dpickett-thinking-sphinx 1.1.4

data/lib/thinking_sphinx/adapters/postgresql_adapter.rb

@@ -0,0 +1,129 @@
+module ThinkingSphinx
+  class PostgreSQLAdapter < AbstractAdapter
+    def setup
+      create_array_accum_function
+      create_crc32_function
+    end
+    
+    def sphinx_identifier
+      "pgsql"
+    end
+    
+    def concatenate(clause, separator = ' ')
+      clause.split(', ').collect { |field|
+        "COALESCE(#{field}, '')"
+      }.join(" || '#{separator}' || ")
+    end
+    
+    def group_concatenate(clause, separator = ' ')
+      "array_to_string(array_accum(#{clause}), '#{separator}')"
+    end
+    
+    def cast_to_string(clause)
+      clause
+    end
+    
+    def cast_to_datetime(clause)
+      "cast(extract(epoch from #{clause}) as int)"
+    end
+    
+    def cast_to_unsigned(clause)
+      clause
+    end
+    
+    def convert_nulls(clause, default = '')
+      default = "'#{default}'" if default.is_a?(String)
+      
+      "COALESCE(#{clause}, #{default})"
+    end
+    
+    def boolean(value)
+      value ? 'TRUE' : 'FALSE'
+    end
+    
+    def crc(clause)
+      "crc32(#{clause})"
+    end
+    
+    def utf8_query_pre
+      nil
+    end
+    
+    def time_difference(diff)
+      "current_timestamp - interval '#{diff} seconds'"
+    end
+    
+    private
+    
+    def execute(command, output_error = false)
+      connection.execute "begin"
+      connection.execute "savepoint ts"
+      begin
+        connection.execute command
+      rescue StandardError => err
+        puts err if output_error
+        connection.execute "rollback to savepoint ts"
+      end
+      connection.execute "release savepoint ts"
+      connection.execute "commit"
+    end
+    
+    def create_array_accum_function
+      if connection.raw_connection.server_version > 80200
+        execute <<-SQL
+          CREATE AGGREGATE array_accum (anyelement)
+          (
+              sfunc = array_append,
+              stype = anyarray,
+              initcond = '{}'
+          );
+        SQL
+      else
+        execute <<-SQL
+          CREATE AGGREGATE array_accum
+          (
+              basetype = anyelement,
+              sfunc = array_append,
+              stype = anyarray,
+              initcond = '{}'
+          );
+        SQL
+      end
+    end
+    
+    def create_crc32_function
+      execute "CREATE LANGUAGE 'plpgsql';"
+      function = <<-SQL
+        CREATE OR REPLACE FUNCTION crc32(word text)
+        RETURNS bigint AS $$
+          DECLARE tmp bigint;
+          DECLARE i int;
+          DECLARE j int;
+          DECLARE word_array bytea;
+          BEGIN
+            i = 0;
+            tmp = 4294967295;
+            word_array = decode(replace(word, E'\\\\', E'\\\\\\\\'), 'escape');
+            LOOP
+              tmp = (tmp # get_byte(word_array, i))::bigint;
+              i = i + 1;
+              j = 0;
+              LOOP
+                tmp = ((tmp >> 1) # (3988292384 * (tmp & 1)))::bigint;
+                j = j + 1;
+                IF j >= 8 THEN
+                  EXIT;
+                END IF;
+              END LOOP;
+              IF i >= char_length(word) THEN
+                EXIT;
+              END IF;
+            END LOOP;
+            return (tmp # 4294967295);
+          END
+        $$ IMMUTABLE STRICT LANGUAGE plpgsql;
+      SQL
+      execute function, true
+    end
+  end
+end

