pg-replication-protocol 0.0.1 → 0.0.3

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: b723f4c6b03faf3c0ef5a638da7c08d4d65e6a0b723d4a7f00578333705803fc
-  data.tar.gz: 87577bceaff85726653296a50fd508113be910ecbe305c120d86c12f05ae1276
+  metadata.gz: 44b3fb2ad8e9ddb30400a0cf8acaf04a70c09d9ae0641e5e475bdf1394c85d5e
+  data.tar.gz: fd766d8b6157af89d017f9dbff382e80278614fe005320c26d7eb89c4f3c4701
 SHA512:
-  metadata.gz: 7693f24a4bee5fd0fdf232c18a126fc9f1728a2bad982c4225ebaa7ec7452158a1ec4ffe95f27b7793d7d2184cdeba6d8eb0561c61984c3687eb67d4ad3715b0
-  data.tar.gz: 97e43186afd2876c52dbaa7d0e8b9235bbdddcc9edcb0e5a5cede9d69617efb08040468de0dcdfbc13966ee93fedcfd6bd9387edb8c63ec173df4622dd3ca6cf
+  metadata.gz: 387313faa12bfa5084ea78920f8d860b048d57050e574793e99c9b8fd922cfb60dff8f843a68ce1ba3d05dc010aa011a3b5816a9a8c703e522411c4fff48c1e9
+  data.tar.gz: 5df28c6e5fe755a34cf1544607b6439f903a6da5c38f355d41c9d4b8e6e1dbd834a01f2d2ab6cc46d0c9abd20ba98d0fc47fcbbda57146394f962ee59c2f22e1

data/.rspec

@@ -0,0 +1 @@
+--require spec_helper

data/Gemfile

@@ -5,3 +5,8 @@ source "https://rubygems.org"
 gemspec
 
 gem "rake", "~> 13.0"
+
+group :test do
+  gem "rspec", "~> 3.0"
+  gem "testcontainers-postgres", require: "testcontainers/postgres"
+end

data/Gemfile.lock

@@ -1,14 +1,37 @@
 PATH
   remote: .
   specs:
-    pg-replication-protocol (0.0.1)
+    pg-replication-protocol (0.0.2)
       pg (~> 1.0)
 
 GEM
   remote: https://rubygems.org/
   specs:
+    diff-lcs (1.5.1)
+    docker-api (2.4.0)
+      excon (>= 0.64.0)
+      multi_json
+    excon (1.2.1)
+    multi_json (1.15.0)
     pg (1.5.9)
     rake (13.2.1)
+    rspec (3.13.0)
+      rspec-core (~> 3.13.0)
+      rspec-expectations (~> 3.13.0)
+      rspec-mocks (~> 3.13.0)
+    rspec-core (3.13.2)
+      rspec-support (~> 3.13.0)
+    rspec-expectations (3.13.3)
+      diff-lcs (>= 1.2.0, < 2.0)
+      rspec-support (~> 3.13.0)
+    rspec-mocks (3.13.2)
+      diff-lcs (>= 1.2.0, < 2.0)
+      rspec-support (~> 3.13.0)
+    rspec-support (3.13.1)
+    testcontainers-core (0.2.0)
+      docker-api (~> 2.2)
+    testcontainers-postgres (0.2.0)
+      testcontainers-core (~> 0.1)
 
 PLATFORMS
   x86_64-linux
@@ -16,6 +39,8 @@ PLATFORMS
 DEPENDENCIES
   pg-replication-protocol!
   rake (~> 13.0)
+  rspec (~> 3.0)
+  testcontainers-postgres
 
