dalli 5.0.0 → 5.0.1

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 16d7d5950137f4a51ae41392eaa576732595ac4020ae7be972d6fb4b33b9861d
-  data.tar.gz: ebb799522259a9a32a86e1dbfd13e810427f787f8110415509288655a5a5460b
+  metadata.gz: 02d0fa949b7a065f86fb3ac7b511a3feacc50b700e82dcb385e0e79c56583a9d
+  data.tar.gz: d72b9e4b014ae1ae0b0d5a1ebe09ea48cdf9fbf0a2cda000efc19f17d5d34368
 SHA512:
-  metadata.gz: 74e32eace9001bd50be0617f10234f1440f86e506facc7247cf6041cf81fa85e966a5721e2a7ff364eb89da19bd46cc92b2631d99d4cbb21aee8240a475a76b5
-  data.tar.gz: 8e2bc294a1fa08a04f0bdfa300394751e63a7c1a21cac6aa1080a37e9491fe33091150047f664607ae42596d0bb2ac60f05c06ad97aee83165de754b9cceeed2
+  metadata.gz: bf26484aa345df2d43a78da570027b68a7fd0dd70d7a62275ebe95a5e29ed36c11a8b1385ff0c71818f55184245e085cd75962510f9f8559aad10b0d284f8d71
+  data.tar.gz: 0b3913a33f61873d6e3da0d62ec643dbf49f679740d3469b7cb826083f02f3b32d37e2a20548634d09cecdc4389da7c5bdc43af530bb48956cb4084f91f10d9e

data/CHANGELOG.md

@@ -1,6 +1,29 @@
 Dalli Changelog
 =====================
 
+5.0.1
+==========
+
+Performance:
+
+- Reduce object allocations in pipelined get response processing (#1072, #1078)
+  - Offset-based `ResponseBuffer`: track a read offset instead of slicing a new string after every parsed response; compact only when the consumed portion exceeds 4KB and more than half the buffer
+  - Inline response processor parsing: avoid intermediate array allocations from `split`-based header parsing
+  - Block-based `pipeline_next_responses`: yield `(key, value, cas)` directly when a block is given, avoiding per-call Hash allocation
+  - `PipelinedGetter`: replace Hash-based socket-to-server mapping with linear scan (faster for typical 1-5 server counts); use `Process.clock_gettime(CLOCK_MONOTONIC)` instead of `Time.now`
+- Add cross-version benchmark script (`bin/compare_versions`) for reproducible performance comparisons across Dalli versions
+
+Bug Fixes:
+
+- Rescue `IOError` in connection manager `write`/`flush` methods (#1075)
+  - Prevents unhandled exceptions when a connection is closed mid-operation
+  - Thanks to Graham Cooper (Shopify) for this fix
+
+Development:
+
+- Add `rubocop-thread_safety` for detecting thread-safety issues (#1076)
+- Add CONTRIBUTING.md with AI contribution policy (#1074)
+
 5.0.0
 ==========
 
@@ -44,6 +67,37 @@ Internal:
 - Removed deprecated binary protocol files and SASL authentication code
 - Removed `require 'set'` (autoloaded in Ruby 3.3+)
 
+4.3.3
+==========
+
+Performance:
+
+- Reduce object allocations in pipelined get response processing (#1072)
+  - Offset-based `ResponseBuffer`: track a read offset instead of slicing a new string after every parsed response; compact only when the consumed portion exceeds 4KB and more than half the buffer
+  - Inline response processor parsing: avoid intermediate array allocations from `split`-based header parsing in both binary and meta protocols
+  - Block-based `pipeline_next_responses`: yield `(key, value, cas)` directly when a block is given, avoiding per-call Hash allocation
+  - `PipelinedGetter`: replace Hash-based socket-to-server mapping with linear scan (faster for typical 1-5 server counts); use `Process.clock_gettime(CLOCK_MONOTONIC)` instead of `Time.now`
+- Add cross-version benchmark script (`bin/compare_versions`) for reproducible performance comparisons across Dalli versions
+
+Bug Fixes:
+
+- Skip OTel integration tests when meta protocol is unavailable (#1072)
+
+4.3.2
+==========
+
+OpenTelemetry:
+
+- Migrate to stable OTel semantic conventions
+  - `db.system` renamed to `db.system.name`
+  - `db.operation` renamed to `db.operation.name`
+  - `server.address` now contains hostname only; `server.port` is a separate integer attribute
+  - `get_with_metadata` and `fetch_with_lock` now include `server.address`/`server.port`
+- Add `db.query.text` span attribute with configurable modes
+  - `:otel_db_statement` option: `:include`, `:obfuscate`, or `nil` (default: omitted)
+- Add `peer.service` span attribute
+  - `:otel_peer_service` option for logical service naming
+
 4.3.1
 ==========
 

data/Gemfile

@@ -22,6 +22,7 @@ group :development, :test do
   gem 'rubocop-minitest'
   gem 'rubocop-performance'
   gem 'rubocop-rake'
+  gem 'rubocop-thread_safety'
   gem 'simplecov'
 end
 

data/README.md

@@ -113,7 +113,7 @@ To install this gem onto your local machine, run `bundle exec rake install`.
 
