bmg 0.16.7 → 0.17.0

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 3ca0b8ab7a34fca3dbf7cb37774b099289a8e15cf42fd381b9ee88f3ac01ff81
-  data.tar.gz: e96b6905b920b0cf70e52227ffe19dcd40ce3a9263deadf032a628348f4c340c
+  metadata.gz: 0d94c440e4e3c70b16735c6d31e3ba50a4be12430d347d9ac8cff6622bc64fb3
+  data.tar.gz: 1443c66b0c1c621f97b2f2f8bea2c522c07151ec4bbef71b45585078635c5b7b
 SHA512:
-  metadata.gz: 6727a3c5b475e5a69765b0f2d4093df4f2aa2ee67364589ee8c0a8e3d368acf50cbe092e7651791e3b13f70ee4d51718921166cac62595998f0b0de2775b1b54
-  data.tar.gz: ef2c1b8eb520f5a4a3a7a24a95f7afcbb96be47e77872dbd55100a0369bd65d7c4d173234d4966591a803106b4937505c5d36ab8ee59d8fde2d6c261988e3ab8
+  metadata.gz: 7eba5332a684097f8cd95a1c28463bbb1e5b004788bf3b3133a55a0c1cf1b5b5335a41924b9ba98cae10fe81f2c4b0ad70bb27d4f97618a8d18a7dbc0f48265f
+  data.tar.gz: 0241a9d8b895ca70df1a7f705ceecce062369810a8b4f60e1f314db0ab2888dc0060ad2f37697b7c83ab11cd2eaefd13e8445d1486d9827f294766bd9f3471aa

data/Gemfile

@@ -1,2 +1,5 @@
 source "https://rubygems.org"
 gemspec
+
+gem "predicate", github: "enspirit/predicate", branch: "placeholders"
+# gem "predicate", path: "../predicate"

data/lib/bmg/operator/image.rb

@@ -12,7 +12,21 @@ module Bmg
 
         # Whether we need to convert each image as an Array,
         # instead of keeping a Relation instance
-        array: false
+        array: false,
+
+        # The strategy to use for actual image algorithm. Default is
+        # :refilter_right. Possible values are:
+        #
+        # - :index_right : builds a memory index with tuples from right, then
+        #   passes left tuples and joins them with the index values.
+        #
+        # - :refilter_right : the left operand is materialized and all
+        #   distinct values collected. The right operand is lately restricted
+        #   to only those matching values. :index_right is then applied on
+        #   resulting operabds. This option only applies when (optimized) `on`
+        #   contains one attribute only. ; it fallbacks on :index_right
+        #   otherwise.
+        strategy: :refilter_right
 
       }
 
@@ -31,7 +45,55 @@ module Bmg
 
     public
 
-      def each
+      def each(*args, &bl)
+        (options[:jit_optimized] ? self : jit_optimize)._each(*args, &bl)
+      end
+
+      def to_ast
+        [ :image, left.to_ast, right.to_ast, as, on, options.dup ]
+      end
+
+    protected
+
+      def _each(*args, &bl)
+        case s = options[:strategy]
+        when :index_right then _each_index_right(*args, &bl)
+        when :refilter_right then _each_refilter_right(*args, &bl)
+        else
+          raise ArgumentError, "Unknown strategy `#{s}`"
+        end
+      end
+
+      def _each_index_right(*args, &bl)
+        left_rel, right_rel = self.left, self.right
+        _each_implem(left_rel, right_rel, *args, &bl)
+      end
+
+      def _each_refilter_right(*args, &bl)
+        left_rel, right_rel = self.left, self.right
+
+        # find matching keys on left and rebind the right
+        # placeholder to them
+        values = left_rel.map{|t| t[on.first] }
+        placeholder = options[:refilter_right][:placeholder]
+        right_rel = right_rel.bind(placeholder => values)
+
+        _each_implem(left_rel, right_rel, *args, &bl)
+      end
+
+      def _each_implem(left_rel, right_rel, *args)
+        # build right index
+        index = build_right_index(right_rel)
+
+        # each left with image from right index
+        left_rel.each do |tuple|
+          key = tuple_project(tuple, on)
+          image = index[key] || (options[:array] ? [] : empty_image)
+          yield tuple.merge(as => image)
+        end
+      end
+
+      def build_right_index(right)
         index = Hash.new{|h,k| h[k] = empty_image }
         right.each_with_object(index) do |t, index|
           key = tuple_project(t, on)
@@ -42,15 +104,54 @@ module Bmg
             ix[k] = v.to_a
           end
         end
-        left.each do |tuple|
-          key = tuple_project(tuple, on)
-          image = index[key] || (options[:array] ? [] : empty_image)
-          yield tuple.merge(as => image)
+        index
+      end
+
+    protected ### jit_optimization
+
+      def jit_optimize
+        case s = options[:strategy]
+        when :index_right then jit_index_right
+        when :refilter_right then jit_refilter_right
+        else
+          raise ArgumentError, "Unknown strategy `#{s}`"
         end
       end
 