 BUNDLED WITH
    2.4.1

data/README.md

@@ -1 +1,63 @@
-# Pg::Replication::Protocol
+# PG::Replication
+
+## Usage
+
+Add to your Gemfile:
+
+```ruby
+gem "pg-replication-protocol", require: "pg/replication"
+```
+
+## Demo
+
+```ruby
+require "pg"
+require "pg/replication"
+
+# Important to create a connection with the `replication: "database"` flag
+connection = PG.connect(..., replication: "database")
+
+# Create a publication and a slot (in a real use case the slot will not be temporary)
+connection.query("CREATE PUBLICATION some_publication FOR ALL TABLES")
+connection.query('CREATE_REPLICATION_SLOT some_slot TEMPORARY LOGICAL "pgoutput"')
+
+# Just a storage for our table relation data
+tables = {}
+
+# Start a pgoutput plugin replication slot message stream
+connection.start_pgoutput_replication_slot(slot, publications).each do |msg|
+  case msg
+  in PG::Replication::PGOutput::Relation(oid:, name:, columns:)
+    # We receive this message on the first row of each table
+    tables[oid] = { name:, columns: }
+
+  in PG::Replication::PGOutput::Begin
+    puts "Transaction start"
+
+  in PG::Replication::PGOutput::Commit
+    puts "Transaction end"
+
+  in PG::Replication::PGOutput::Insert(oid:, new:)
+    puts "Insert #{tables[oid][:name]}"
+    new.zip(tables[oid][:columns]).each do |tuple, col|
+      puts "#{col.name}: #{tuple.data || "NULL"}"
+    end
+
+  in PG::Replication::PGOutput::Update(oid:, new:, old:)
+    puts "Update #{tables[oid][:name]}"
+    if !old.empty? && new != old
+      new.zip(old, tables[oid][:columns]).each do |new, old, col|
+        if new != old
+          puts "Changed #{col.name}: #{old.data || "NULL"} > #{new.data || "NULL"}"
+        end
+      end
+    end
+
+  in PG::Replication::PGOutput::Delete(oid:)
+    puts "Delete #{tables[oid][:name]}"
+
+  else
+    nil
+  end
+end
+```

data/lib/pg/replication/buffer.rb

@@ -1,19 +1,15 @@
 # frozen_string_literal: true
+
+require "delegate"
 require "stringio"
 
 module PG
   module Replication
-    class Buffer
-      attr_reader :buffer
-
+    class Buffer < SimpleDelegator
       def self.from_string(str)
         new(StringIO.new(str))
       end
 
-      def initialize(buffer)
-        @buffer = buffer
-      end
-
       def read_char
         read_int8.chr
       end
@@ -23,19 +19,19 @@ module PG
       end
 
       def read_int8
-        @buffer.read(1).unpack("C").first
+        read(1).unpack("C").first
       end
 
       def read_int16
-        @buffer.read(2).unpack("n").first
+        read(2).unpack("n").first
       end
 
       def read_int32
-        @buffer.read(4).unpack("N").first
+        read(4).unpack("N").first
       end
 
       def read_int64
-        @buffer.read(8).unpack("Q>").first
+        read(8).unpack("Q>").first
       end
 
       def read_timestamp
@@ -54,14 +50,6 @@ module PG
           end
         end
       end
-
-      def read(size = nil)
-        @buffer.read(size)
-      end
-
-      def eof?
-        @buffer.eof?
-      end
     end
   end
 end

data/lib/pg/replication/pg_output.rb

@@ -11,10 +11,14 @@ module PG
       Type = Data.define(:oid, :namespace, :name)
       Insert = Data.define(:oid, :new)
       Update = Data.define(:oid, :key, :old, :new)
-      Delete = Data.define(:oid, :old)
+      Delete = Data.define(:oid, :key, :old)
       Truncate = Data.define(:oid)
       Tuple = Data.define(:type, :data)
-      Column = Data.define(:flags, :name, :oid, :modifier)
+      Column = Data.define(:flags, :name, :oid, :modifier) do
+        def key?
+          flags == 1
+        end
+      end
 
       def self.read_message(buffer)
         case buffer.read_char
@@ -39,7 +43,7 @@ module PG
           )
 
         in "C"
-          buffer.read_int8 # Unused bytes
+          buffer.read_int8 # Unused byte
           PGOutput::Commit.new(
             lsn: buffer.read_int64,
             end_lsn: buffer.read_int64,
@@ -55,21 +59,16 @@ module PG
         in "R"
           PGOutput::Relation.new(
             oid: buffer.read_int32,
-            namespace: buffer.read_cstring,
+            namespace: buffer.read_cstring.then { |ns| ns == "" ? "pg_catalog" : ns },
             name: buffer.read_cstring,
-            replica_identity: buffer.read_int8,
-            columns: case buffer.read_int16
-              in 0
-                []
-              in size
-                size.times.map do
-                  PGOutput::Column.new(
-                    flags: buffer.read_int8,
-                    name: buffer.read_cstring,
-                    oid: buffer.read_int32,
-                    modifier: buffer.read_int32,
-                  )
-                end
+            replica_identity: buffer.read_char,
+            columns: buffer.read_int16.times.map do
+              PGOutput::Column.new(
+                flags: buffer.read_int8,
+                name: buffer.read_cstring,
+                oid: buffer.read_int32,
+                modifier: buffer.read_int32,
+              )
             end
           )
 