data/lib/thinking_sphinx/association.rb

@@ -0,0 +1,144 @@
+module ThinkingSphinx
+  # Association tracks a specific reflection and join to reference data that
+  # isn't in the base model. Very much an internal class for Thinking Sphinx -
+  # perhaps because I feel it's not as strong (or simple) as most of the rest.
+  # 
+  class Association
+    attr_accessor :parent, :reflection, :join
+    
+    # Create a new association by passing in the parent association, and the
+    # corresponding reflection instance. If there is no parent, pass in nil.
+    # 
+    #   top   = Association.new nil, top_reflection
+    #   child = Association.new top, child_reflection
+    # 
+    def initialize(parent, reflection)
+      @parent, @reflection = parent, reflection
+      @children = {}
+    end
+    
+    # Get the children associations for a given association name. The only time
+    # that there'll actually be more than one association is when the
+    # relationship is polymorphic. To keep things simple though, it will always
+    # be an Array that gets returned (an empty one if no matches).
+    #
+    #   # where pages is an association on the class tied to the reflection.
+    #   association.children(:pages)
+    # 
+    def children(assoc)
+      @children[assoc] ||= Association.children(@reflection.klass, assoc, self)
+    end
+    
+    # Get the children associations for a given class, association name and
+    # parent association. Much like the instance method of the same name, it
+    # will return an empty array if no associations have the name, and only
+    # have multiple association instances if the underlying relationship is
+    # polymorphic.
+    # 
+    #   Association.children(User, :pages, user_association)
+    # 
+    def self.children(klass, assoc, parent=nil)
+      ref = klass.reflect_on_association(assoc)
+      
+      return [] if ref.nil?
+      return [Association.new(parent, ref)] unless ref.options[:polymorphic]
+      
+      # association is polymorphic - create associations for each
+      # non-polymorphic reflection.
+      polymorphic_classes(ref).collect { |klass|
+        Association.new parent, ::ActiveRecord::Reflection::AssociationReflection.new(
+          ref.macro,
+          "#{ref.name}_#{klass.name}".to_sym,
+          casted_options(klass, ref),
+          ref.active_record
+        )
+      }
+    end
+    
+    # Link up the join for this model from a base join - and set parent
+    # associations' joins recursively.
+    #
+    def join_to(base_join)
+      parent.join_to(base_join) if parent && parent.join.nil?
+      
+      @join ||= ::ActiveRecord::Associations::ClassMethods::JoinDependency::JoinAssociation.new(
+        @reflection, base_join, parent ? parent.join : base_join.joins.first
+      )
+    end
+    
+    # Returns the association's join SQL statements - and it replaces
+    # ::ts_join_alias:: with the aliased table name so the generated reflection
+    # join conditions avoid column name collisions.
+    # 
+    def to_sql
+      @join.association_join.gsub(/::ts_join_alias::/,
+        "#{@reflection.klass.connection.quote_table_name(@join.parent.aliased_table_name)}"
+      )
+    end
+    
+    # Returns true if the association - or a parent - is a has_many or
+    # has_and_belongs_to_many.
+    # 
+    def is_many?
+      case @reflection.macro
+      when :has_many, :has_and_belongs_to_many
+        true
+      else
+        @parent ? @parent.is_many? : false
+      end
+    end
+    
+    # Returns an array of all the associations that lead to this one - starting
+    # with the top level all the way to the current association object.
+    # 
+    def ancestors
+      (parent ? parent.ancestors : []) << self
+    end
+    
+    def has_column?(column)
+      @reflection.klass.column_names.include?(column.to_s)
+    end
+    
+    private
+    
+    # Returns all the objects that could be currently instantiated from a
+    # polymorphic association. This is pretty damn fast if there's an index on
+    # the foreign type column - but if there isn't, it can take a while if you
+    # have a lot of data.
+    # 
+    def self.polymorphic_classes(ref)
+      ref.active_record.connection.select_all(
+        "SELECT DISTINCT #{ref.options[:foreign_type]} " +
+        "FROM #{ref.active_record.table_name} " +
+        "WHERE #{ref.options[:foreign_type]} IS NOT NULL"
+      ).collect { |row|
+        row[ref.options[:foreign_type]].constantize
+      }
+    end
+    
+    # Returns a new set of options for an association that mimics an existing
+    # polymorphic relationship for a specific class. It adds a condition to
+    # filter by the appropriate object.
+    # 
+    def self.casted_options(klass, ref)
+      options = ref.options.clone
+      options[:polymorphic]   = nil
+      options[:class_name]    = klass.name
+      options[:foreign_key] ||= "#{ref.name}_id"
+      
+      quoted_foreign_type = klass.connection.quote_column_name ref.options[:foreign_type]
+      case options[:conditions]
+      when nil
+        options[:conditions] = "::ts_join_alias::.#{quoted_foreign_type} = '#{klass.name}'"
+      when Array
+        options[:conditions] << "::ts_join_alias::.#{quoted_foreign_type} = '#{klass.name}'"
+      when Hash
+        options[:conditions].merge!(ref.options[:foreign_type] => klass.name)
+      else
+        options[:conditions] << " AND ::ts_join_alias::.#{quoted_foreign_type} = '#{klass.name}'"
+      end
+      
+      options
+    end
+  end
+end