-      def to_ast
-        [ :image, left.to_ast, right.to_ast, as, on, options.dup ]
+      def jit_index_right
+        Image.new(
+          type,
+          left,
+          right,
+          as,
+          on,
+          options.merge(jit_optimized: true))
+      end
+
+      def jit_refilter_right
+        ltc = left.type.predicate.constants
+        rtc = right.type.predicate.constants
+        jit_allbut, jit_on = on.partition{|attr|
+          ltc.has_key?(attr) && rtc.has_key?(attr) && ltc[attr] == rtc[attr]
+        }
+        if jit_on.size == 1
+          p = Predicate.placeholder
+          Image.new(
+            type,
+            left.materialize,
+            right.restrict(Predicate.in(jit_on.first, p)).allbut(jit_allbut),
+            as,
+            jit_on,
+            options.merge(jit_optimized: true, refilter_right: { placeholder: p }))
+        else
+          Image.new(
+            type,
+            left,
+            right.allbut(jit_allbut),
+            as,
+            jit_on,
+            options.merge(jit_optimized: true, strategy: :index_right))
+        end
       end
 
     protected ### optimization
@@ -68,7 +169,7 @@ module Bmg
       def _restrict(type, predicate)
         on_as, rest = predicate.and_split([as])
         if rest.tautology?
-          # push none situation: on_as is still the full predicate
+          # push index_right situation: on_as is still the full predicate
           super
         else
           # rest makes no reference to `as` and can be pushed
@@ -126,6 +227,16 @@ module Bmg
         Relation::InMemory.new(image_type, Set.new)
       end
 
+    public
+
+      def to_s
+        options[:jit_optimized] ? super : jit_optimize.to_s
+      end
+
+      def inspect
+        options[:jit_optimized] ? super : jit_optimize.inspect
+      end
+
     end # class Project
   end # module Operator
 end # module Bmg

data/lib/bmg/operator/restrict.rb

@@ -21,6 +21,10 @@ module Bmg
 
     public
 
+      def bind(binding)
+        Restrict.new(type, operand.bind(binding), predicate.bind(binding))
+      end
+
       def each
         @operand.each do |tuple|
           yield(tuple) if @predicate.evaluate(tuple)

data/lib/bmg/operator/shared/binary.rb

@@ -3,9 +3,13 @@ module Bmg
     module Binary
       include Operator
 
+      def bind(binding)
+        _with_operands(left.bind(binding), right.bind(binding))
+      end
+
     protected
 
-      attr_reader :left, :right
+      attr_accessor :left, :right
 
       def _visit(parent, visitor)
         visitor.call(self, parent)
@@ -13,6 +17,13 @@ module Bmg
         right.send(:_visit, self, visitor)
       end
 
+      def _with_operands(left, right)
+        dup.tap{|d|
+          d.left = left
+          d.right = right
+        }
+      end
+
       def operands
         [left, right]
       end

data/lib/bmg/operator/shared/nary.rb

@@ -3,9 +3,17 @@ module Bmg
     module Nary
       include Operator
 
+      def bind(binding)
+        _with_operands(operands.map{|op| op.bind(binding) })
+      end
+
     protected
 
-      attr_reader :operands
+      attr_accessor :operands
+
+      def _with_operands(operands)
+        dup.tap{|d| d.operands = operands }
+      end
 
       def _visit(parent, visitor)
         visitor.call(self, parent)

data/lib/bmg/operator/shared/unary.rb

@@ -3,15 +3,23 @@ module Bmg
     module Unary
       include Operator
 
+      def bind(binding)
+        _with_operand(operand.bind(binding))
+      end
+
     protected
 
-      attr_reader :operand
+      attr_accessor :operand
 
       def _visit(parent, visitor)
         visitor.call(self, parent)
         operand._visit(self, visitor)
       end
 