@@ -116,14 +115,23 @@ module PG
           )
 
         in "D"
+          oid = buffer.read_int32
+          key = []
+          old = []
+
+          until buffer.eof?
+            case buffer.read_char
+            when "K"
+              key = PGOutput.read_tuples(buffer)
+            when "O"
+              old = PGOutput.read_tuples(buffer)
+            end
+          end
+
           PGOutput::Delete.new(
-            oid: buffer.read_int32,
-            old: case buffer.read_char
-              when "N", "K"
-                PGOutput.read_tuples(buffer)
-              else
-                []
-              end,
+            oid:,
+            key:,
+            old:,
           )
 
         in "T"

data/lib/pg/replication/protocol.rb

@@ -1,11 +1,5 @@
 # frozen_string_literal: true
 
-require "pg"
-require_relative "version"
-require_relative "buffer"
-require_relative "pg_output"
-require_relative "stream"
-
 module PG
   module Replication
     module Protocol
@@ -26,10 +20,7 @@ module PG
             lsn: buffer.read_int64,
             current_lsn: buffer.read_int64,
             server_time: buffer.read_timestamp,
-            data: buffer
-              .read
-              .then { |msg| Buffer.from_string(msg) }
-              .then { |msg| PGOutput.read_message(msg) },
+            data: buffer.read,
           )
 
         in unknown
@@ -37,42 +28,5 @@ module PG
         end
       end
     end
-
-    def start_pgoutput_replication_slot(slot, publications, location: "0/0", keep_alive_time: 10)
-      publications = Array(publications)
-        .map { |name| quote_ident(name) }
-        .map { |name| "'#{name}'" }
-        .join(",")
-
-      query(<<~SQL)
-        START_REPLICATION SLOT
-        #{slot} LOGICAL #{location}
-        ("proto_version" '1', "publication_names" #{publications})
-      SQL
-
-      last_keep_alive = Time.now
-      last_processed_lsn = 0
-
-      stream = PG::Replication::Stream.new(self)
-      stream.lazy.map do |msg|
-        case msg
-        in PG::Replication::Protocol::XLogData(lsn:, data:)
-          last_processed_lsn = lsn
-          data
-        in PG::Replication::Protocol::PrimaryKeepalive(server_time:, asap: true)
-          stream.standby_status_update(write_lsn: last_processed_lsn)
-          next
-        in PG::Replication::Protocol::PrimaryKeepalive(server_time:)
-          now = Time.now
-          if now - last_keep_alive > keep_alive_time
-            stream.standby_status_update(write_lsn: last_processed_lsn)
-            last_keep_alive = now
-          end
-          next
-        end
-      end
-    end
   end
-
-  Connection.send(:include, Replication)
 end

data/lib/pg/replication/version.rb

@@ -2,6 +2,6 @@
 
 module PG
   module Replication
-    VERSION = "0.0.1"
+    VERSION = "0.0.3"
   end
 end

