djb2 0.1.1 → 0.2.0

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 69efd4b5520aefff0c661214b57a87aa95a3eb64baa0fa298bc6a4a282fea476
-  data.tar.gz: 848fa3e0e64c9fcc8d595232c4507478b0ed530bfa7d751a85d3dad5f68f0dc4
+  metadata.gz: 2daeb6437c508af570b7efcab866f08594ca051704f4d1da3b43e4a0da2a47c3
+  data.tar.gz: b4964f45412483fd9ed4d8d77f4f74f92fa361efd174a650c98386270ef60363
 SHA512:
-  metadata.gz: b08618ceb92187f72f6e6bc743ae199e332a86c62d6075e2fee824641565ea815c6121af9fa4d8ecb7ce2750363d670147b380880734f3edf0fcf8796c9ea13a
-  data.tar.gz: 4988363fafdd18547b51d683b94eca713a828d523e1a30550bb6a62e8aa1cba23d9d26c040fab4daed2263c13d793a330eb2b63da7143f3c6738c9e22d52f96d
+  metadata.gz: 2460212ef211b32a57c3e1885a8c7d8f57319db9fbcb0430529d35e6a85a54769f3449d98d16073e960ae07413b8ec9ea8b05a2a785c15e78d241fd07024c2f4
+  data.tar.gz: 8785a9c2dcdb3bdacf3285d1c0a0f0f5fc8f9c6b430f28276f4543ee3aab8e6546e5906b137a7fa8ee1c0865860c2c3b986517ece8b076e2c1462a0b1e6f7b8c

data/README.md

@@ -1,6 +1,6 @@
 # DJB2
 
-Native djb2 hash implementation
+Pure Ruby djb2 hash implementation, optimized for YJIT.
 
 ## Installation
 

data/Rakefile

@@ -9,14 +9,4 @@ Rake::TestTask.new(:test) do |t|
   t.test_files = FileList["test/**/test_*.rb"]
 end
 
-require "rake/extensiontask"
-
-task build: :compile
-
-GEMSPEC = Gem::Specification.load("djb2.gemspec")
-
-Rake::ExtensionTask.new("djb2", GEMSPEC) do |ext|
-  ext.lib_dir = "lib/djb2"
-end
-
-task default: %i[clobber compile test]
+task default: :test

data/benchmark/benchmark.rb

@@ -0,0 +1,71 @@
+# frozen_string_literal: true
+
+# Re-exec with --yjit if YJIT is available but not yet enabled.
+if defined?(RubyVM::YJIT) && !RubyVM::YJIT.enabled?
+  exec(RbConfig.ruby, "--yjit", $0, *ARGV)
+end
+
+require "bundler/inline"
+
+gemfile do
+  source "https://rubygems.org"
+  gem "benchmark-ips"
+  gem "djb2", "0.1.1"
+end
+
+# ── Load the C extension (released gem) ──────────────────────────────────────
+require "djb2"
+c_digest = DJB2.method(:digest)
+
+# ── Pure Ruby implementation (this branch) ───────────────────────────────────
+load File.expand_path("../lib/djb2.rb", __dir__)
+ruby_digest = DJB2.method(:digest)
+
+# ── Verify correctness ──────────────────────────────────────────────────────
+test_strings = ["foo", "bar", "hello world", "x" * 100, Random.bytes(1000)]
+test_strings.each do |s|
+  c_result = c_digest.call(s)
+  ruby_result = ruby_digest.call(s)
+  unless c_result == ruby_result
+    abort "MISMATCH on #{s.inspect[0..40]}: C=#{c_result} Ruby=#{ruby_result}"
+  end
+end
+puts "Correctness verified: both implementations produce identical results."
+puts
+
+# ── Environment ──────────────────────────────────────────────────────────────
+puts "Ruby:  #{RUBY_VERSION} (#{RUBY_PLATFORM})"
+puts "YJIT:  #{RubyVM::YJIT.enabled? ? "enabled" : "DISABLED"}"
+puts
+
+# ── Warmup YJIT ─────────────────────────────────────────────────────────────
+warmup_str = "warmup" * 50
+20_000.times { c_digest.call(warmup_str) }
+20_000.times { ruby_digest.call(warmup_str) }
+puts "YJIT warmup complete (20k iterations each)."
+puts
+
+# ── Benchmark ────────────────────────────────────────────────────────────────
+inputs = {
+  "short (5B)"    => "hello",
+  "realistic (134B)" => "{\"template_name\":\"customers\\/account.json\",\"section_id\":\"section-id\",\"block_id\":\"abc\\/\\/dc\\u0026f\",\"setting_id\":\"setting11111-id\"}",
+  "medium (100B)" => "x" * 100,
+  "long (1KB)"    => "y" * 1024,
+  "long (10KB)"   => "z" * 10240,
+}
+
+inputs.each do |label, str|
+  puts "=" * 60
+  puts "  #{label}  (#{str.bytesize} bytes)"
+  puts "=" * 60
+
+  Benchmark.ips do |x|
+    x.config(warmup: 2, time: 5)
+
+    x.report("C extension") { c_digest.call(str) }
+    x.report("Pure Ruby")   { ruby_digest.call(str) }
+
+    x.compare!
+  end
+  puts
+end