 ## Contributing
 
-If you have a fix you wish to provide, please fork the code, fix in your local project and then send a pull request on github.  Please ensure that you include a test which verifies your fix and update the [changelog](CHANGELOG.md) with a one sentence description of your fix so you get credit as a contributor.
+Please see [CONTRIBUTING.md](CONTRIBUTING.md) for guidelines on how to contribute, including our policy on AI-authored contributions.
 
 ## Appreciation
 

data/lib/dalli/instrumentation.rb

@@ -61,11 +61,13 @@ module Dalli
       # Uses the library name 'dalli' and current Dalli::VERSION.
       #
       # @return [OpenTelemetry::Trace::Tracer, nil] the tracer or nil if OTel unavailable
+      # rubocop:disable ThreadSafety/ClassInstanceVariable
       def tracer
         return @tracer if defined?(@tracer)
 
         @tracer = (OpenTelemetry.tracer_provider.tracer('dalli', Dalli::VERSION) if defined?(OpenTelemetry))
       end
+      # rubocop:enable ThreadSafety/ClassInstanceVariable
 
       # Returns true if instrumentation is enabled (OpenTelemetry SDK is available).
       #

data/lib/dalli/pid_cache.rb

@@ -13,7 +13,7 @@ module Dalli
         attr_reader :pid
 
         def update!
-          @pid = Process.pid
+          @pid = Process.pid # rubocop:disable ThreadSafety/ClassInstanceVariable
         end
       end
       update!

data/lib/dalli/pipelined_getter.rb

@@ -27,7 +27,7 @@ module Dalli
         # Stores partial results collected during interleaved send phase
         @partial_results = {}
         servers = setup_requests(keys)
-        start_time = Time.now
+        start_time = Process.clock_gettime(Process::CLOCK_MONOTONIC)
 
         # First yield any partial results collected during interleaved send
         yield_partial_results(&block)
@@ -147,7 +147,7 @@ module Dalli
     end
 
     def remaining_time(start, timeout)
-      elapsed = Time.now - start
+      elapsed = Process.clock_gettime(Process::CLOCK_MONOTONIC) - start
       return 0 if elapsed > timeout
 
       timeout - elapsed
@@ -166,8 +166,8 @@ module Dalli
     # Processes responses from a server.  Returns true if there are no
     # additional responses from this server.
     def process_server(server)
-      server.pipeline_next_responses.each_pair do |key, value_list|
-        yield @key_manager.key_without_namespace(key), value_list
+      server.pipeline_next_responses do |key, value, cas|
+        yield @key_manager.key_without_namespace(key), [value, cas]
       end
 
       server.pipeline_complete?
@@ -176,18 +176,13 @@ module Dalli
     def servers_with_response(servers, timeout)
       return [] if servers.empty?
 
-      # TODO: - This is a bit challenging.  Essentially the PipelinedGetter
-      # is a reactor, but without the benefit of a Fiber or separate thread.
-      # My suspicion is that we may want to try and push this down into the
-      # individual servers, but I'm not sure.  For now, we keep the
-      # mapping between the alerted object (the socket) and the
-      # corrresponding server here.
-      server_map = servers.each_with_object({}) { |s, h| h[s.sock] = s }
-
-      readable, = IO.select(server_map.keys, nil, nil, timeout)
+      sockets = servers.map(&:sock)
+      readable, = IO.select(sockets, nil, nil, timeout)
       return [] if readable.nil?
 
-      readable.map { |sock| server_map[sock] }
+      # For typical server counts (1-5), linear scan is faster than
+      # building and looking up a hash map
+      readable.filter_map { |sock| servers.find { |s| s.sock == sock } }
     end
 
     def groups_for_keys(*keys)

data/lib/dalli/protocol/base.rb

@@ -92,10 +92,13 @@ module Dalli
       # repeatedly whenever this server's socket is readable until
       # #pipeline_complete?.
       #
-      # Returns a Hash of kv pairs received.
-      def pipeline_next_responses
+      # When a block is given, yields (key, value, cas) for each response,
+      # avoiding intermediate Hash allocation. Returns nil.
+      # Without a block, returns a Hash of { key => [value, cas] }.
+      # rubocop:disable Metrics/AbcSize, Metrics/CyclomaticComplexity, Metrics/PerceivedComplexity
+      def pipeline_next_responses(&block)
         reconnect_on_pipeline_complete!
-        values = {}
+        values = nil
 
         response_buffer.read
 
@@ -109,16 +112,24 @@ module Dalli
 
           # If the status is ok and the key is not nil, then this is a
           # getkq response with a value that we want to set in the response hash
-          values[key] = [value, cas] unless key.nil?
+          unless key.nil?
+            if block
+              yield key, value, cas
+            else
+              values ||= {}
+              values[key] = [value, cas]
+            end
+          end
 