+      def _with_operand(operand)
+        dup.tap{|d| d.operand = operand }
+      end
+
       def operands
         [operand]
       end

data/lib/bmg/relation.rb

@@ -13,6 +13,10 @@ module Bmg
       Relation::Empty.new(type)
     end
 
+    def bind(binding)
+      self
+    end
+
     def with_typecheck
       dup.tap{|r|
         r.type = r.type.with_typecheck

data/lib/bmg/sequel/translator.rb

@@ -50,7 +50,7 @@ module Bmg
         dataset   = dataset.select(*selection)
         dataset   = dataset.distinct             if sexpr.distinct?
         dataset   = dataset.where(predicate)     if predicate
-        dataset   = dataset.group(*grouping)     if grouping
+        dataset   = dataset.group(grouping)      if grouping
         dataset   = dataset.order_by(*order)     if order
         dataset   = dataset.limit(limit, offset == 0 ? nil : offset) if limit or offset
         dataset

data/lib/bmg/sql/processor.rb

@@ -85,3 +85,4 @@ require_relative 'processor/semi_join'
 require_relative 'processor/flatten'
 require_relative 'processor/requalify'
 require_relative 'processor/summarize'
+require_relative 'processor/bind'

data/lib/bmg/sql/processor/bind.rb

@@ -0,0 +1,23 @@
+module Bmg
+  module Sql
+    class Processor
+      class Bind < Processor
+
+        def initialize(binding, builder)
+          super(builder)
+          @binding = binding
+        end
+
+        def on_select_exp(sexpr)
+          if w = sexpr.where_clause
+            pred = Predicate::Grammar.sexpr(w.predicate.bind(@binding))
+            sexpr.with_update(:where_clause, [ :where_clause, pred ])
+          else
+            sexpr
+          end
+        end
+
+      end # class Bind
+    end # class Processor
+  end # module Sql
+end # module Bmg

data/lib/bmg/sql/processor/requalify.rb

@@ -9,7 +9,7 @@ module Bmg
             h[k.to_s] = builder.next_qualifier!
           }
         end
-        attr_reader :requalify 
+        attr_reader :requalify
 
         alias :on_select_exp :copy_and_apply
         alias :on_missing :copy_and_apply

data/lib/bmg/sql/relation.rb

@@ -18,6 +18,12 @@ module Bmg
 
     public
 
+      def bind(binding)
+        expr = before_use(self.expr)
+        expr = Processor::Bind.new(binding, builder).call(expr)
+        _instance(type, builder, expr)
+      end
+
       def each(&bl)
         raise NotImplementedError
       end

data/lib/bmg/version.rb

@@ -1,8 +1,8 @@
 module Bmg
   module Version
     MAJOR = 0
-    MINOR = 16
-    TINY  = 7
+    MINOR = 17
+    TINY  = 0
   end
   VERSION = "#{Version::MAJOR}.#{Version::MINOR}.#{Version::TINY}"
 end

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: bmg
 version: !ruby/object:Gem::Version
-  version: 0.16.7
+  version: 0.17.0
 platform: ruby
 authors:
 - Bernard Lambeau
 autorequire: 
 bindir: bin
 cert_chain: []
-date: 2020-05-15 00:00:00.000000000 Z
+date: 2020-04-21 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: predicate
@@ -16,20 +16,20 @@ dependencies:
     requirements:
     - - "~>"
       - !ruby/object:Gem::Version
-        version: '2.2'
+        version: '2.3'
     - - ">="
       - !ruby/object:Gem::Version
-        version: 2.2.1
+        version: 2.3.0
   type: :runtime
   prerelease: false
   version_requirements: !ruby/object:Gem::Requirement
     requirements:
     - - "~>"
       - !ruby/object:Gem::Version
-        version: '2.2'
+        version: '2.3'
     - - ">="
       - !ruby/object:Gem::Version
-        version: 2.2.1
+        version: 2.3.0
 - !ruby/object:Gem::Dependency
   name: rake
   requirement: !ruby/object:Gem::Requirement
@@ -222,6 +222,7 @@ files:
 - lib/bmg/sql/nodes/with_spec.rb
 - lib/bmg/sql/processor.rb
 - lib/bmg/sql/processor/all.rb
+- lib/bmg/sql/processor/bind.rb
 - lib/bmg/sql/processor/clip.rb
 - lib/bmg/sql/processor/constants.rb
 - lib/bmg/sql/processor/distinct.rb