data/lib/djb2/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module DJB2
-  VERSION = "0.1.1"
+  VERSION = "0.2.0"
 end

data/lib/djb2.rb

@@ -2,16 +2,66 @@
 
 require_relative "djb2/version"
 
-begin
-  # Load the precompiled version of the library
-  ruby_version = /(\d+\.\d+)/.match(RUBY_VERSION)
-  require "djb2/#{ruby_version}/djb2"
-rescue LoadError
-  # It's important to leave for users that can not or don't want to use the gem with precompiled binaries.
-  require "djb2/djb2"
-end
-
 module DJB2
   class Error < StandardError; end
-  # Your code goes here...
+
+  # Computes the djb2 hash (xor variant) of the given string.
+  #
+  # The hash is computed using 64-bit arithmetic split into two 32-bit
+  # halves to keep all intermediate values within Ruby's Fixnum range,
+  # avoiding Bignum allocation in the hot loop. This makes the
+  # implementation YJIT-friendly: the JIT can emit efficient native
+  # code for the entire loop without any object allocations.
+  #
+  # @param string [String] the string to hash (binary-safe, operates on raw bytes)
+  # @return [Integer] a 64-bit unsigned integer hash value
+  # @raise [TypeError] if the argument is not a String
+  def self.digest(string)
+    raise TypeError, "no implicit conversion of #{string.class} into String" unless string.is_a?(String)
+
+    hi = 0       # upper 32 bits of the hash
+    lo = 5381    # lower 32 bits of the hash
+    i = 0
+    len = string.bytesize
+
+    # Process 4 bytes at a time to reduce loop overhead.
+    stop = len - (len & 3)
+    while i < stop
+      # Multiply (hi:lo) by 33 using: x * 33 = (x << 5) + x
+      # lo half: must mask (lo << 5) to 32 bits BEFORE adding lo, so that
+      # the carry into the hi half is correct (0 or 1, never more).
+      t = ((lo << 5) & 0xFFFFFFFF) + lo
+      # hi half: (hi << 5) can be up to 37 bits, but the total expression
+      # still fits in a Fixnum (< 62 bits), so we only mask at the end.
+      hi = ((hi << 5) + (lo >> 27) + hi + (t >> 32)) & 0xFFFFFFFF
+      # XOR the current byte into the lo half.
+      lo = (t & 0xFFFFFFFF) ^ string.getbyte(i)
+
+      t = ((lo << 5) & 0xFFFFFFFF) + lo
+      hi = ((hi << 5) + (lo >> 27) + hi + (t >> 32)) & 0xFFFFFFFF
+      lo = (t & 0xFFFFFFFF) ^ string.getbyte(i + 1)
+
+      t = ((lo << 5) & 0xFFFFFFFF) + lo
+      hi = ((hi << 5) + (lo >> 27) + hi + (t >> 32)) & 0xFFFFFFFF
+      lo = (t & 0xFFFFFFFF) ^ string.getbyte(i + 2)
+
+      t = ((lo << 5) & 0xFFFFFFFF) + lo
+      hi = ((hi << 5) + (lo >> 27) + hi + (t >> 32)) & 0xFFFFFFFF
+      lo = (t & 0xFFFFFFFF) ^ string.getbyte(i + 3)
+
+      i += 4
+    end
+
+    # Handle remaining 0-3 bytes.
+    while i < len
+      t = ((lo << 5) & 0xFFFFFFFF) + lo
+      hi = ((hi << 5) + (lo >> 27) + hi + (t >> 32)) & 0xFFFFFFFF
+      lo = (t & 0xFFFFFFFF) ^ string.getbyte(i)
+      i += 1
+    end
+
+    # Combine the two halves into a 64-bit result. Using multiply instead
+    # of (hi << 32) | lo avoids a YJIT side exit caused by left shift overflow.
+    hi * 0x100000000 + lo
+  end
 end