           # Get the next response from the buffer
           status, cas, key, value = response_buffer.process_single_getk_response
         end
 
-        values
+        values || {}
       rescue SystemCallError, *TIMEOUT_ERRORS, *SSL_ERRORS, EOFError => e
         @connection_manager.error_on_request!(e)
       end
+      # rubocop:enable Metrics/AbcSize, Metrics/CyclomaticComplexity, Metrics/PerceivedComplexity
 
       # Abort current pipelined get. Generally used to signal an external
       # timeout during pipelined get.  The underlying socket is

data/lib/dalli/protocol/connection_manager.rb

@@ -169,13 +169,13 @@ module Dalli
 
       def write(bytes)
         @sock.write(bytes)
-      rescue SystemCallError, *TIMEOUT_ERRORS, *SSL_ERRORS => e
+      rescue SystemCallError, *TIMEOUT_ERRORS, *SSL_ERRORS, IOError => e
         error_on_request!(e)
       end
 
       def flush
         @sock.flush
-      rescue SystemCallError, *TIMEOUT_ERRORS, *SSL_ERRORS => e
+      rescue SystemCallError, *TIMEOUT_ERRORS, *SSL_ERRORS, IOError => e
         error_on_request!(e)
       end
 

data/lib/dalli/protocol/response_buffer.rb

@@ -7,38 +7,39 @@ module Dalli
   module Protocol
     ##
     # Manages the buffer for responses from memcached.
+    # Uses an offset-based approach to avoid string allocations
+    # when advancing through parsed responses.
     ##
     class ResponseBuffer
+      # Compact the buffer when the consumed portion exceeds this
+      # threshold and represents more than half the buffer
+      COMPACT_THRESHOLD = 4096
+
       def initialize(io_source, response_processor)
         @io_source = io_source
         @response_processor = response_processor
         @buffer = nil
+        @offset = 0
       end
 
       def read
         @buffer << @io_source.read_nonblock
       end
 
-      # Attempts to process a single response from the buffer.  Starts
-      # by advancing the buffer to the specified start position
+      # Attempts to process a single response from the buffer,
+      # advancing the offset past the consumed bytes.
       def process_single_getk_response
-        bytes, status, cas, key, value = @response_processor.getk_response_from_buffer(@buffer)
-        advance(bytes)
+        bytes, status, cas, key, value = @response_processor.getk_response_from_buffer(@buffer, @offset)
+        @offset += bytes
+        compact_if_needed
         [status, cas, key, value]
       end
 
-      # Advances the internal response buffer by bytes_to_advance
-      # bytes.  The
-      def advance(bytes_to_advance)
-        return unless bytes_to_advance.positive?
-
-        @buffer = @buffer.byteslice(bytes_to_advance..-1)
-      end
-
       # Resets the internal buffer to an empty state,
       # so that we're ready to read pipelined responses
       def reset
         @buffer = ''.b
+        @offset = 0
       end
 
       # Ensures the buffer is initialized for reading without discarding
@@ -48,16 +49,30 @@ module Dalli
         return if in_progress?
 
         @buffer = ''.b
+        @offset = 0
       end
 
       # Clear the internal response buffer
       def clear
         @buffer = nil
+        @offset = 0
       end
 
       def in_progress?
         !@buffer.nil?
       end
+
+      private
+
+      # Only compact when we've consumed a significant portion of the buffer.
+      # This avoids per-response string allocation while preventing unbounded
+      # memory growth for large pipelines.
+      def compact_if_needed
+        return unless @offset > COMPACT_THRESHOLD && @offset > @buffer.bytesize / 2
+
+        @buffer = @buffer.byteslice(@offset..)
+        @offset = 0
+      end
     end
   end
 end

data/lib/dalli/protocol/response_processor.rb

@@ -147,14 +147,6 @@ module Dalli
           true
         end
 
-        def tokens_from_header_buffer(buf)
-          header = header_from_buffer(buf)
-          tokens = header.split
-          header_len = header.bytesize + TERMINATOR.length
-          body_len = body_len_from_tokens(tokens)
-          [tokens, header_len, body_len]
-        end
-
         def full_response_from_buffer(tokens, body, resp_size)
           value = @value_marshaller.retrieve(body, bitflags_from_tokens(tokens))
           [resp_size, tokens.first == VA, cas_from_tokens(tokens), key_from_tokens(tokens), value]
@@ -170,11 +162,15 @@ module Dalli
         # The remaining three values in the array are the ResponseHeader,
         # key, and value.
         ##
-        def getk_response_from_buffer(buf)
-          # There's no header in the buffer, so don't advance
-          return [0, nil, nil, nil, nil] unless contains_header?(buf)
+        def getk_response_from_buffer(buf, offset = 0)
+          # Find the header terminator starting from offset
+          term_idx = buf.index(TERMINATOR, offset)
+          return [0, nil, nil, nil, nil] unless term_idx
 
