protocol-hpack 1.4.3 → 1.5.1

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 3192190834a35550921cbff8e50482499716791371f521a77c3f7c4b5a0efcff
-  data.tar.gz: 40d443b19eaac26aaaaae63da3baab5ae384246804643fdc7fb01f5455595dc3
+  metadata.gz: ad81d71708c7c92bb94e9616655872045a31052aa62c3543487064ddea8c99ad
+  data.tar.gz: 1945361f8e8576ef3db44de1e0cf44554795699d910c40b8659232a499ded6b8
 SHA512:
-  metadata.gz: c19a7524c1a544016e0feb7d88099a67fbccf3fe200214f7bc21639512da55e72a9e3442c8917f69fec1568a3027487e701a2f1e99deb652942ebae54c4752a7
-  data.tar.gz: 3f105d7bba7111585325db6bc8f414c34366bfb189f2263fa497e8ec53ed1502c3694c0dcd6749151fb10456c52a5a333daa9b7f5f8b362026d8be745c568375
+  metadata.gz: c0537372d9278691fa9621622b84f32facde4f9efdfbfd0c36eb7573bba4b19ec05bc91f5191a7463405cbed952c4c6ac614531c4c0fc0b26707196db8d800b3
+  data.tar.gz: fc53196d54e9b02a56df5c1294e819f5a4002c49201811c12d2666725a0325e27df75603afc4b04a5fc5c1845237937097bf51352ea4d5f1c3741d1b73fa244a

data/lib/protocol/hpack/compressor.rb

@@ -11,9 +11,10 @@
 # Copyright, 2018, by Kenichi Nakamura.
 # Copyright, 2019, by Jingyi Chen.
 # Copyright, 2020, by Justin Mazzocchi.
+# Copyright, 2024, by Nathan Froyd.
 
-require_relative 'context'
-require_relative 'huffman'
+require_relative "context"
+require_relative "huffman"
 
 module Protocol
 	module HPACK
@@ -74,22 +75,19 @@ module Protocol
 			# @param bits [Integer] number of available bits
 			# @return [String] binary string
 			def write_integer(value, bits)
-				limit = 2**bits - 1
+				limit = (1 << bits) - 1
 				
-				return write_bytes([value].pack('C')) if value < limit
+				return @buffer << value if value < limit
 				
-				bytes = []
-				bytes.push(limit) unless bits.zero?
+				@buffer << limit unless bits.zero?
 				
 				value -= limit
 				while value >= 128
-					bytes.push((value % 128) + 128)
+					@buffer << ((value & 0x7f) + 128)
 					value /= 128
 				end
 				
-				bytes.push(value)
-				
-				write_bytes(bytes.pack('C*'))
+				@buffer << value
 			end
 			
 			def huffman
@@ -118,7 +116,7 @@ module Protocol
 			# @return [String] binary string
 			def write_string(string, huffman = self.huffman)
 				if huffman != :never
-					encoded = Huffman.new.encode(string)
+					encoded = Huffman.encode(string)
 					
 					if huffman == :shorter and encoded.bytesize >= string.bytesize
 						encoded = nil

data/lib/protocol/hpack/context.rb

@@ -2,8 +2,10 @@
 
 # Released under the MIT License.
 # Copyright, 2018-2024, by Samuel Williams.
+# Copyright, 2024, by Maruth Goyal.
+# Copyright, 2024, by Nathan Froyd.
 
-require_relative 'huffman'
+require_relative "huffman"
 
 module Protocol
 	# Implementation of header compression for HTTP 2.0 (HPACK) format adapted
@@ -94,7 +96,20 @@ module Protocol
 				["via", ""],
 				["www-authenticate", ""],
 			].each(&:freeze).freeze
-			
+
+			STATIC_EXACT_LOOKUP = {}
+			STATIC_NAME_LOOKUP = {}
+
+			STATIC_TABLE.each_with_index do |(name, value), i|
+				exact_header_values = (STATIC_EXACT_LOOKUP[name] ||= [])
+				exact_header_values << [value, i]
+				STATIC_NAME_LOOKUP[name] = i if STATIC_NAME_LOOKUP[name].nil?
+			end
+
+			STATIC_EXACT_LOOKUP.each {|k, v| v.freeze}
+			STATIC_EXACT_LOOKUP.freeze
+			STATIC_NAME_LOOKUP.freeze
+
 			# Initializes compression context with appropriate client/server defaults and maximum size of the dynamic table.
 			#
 			# @param table [Array] Table of header key-value pairs.
@@ -233,29 +248,42 @@ module Protocol
 			#  :static  Use static table only.
 			#  :all     Use all of them.
 			#
-			# @param header [Array] +[name, value]+
+			# @param name [String]
+			# @param value [String]
 			# @return [Hash] command
-			def add_command(*header)
+			def add_command(name, value)
 				exact = nil
 				name_only = nil
 