data/lib/pg/replication.rb

@@ -0,0 +1,105 @@
+# frozen_string_literal: true
+
+require "pg"
+require_relative "replication/version"
+require_relative "replication/buffer"
+require_relative "replication/pg_output"
+require_relative "replication/protocol"
+
+module PG
+  module Replication
+    def start_replication_slot(slot, logical: true, location: "0/0", **params)
+      keep_alive_secs = query(<<~SQL).getvalue(0, 0)&.to_i || 10
+        SELECT setting FROM pg_catalog.pg_settings WHERE name = 'wal_receiver_status_interval'
+      SQL
+
+      start_query = "START_REPLICATION SLOT #{slot} #{logical ? "LOGICAL" : "PHYSICAL"} #{location}"
+      unless params.empty?
+        start_query << "("
+        start_query << params
+          .map { |k, v| "#{quote_ident(k.to_s)} '#{escape_string(v.to_s)}'" }
+          .join(", ")
+        start_query << ")"
+      end
+      query(start_query)
+
+      last_processed_lsn = 0
+      last_keep_alive = Time.now
+
+      Enumerator
+        .new do |y|
+          loop do
+            if Time.now - last_keep_alive > keep_alive_secs
+              standby_status_update(write_lsn: last_processed_lsn)
+              last_keep_alive = Time.now
+            end
+
+            consume_input
+            next if is_busy
+
+            case get_copy_data(async: true)
+            in nil
+              get_last_result
+              break
+
+            in false
+              IO.select([socket_io], nil, nil, keep_alive_secs)
+              next
+
+            in data
+              buffer = Buffer.new(StringIO.new(data))
+              y << Protocol.read_message(buffer)
+            end
+          end
+        end
+        .lazy
+        .filter_map do |msg|
+          case msg
+          in Protocol::XLogData(lsn:, data:)
+            last_processed_lsn = lsn
+            standby_status_update(write_lsn: last_processed_lsn)
+            last_keep_alive = Time.now
+            data
+
+          in Protocol::PrimaryKeepalive(server_time:, asap: true)
+            standby_status_update(write_lsn: last_processed_lsn)
+            last_keep_alive = Time.now
+            next
+
+          else
+            next
+          end
+        end
+    end
+
+    def start_pgoutput_replication_slot(slot, publication_names, **kwargs)
+      publication_names = publication_names.join(",")
+
+      start_replication_slot(slot, **kwargs.merge(proto_version: "1", publication_names:))
+        .map { |data| data.force_encoding(internal_encoding) }
+        .map { |data| PGOutput.read_message(Buffer.from_string(data)) }
+    end
+
+    def standby_status_update(
+      write_lsn:,
+      flush_lsn: write_lsn,
+      apply_lsn: write_lsn,
+      timestamp: Time.now,
+      reply: false
+    )
+      msg = [
+        "r".bytes.first,
+        write_lsn,
+        flush_lsn,
+        apply_lsn,
+        (timestamp - Time.new(2_000, 1, 1, 0, 0, 0, 0)) * 10**6,
+        reply ? 1 : 0,
+      ].pack("CQ>Q>Q>Q>C")
+
+      put_copy_data(msg)
+      flush
+    end
+  end
+
+  Connection.send(:include, Replication)
+end

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: pg-replication-protocol
 version: !ruby/object:Gem::Version
-  version: 0.0.1
+  version: 0.0.3
 platform: ruby
 authors:
 - Rodrigo Navarro
 autorequire:
 bindir: bin
 cert_chain: []
-date: 2024-11-13 00:00:00.000000000 Z
+date: 2024-11-29 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: pg
@@ -31,15 +31,16 @@ executables: []
 extensions: []
 extra_rdoc_files: []
 files:
+- ".rspec"
 - Gemfile
 - Gemfile.lock
 - LICENSE.txt
 - README.md
 - Rakefile
+- lib/pg/replication.rb
 - lib/pg/replication/buffer.rb
 - lib/pg/replication/pg_output.rb
 - lib/pg/replication/protocol.rb
-- lib/pg/replication/stream.rb
 - lib/pg/replication/version.rb
 - pg-replication-protocol.gemspec
 homepage:

data/lib/pg/replication/stream.rb

@@ -1,53 +0,0 @@
-# frozen_string_literal: true
-
-module PG
-  module Replication
-    class Stream
-      include Enumerable
-
-      def initialize(connection)
-        @connection = connection
-      end
-
-      def each
-        loop do
-          @connection.consume_input
-
-          next if @connection.is_busy
-
-          case @connection.get_copy_data(async: true)
-          in nil
-            @connection.get_last_result
-            break
-          in false
-            IO.select([@connection.socket_io], nil, nil)
-            next
-          in data
-            buffer = Buffer.new(StringIO.new(data))
-            yield Protocol.read_message(buffer)
-          end
-        end
-      end
-
-      def standby_status_update(
-        write_lsn:,
-        flush_lsn: write_lsn,
-        apply_lsn: write_lsn,
-        timestamp: Time.now,
-        reply: false
-      )
-        msg = [
-          "r".bytes.first,
-          write_lsn,
-          flush_lsn,
-          apply_lsn,
-          (timestamp - Time.new(2000, 1, 1, 0, 0, 0, 0)) * 10**6,
-          reply ? 1 : 0,
-        ].pack("CQ>Q>Q>Q>C")
-
-        @connection.put_copy_data(msg)
-        @connection.flush
-      end
-    end
-  end
-end