
+    @buf = FormatParser::IOConstraint.new(io)
+
+    return unless matches_mov_definition?
+
+    box_tree = Measurometer.instrument('format_parser.mov_parser.decoder.build_box_tree') do
+      Decoder.new.build_box_tree(0xffffffff, @buf)
+    end
+
+    width, height = dimensions(box_tree)
+
+    FormatParser::Video.new(
+      format: :mov,
+      width_px: width,
+      height_px: height,
+      frame_rate: frame_rate(box_tree),
+      media_duration_seconds: duration(box_tree),
+      content_type: MOV_MIME_TYPE,
+      codecs: video_codecs(box_tree),
+      intrinsics: box_tree
+    )
+  end
+
+  private
+
+  def matches_mov_definition?
+    skip_bytes(4)
+    matches = read_string(8) == MAGIC_BYTES
+    @buf.seek(0)
+    matches
+  end
+
+  FormatParser.register_parser new, natures: [:video], formats: [:mov], priority: 3
+end

data/lib/parsers/mp4_parser.rb

@@ -0,0 +1,80 @@
+require_relative 'iso_base_media_file_format/decoder'
+
+class FormatParser::MP4Parser
+  include FormatParser::IOUtils
+  include FormatParser::ISOBaseMediaFileFormat
+  include FormatParser::ISOBaseMediaFileFormat::Utils
+
+  MAGIC_BYTES = /^ftyp(iso[m2]|mp4[12]|m4[abprv] )$/i
+
+  BRAND_FORMATS = {
+    'isom' => :mp4, # Prohibited as a major brand by ISO/IEC 14496-12 sec 6.3 paragraph 2, but occasionally used.
+    'iso2' => :mp4, # Prohibited as a major brand by ISO/IEC 14496-12 sec 6.3 paragraph 2, but occasionally used.
+    'mp41' => :mp4,
+    'mp42' => :mp4,
+    'm4a ' => :m4a,
+    'm4b ' => :m4b, # iTunes audiobooks
+    'm4p ' => :m4p, # iTunes audio
+    'm4r ' => :m4r, # iTunes ringtones
+    'm4v ' => :m4v, # iTunes video
+  }
+  AUDIO_FORMATS = Set[:m4a, :m4b, :m4p, :m4r]
+  VIDEO_FORMATS = Set[:mp4, :m4v]
+
+  AUDIO_MIMETYPE = 'audio/mp4'
+  VIDEO_MIMETYPE = 'video/mp4'
+
+  def likely_match?(filename)
+    /\.(mp4|m4[abprv])$/i.match?(filename)
+  end
+
+  def call(io)
+    @buf = FormatParser::IOConstraint.new(io)
+
+    return unless matches_mp4_definition?
+
+    box_tree = Measurometer.instrument('format_parser.mp4_parser.decoder.build_box_tree') do
+      Decoder.new.build_box_tree(0xffffffff, @buf)
+    end
+
+    case file_format = file_format(box_tree)
+    when VIDEO_FORMATS
+      width, height = dimensions(box_tree)
+      FormatParser::Video.new(
+        codecs: video_codecs(box_tree),
+        content_type: VIDEO_MIMETYPE,
+        format: file_format,
+        frame_rate: frame_rate(box_tree),
+        height_px: height,
+        intrinsics: box_tree,
+        media_duration_seconds: duration(box_tree),
+        width_px: width,
+      )
+    when AUDIO_FORMATS
+      FormatParser::Audio.new(
+        content_type: AUDIO_MIMETYPE,
+        format: file_format,
+        intrinsics: box_tree,
+        media_duration_seconds: duration(box_tree),
+      )
+    else
+      nil
+    end
+  end
+
+  private
+
+  def file_format(box_tree)
+    major_brand = box_tree.find { |box| box.type == 'ftyp' }&.fields[:major_brand]
+    BRAND_FORMATS[major_brand.downcase] if major_brand
+  end
+
+  def matches_mp4_definition?
+    skip_bytes(4)
+    matches = MAGIC_BYTES.match?(read_string(8))
+    @buf.seek(0)
+    matches
+  end
+
+  FormatParser.register_parser new, natures: [:audio, :video], formats: BRAND_FORMATS.values.uniq, priority: 3
+end