-				if [:all, :static].include?(@index)
-					STATIC_TABLE.each_index do |i|
-						if STATIC_TABLE[i] == header
-							exact ||= i
-							break
-						elsif STATIC_TABLE[i].first == header.first
-							name_only ||= i
+				if @index == :all || @index == :static
+					if (values_and_indices = STATIC_EXACT_LOOKUP[name])
+						values_and_indices.each do |known_value, index|
+							if value == known_value
+								exact = index
+								break
+							end
 						end
+						
+						needs_name_lookup = exact.nil?
+					else
+						needs_name_lookup = true
+					end
+
+					if needs_name_lookup && (static_value = STATIC_NAME_LOOKUP[name])
+						name_only = static_value
 					end
 				end
-				if [:all].include?(@index) && !exact
+
+				if @index == :all && !exact
 					@table.each_index do |i|
-						if @table[i] == header
-							exact ||= i + STATIC_TABLE.size
-							break
-						elsif @table[i].first == header.first
-							name_only ||= i + STATIC_TABLE.size
+						entry = @table[i]
+						if entry.first == name
+							if entry.last == value
+								exact ||= i + STATIC_TABLE.size
+								break
+							else
+								name_only ||= i + STATIC_TABLE.size
+							end
 						end
 					end
 				end
@@ -263,9 +291,9 @@ module Protocol
 				if exact
 					{name: exact, type: :indexed}
 				elsif name_only
-					{name: name_only, value: header.last, type: :incremental}
+					{name: name_only, value: value, type: :incremental}
 				else
-					{name: header.first, value: header.last, type: :incremental}
+					{name: name, value: value, type: :incremental}
 				end
 			end
 

data/lib/protocol/hpack/decompressor.rb

@@ -2,9 +2,11 @@
 
 # Released under the MIT License.
 # Copyright, 2018-2024, by Samuel Williams.
+# Copyright, 2024, by Maruth Goyal.
+# Copyright, 2024, by Nathan Froyd.
 
-require_relative 'context'
-require_relative 'huffman'
+require_relative "context"
+require_relative "huffman"
 
 module Protocol
 	module HPACK
@@ -88,7 +90,7 @@ module Protocol
 				
 				raise CompressionError, "Invalid string length, got #{string.bytesize}, expecting #{length}!" unless string.bytesize == length
 				
-				string = Huffman.new.decode(string) if huffman
+				string = Huffman.decode(string) if huffman
 				