-          tokens, header_len, body_len = tokens_from_header_buffer(buf)
+          header = buf.byteslice(offset, term_idx - offset)
+          tokens = header.split
+          header_len = header.bytesize + TERMINATOR.length
+          body_len = body_len_from_tokens(tokens)
 
           # We have a complete response that has no body.
           # This is either the response to the terminating
@@ -185,22 +181,14 @@ module Dalli
           resp_size = header_len + body_len + TERMINATOR.length
           # The header is in the buffer, but the body is not.  As we don't have
           # a complete response, don't advance the buffer
-          return [0, nil, nil, nil, nil] unless buf.bytesize >= resp_size
+          return [0, nil, nil, nil, nil] unless buf.bytesize >= offset + resp_size
 
           # The full response is in our buffer, so parse it and return
           # the values
-          body = buf.slice(header_len, body_len)
+          body = buf.byteslice(offset + header_len, body_len)
           full_response_from_buffer(tokens, body, resp_size)
         end
 
-        def contains_header?(buf)
-          buf.include?(TERMINATOR)
-        end
-
-        def header_from_buffer(buf)
-          buf.split(TERMINATOR, 2).first
-        end
-
         def error_on_unexpected!(expected_codes)
           tokens = next_line_to_tokens
 

data/lib/dalli/socket.rb

@@ -108,12 +108,14 @@ module Dalli
       # Detect and cache whether TCPSocket supports the connect_timeout: keyword argument.
       # Returns false if TCPSocket#initialize has been monkey-patched by gems like
       # socksify or resolv-replace, which don't support keyword arguments.
+      # rubocop:disable ThreadSafety/ClassInstanceVariable
       def self.supports_connect_timeout?
         return @supports_connect_timeout if defined?(@supports_connect_timeout)
 
         @supports_connect_timeout = RUBY_VERSION >= '3.0' &&
                                     ::TCPSocket.instance_method(:initialize).parameters == TCPSOCKET_NATIVE_PARAMETERS
       end
+      # rubocop:enable ThreadSafety/ClassInstanceVariable
 
       def self.create_socket_with_timeout(host, port, options)
         if supports_connect_timeout?
@@ -170,12 +172,14 @@ module Dalli
 
       # Detect and cache the correct pack format for struct timeval on this platform.
       # Different architectures have different sizes for time_t and suseconds_t.
+      # rubocop:disable ThreadSafety/ClassInstanceVariable
       def self.timeval_pack_format(sock)
         @timeval_pack_format ||= begin
           expected_size = sock.getsockopt(::Socket::SOL_SOCKET, ::Socket::SO_RCVTIMEO).data.bytesize
           TIMEVAL_PACK_FORMATS.find { |fmt| TIMEVAL_TEST_VALUES.pack(fmt).bytesize == expected_size } || 'll'
         end
       end
+      # rubocop:enable ThreadSafety/ClassInstanceVariable
 
       def self.pack_timeval(sock, seconds, microseconds)
         [seconds, microseconds].pack(timeval_pack_format(sock))

data/lib/dalli/version.rb

@@ -1,7 +1,7 @@
 # frozen_string_literal: true
 
 module Dalli
-  VERSION = '5.0.0'
+  VERSION = '5.0.1'
 
   MIN_SUPPORTED_MEMCACHED_VERSION = '1.6'
 end

data/lib/dalli.rb

@@ -38,7 +38,7 @@ module Dalli
   QUIET = :dalli_multi
 
   def self.logger
-    @logger ||= rails_logger || default_logger
+    @logger ||= rails_logger || default_logger # rubocop:disable ThreadSafety/ClassInstanceVariable
   end
 
   def self.rails_logger
@@ -54,7 +54,7 @@ module Dalli
   end
 
   def self.logger=(logger)
-    @logger = logger
+    @logger = logger # rubocop:disable ThreadSafety/ClassInstanceVariable
   end
 end
 

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: dalli
 version: !ruby/object:Gem::Version
-  version: 5.0.0
+  version: 5.0.1
 platform: ruby
 authors:
 - Peter M. Goldstein