data/lib/parsers/pdf_parser.rb

@@ -6,7 +6,7 @@ class FormatParser::PDFParser
   #
   # There are however exceptions, which are left out for now.
   #
-  PDF_MARKER = /%PDF-1\.[0-8]{1}/
+  PDF_MARKER = /%PDF-[12]\.[0-8]{1}/
   PDF_CONTENT_TYPE = 'application/pdf'
 
   def likely_match?(filename)
@@ -16,9 +16,12 @@ class FormatParser::PDFParser
   def call(io)
     io = FormatParser::IOConstraint.new(io)
 
-    return unless safe_read(io, 9) =~ PDF_MARKER
+    header = safe_read(io, 9)
+    return unless header =~ PDF_MARKER
 
     FormatParser::Document.new(format: :pdf, content_type: PDF_CONTENT_TYPE)
+  rescue FormatParser::IOUtils::InvalidRead
+    nil
   end
 
   FormatParser.register_parser new, natures: :document, formats: :pdf, priority: 3

data/lib/parsers/webp_parser.rb

@@ -69,7 +69,7 @@ class FormatParser::WebpParser
     # The subsequent 4 bytes contain the image width and height, respectively, as 14-bit unsigned little endian
     # integers (minus one). The 4 remaining bits consist of a 1-bit flag indicating whether alpha is used, and a 3-bit
     # version that is always zero.
-    dimensions = read_little_endian_int_32
+    dimensions = read_int(big_endian: false)
     width = (dimensions & 0x3fff) + 1
     height = (dimensions >> 14 & 0x3fff) + 1
     has_transparency = (dimensions >> 28 & 0x1) == 1
@@ -92,7 +92,7 @@ class FormatParser::WebpParser
     #   - E = Set if file contains Exif metadata.
     #   - X = Set if file contains XMP metadata.
     #   - A = Set if file is an animated image.
-    flags = read_int_8
+    flags = read_int(n: 1)
     has_transparency = flags & 0x10 != 0
     has_exif_metadata = flags & 0x08 != 0
     has_xmp_metadata = flags & 0x04 != 0

data/spec/format_parser_spec.rb

@@ -184,7 +184,7 @@ describe FormatParser do
         'FormatParser::GIFParser',
         'Class',
         'FormatParser::PNGParser',
-        'FormatParser::MOOVParser',
+        'FormatParser::MP4Parser',
         'FormatParser::CR2Parser',
         'FormatParser::CR3Parser',
         'FormatParser::DPXParser',

data/spec/parsers/cr3_parser_spec.rb

@@ -18,9 +18,9 @@ describe FormatParser::CR3Parser do
     expect(result).to be_nil
   end
 
-  it 'should return nil if no CMT1 atom is present' do
+  it 'should return nil if no CMT1 box is present' do
     # This is a MOV file with the ftyp header modified to masquerade as a CR3 file. It is therefore missing the
-    # CR3-specific CMT1 atom containing the image metadata.
+    # CR3-specific CMT1 box containing the image metadata.
     result = subject.call(File.open(fixtures_dir + '/CR3/invalid'))
     expect(result).to be_nil
   end
@@ -50,7 +50,7 @@ describe FormatParser::CR3Parser do
     expect(result.display_height_px).to eq(4000)
     expect(result.content_type).to eq('image/x-canon-cr3')
     expect(result.intrinsics).not_to be_nil
-    expect(result.intrinsics[:atom_tree]).not_to be_nil
+    expect(result.intrinsics[:box_tree]).not_to be_nil
     expect(result.intrinsics[:exif]).not_to be_nil
     expect(result.intrinsics[:exif][:image_length]).to eq(result.height_px)
     expect(result.intrinsics[:exif][:image_width]).to eq(result.width_px)