metadata

@@ -1,21 +1,18 @@
 --- !ruby/object:Gem::Specification
 name: djb2
 version: !ruby/object:Gem::Version
-  version: 0.1.1
+  version: 0.2.0
 platform: ruby
 authors:
 - Jean Boussier
-autorequire: 
 bindir: exe
 cert_chain: []
-date: 2026-02-19 00:00:00.000000000 Z
+date: 1980-01-02 00:00:00.000000000 Z
 dependencies: []
-description: 
 email:
 - jean.boussier@gmail.com
 executables: []
-extensions:
-- ext/djb2/extconf.rb
+extensions: []
 extra_rdoc_files: []
 files:
 - ".ruby-version"
@@ -23,8 +20,7 @@ files:
 - LICENSE.txt
 - README.md
 - Rakefile
-- ext/djb2/djb2.c
-- ext/djb2/extconf.rb
+- benchmark/benchmark.rb
 - lib/djb2.rb
 - lib/djb2/version.rb
 homepage: https://github.com/Shopify/djb2
@@ -34,7 +30,6 @@ metadata:
   bug_tracker_uri: https://github.com/Shopify/djb2/issues
   source_code_uri: https://github.com/Shopify/djb2
   allowed_push_host: https://rubygems.org
-post_install_message: 
 rdoc_options: []
 require_paths:
 - lib
@@ -42,15 +37,14 @@ required_ruby_version: !ruby/object:Gem::Requirement
   requirements:
   - - ">="
     - !ruby/object:Gem::Version
-      version: 2.6.0
+      version: 3.2.0
 required_rubygems_version: !ruby/object:Gem::Requirement
   requirements:
   - - ">="
     - !ruby/object:Gem::Version
       version: '0'
 requirements: []
-rubygems_version: 3.3.27
-signing_key: 
+rubygems_version: 4.0.3
 specification_version: 4
-summary: Native djb2 hash implementation
+summary: Pure Ruby djb2 hash implementation
 test_files: []

data/ext/djb2/djb2.c

@@ -1,28 +0,0 @@
-#include "ruby.h"
-
-VALUE rb_mDJB2;
-
-uint64_t
-djb_hash_2(const char *string, long length)
-{
-    uint64_t hash = 5381;
-    long index;
-    for (index = 0; index < length; index++) {
-        hash = (hash * 33) ^ (unsigned char)string[index];
-    }
-    return hash;
-}
-
-static VALUE
-rb_djb_hash_2(VALUE self, VALUE str)
-{
-    Check_Type(str, T_STRING);
-    return ULL2NUM(djb_hash_2(RSTRING_PTR(str), RSTRING_LEN(str)));
-}
-
-RUBY_FUNC_EXPORTED void
-Init_djb2(void)
-{
-  rb_mDJB2 = rb_define_module("DJB2");
-  rb_define_singleton_method(rb_mDJB2, "digest", rb_djb_hash_2, 1);
-}

data/ext/djb2/extconf.rb

@@ -1,10 +0,0 @@
-# frozen_string_literal: true
-
-require "mkmf"
-
-# Makes all symbols private by default to avoid unintended conflict
-# with other gems. To explicitly export symbols you can use RUBY_FUNC_EXPORTED
-# selectively, or entirely remove this flag.
-append_cflags("-fvisibility=hidden")
-
-create_makefile("djb2/djb2")