 				return string.force_encoding(Encoding::UTF_8)
 			end

data/lib/protocol/hpack/huffman/generator.rb

@@ -0,0 +1,186 @@
+# frozen_string_literal: true
+
+# Released under the MIT License.
+# Copyright, 2014, by Kaoru Maeda.
+# Copyright, 2015, by Tamir Duberstein.
+# Copyright, 2015, by Ilya Grigorik.
+# Copyright, 2016, by George Ulmer.
+# Copyright, 2018-2024, by Samuel Williams.
+# Copyright, 2024, by Nathan Froyd.
+
+require_relative "../huffman"
+require "set"
+
+module Protocol
+	module HPACK
+		class Huffman
+			module Generator
+				EOS = 256
+				
+				class Node
+					attr_accessor :next, :emit, :final, :depth
+					attr_accessor :transitions
+					attr_accessor :id
+					
+					@@id = 0
+					
+					def initialize(depth)
+						@next = [nil, nil]
+						@id = @@id
+						@@id += 1
+						@final = false
+						@depth = depth
+					end
+					
+					def add(code, len, chr)
+						self.final = true if chr == EOS && @depth <= 7
+						if len.zero?
+							@emit = chr
+						else
+							bit = (code & (1 << (len - 1))).zero? ? 0 : 1
+							node = @next[bit] ||= Node.new(@depth + 1)
+							node.add(code, len - 1, chr)
+						end
+					end
+					
+					class Transition
+						attr_accessor :emit, :node
+						def initialize(emit, node)
+							@emit = emit
+							@node = node
+						end
+					end
+					
+					def self.generate_tree
+						@root = new(0)
+						Protocol::HPACK::Huffman::CODES.each_with_index do |c, chr|
+							code, len = c
+							@root.add(code, len, chr)
+						end
+						puts "#{@@id} nodes"
+						@root
+					end
+					
+					def self.generate_machine
+						generate_tree
+						
+						# Using un-ordered sets (potentially) produces non-deterministic results:
+						togo = Set[@root]
+						@states = Set[@root]
+						
+						until togo.empty?
+							node = togo.first
+							togo.delete(node)
+							
+							next if node.transitions
+							node.transitions = Array[1 << BITS_AT_ONCE]
+							
+							(1 << BITS_AT_ONCE).times do |input|
+								n = node
+								emit = +""
+								(BITS_AT_ONCE - 1).downto(0) do |i|
+									bit = (input & (1 << i)).zero? ? 0 : 1
+									n = n.next[bit]
+									next unless n.emit
+									if n.emit == EOS
+										emit = EOS # cause error on decoding
+									else
+										emit << n.emit.chr(Encoding::BINARY) unless emit == EOS
+									end
+									n = @root
+								end
+								node.transitions[input] = Transition.new(emit, n)
+								togo << n
+								@states << n
+							end
+						end
+						puts "#{@states.size} states"
+						@root
+					end
+					
+					MACHINE_PATH = File.expand_path("machine.rb", __dir__)
+					
+					def self.generate_state_table(output_path = MACHINE_PATH)
+						generate_machine
+						state_id = {}
+						id_state = {}
+						state_id[@root] = 0
+						id_state[0] = @root
+						max_final = 0
+						id = 1
+						(@states - [@root]).sort_by {|s| s.final ? 0 : 1}.each do |s|
+							state_id[s] = id
+							id_state[id] = s
+							max_final = id if s.final
+							id += 1
+						end
+
+						File.open(output_path, "w") do |file|
+							file.print <<~HEADER
+								# frozen_string_literal: true
+
+								# Released under the MIT License.
+								# Copyright, 2018-2024, by Samuel Williams.
+
+								# Machine generated Huffman decoder state machine.
+								# DO NOT EDIT THIS FILE.
+
+								module Protocol
+									module HPACK
+										class Huffman
+											# :nodoc:
+											MAX_FINAL_STATE = #{max_final}
+											MACHINE = [
+							HEADER
+							
+							id.times do |i|
+								n = id_state[i]
+								file.print "\t\t\t\t["
+								string = (1 << BITS_AT_ONCE).times.map do |t|
+									transition = n.transitions.fetch(t)
+									emit = transition.emit
+									unless emit == EOS
+										bytes = emit.bytes
+										fail ArgumentError if bytes.size > 1
+										emit = bytes.first
+									end
+									"[#{emit.inspect}, #{state_id.fetch(transition.node)}]"
+								end.join(", ")
+								file.print(string)
+								file.print "],\n"
+							end
+							
+							file.print <<~FOOTER
+											].each {|arr| arr.each {|subarr| subarr.each(&:freeze)}.freeze}.freeze
+										end
+									end
+								end
+							FOOTER
+						end
+					end
+					
+					class << self
+						attr_reader :root
+					end
+					
+					# Test decoder
+					def self.decode(input)
+						emit = ""
+						n = root
+						nibbles = input.unpack("C*").flat_map {|b| [((b & 0xf0) >> 4), b & 0xf]}
+						until nibbles.empty?
+							nb = nibbles.shift
+							t = n.transitions[nb]
+							emit << t.emit
+							n = t.node
+						end
+						unless n.final && nibbles.all? {|x| x == 0xf}
+							puts "len = #{emit.size} n.final = #{n.final} nibbles = #{nibbles}"
+						end
+						emit
+					end
+				end
+			end
+		end
+	end
+end