data/lib/thinking_sphinx/attribute.rb

@@ -0,0 +1,254 @@
+module ThinkingSphinx
+  # Attributes - eternally useful when it comes to filtering, sorting or
+  # grouping. This class isn't really useful to you unless you're hacking
+  # around with the internals of Thinking Sphinx - but hey, don't let that
+  # stop you.
+  #
+  # One key thing to remember - if you're using the attribute manually to
+  # generate SQL statements, you'll need to set the base model, and all the
+  # associations. Which can get messy. Use Index.link!, it really helps.
+  # 
+  class Attribute
+    attr_accessor :alias, :columns, :associations, :model, :faceted
+    
+    # To create a new attribute, you'll need to pass in either a single Column
+    # or an array of them, and some (optional) options.
+    #
+    # Valid options are:
+    # - :as   => :alias_name 
+    # - :type => :attribute_type
+    #
+    # Alias is only required in three circumstances: when there's
+    # another attribute or field with the same name, when the column name is
+    # 'id', or when there's more than one column.
+    # 
+    # Type is not required, unless you want to force a column to be a certain
+    # type (but keep in mind the value will not be CASTed in the SQL
+    # statements). The only time you really need to use this is when the type
+    # can't be figured out by the column - ie: when not actually using a
+    # database column as your source.
+    # 
+    # Example usage:
+    #
+    #   Attribute.new(
+    #     Column.new(:created_at)
+    #   )
+    #
+    #   Attribute.new(
+    #     Column.new(:posts, :id),
+    #     :as => :post_ids
+    #   )
+    #
+    #   Attribute.new(
+    #     [Column.new(:pages, :id), Column.new(:articles, :id)],
+    #     :as => :content_ids
+    #   )
+    #
+    #   Attribute.new(
+    #     Column.new("NOW()"),
+    #     :as   => :indexed_at,
+    #     :type => :datetime
+    #   )
+    #
+    # If you're creating attributes for latitude and longitude, don't forget
+    # that Sphinx expects these values to be in radians.
+    #  
+    def initialize(columns, options = {})
+      @columns      = Array(columns)
+      @associations = {}
+      
+      raise "Cannot define a field with no columns. Maybe you are trying to index a field with a reserved name (id, name). You can fix this error by using a symbol rather than a bare name (:id instead of id)." if @columns.empty? || @columns.any? { |column| !column.respond_to?(:__stack) }
+      
+      @alias    = options[:as]
+      @type     = options[:type]
+      @faceted  = options[:facet]
+    end
+    
+    # Get the part of the SELECT clause related to this attribute. Don't forget
+    # to set your model and associations first though.
+    #
+    # This will concatenate strings and arrays of integers, and convert
+    # datetimes to timestamps, as needed.
+    # 
+    def to_select_sql
+      clause = @columns.collect { |column|
+        column_with_prefix(column)
+      }.join(', ')
+      
+      separator = all_ints? ? ',' : ' '
+      
+      clause = adapter.concatenate(clause, separator)       if concat_ws?
+      clause = adapter.group_concatenate(clause, separator) if is_many?
+      clause = adapter.cast_to_datetime(clause)             if type == :datetime
+      clause = adapter.convert_nulls(clause)                if type == :string
+      
+      "#{clause} AS #{quote_column(unique_name)}"
+    end
+    
+    # Get the part of the GROUP BY clause related to this attribute - if one is
+    # needed. If not, all you'll get back is nil. The latter will happen if
+    # there isn't actually a real column to get data from, or if there's
+    # multiple data values (read: a has_many or has_and_belongs_to_many
+    # association).
+    # 
+    def to_group_sql
+      case
+      when is_many?, is_string?, ThinkingSphinx.use_group_by_shortcut?
+        nil
+      else
+        @columns.collect { |column|
+          column_with_prefix(column)
+        }
+      end
+    end
+    
+    def type_to_config
+      {
+        :multi    => :sql_attr_multi,
+        :datetime => :sql_attr_timestamp,
+        :string   => :sql_attr_str2ordinal,
+        :float    => :sql_attr_float,
+        :boolean  => :sql_attr_bool,
+        :integer  => :sql_attr_uint
+      }[type]
+    end
+    
+    def config_value
+      if type == :multi
+        "uint #{unique_name} from field"
+      else
+        unique_name
+      end
+    end
+    
+    # Returns the unique name of the attribute - which is either the alias of
+    # the attribute, or the name of the only column - if there is only one. If
+    # there isn't, there should be an alias. Else things probably won't work.
+    # Consider yourself warned.
+    # 
+    def unique_name
+      if @columns.length == 1
+        @alias || @columns.first.__name
+      else
+        @alias
+      end
+    end
+    
+    # Returns the type of the column. If that's not already set, it returns
+    # :multi if there's the possibility of more than one value, :string if
+    # there's more than one association, otherwise it figures out what the
+    # actual column's datatype is and returns that.
+    def type
+      @type ||= case
+      when is_many?
+        :multi
+      when @associations.values.flatten.length > 1
+        :string
+      else
+        translated_type_from_database
+      end
+    end
+    
+    def to_facet
+      return nil unless @faceted
+      
+      ThinkingSphinx::Facet.new(self)
+    end
+    
+    private
+    
+    def adapter
+      @adapter ||= @model.sphinx_database_adapter
+    end
+    
+    def quote_column(column)
+      @model.connection.quote_column_name(column)
+    end
+    
+    # Indication of whether the columns should be concatenated with a space
+    # between each value. True if there's either multiple sources or multiple
+    # associations.
+    # 
+    def concat_ws?
+      multiple_associations? || @columns.length > 1
+    end
+        
+    # Checks whether any column requires multiple associations (which only
+    # happens for polymorphic situations).
+    # 
+    def multiple_associations?
+      associations.any? { |col,assocs| assocs.length > 1 }
+    end
+    
+    # Builds a column reference tied to the appropriate associations. This
+    # dives into the associations hash and their corresponding joins to
+    # figure out how to correctly reference a column in SQL.
+    # 
+    def column_with_prefix(column)
+      if column.is_string?
+        column.__name
+      elsif associations[column].empty?
+        "#{@model.quoted_table_name}.#{quote_column(column.__name)}"
+      else
+        associations[column].collect { |assoc|
+          assoc.has_column?(column.__name) ?
+          "#{@model.connection.quote_table_name(assoc.join.aliased_table_name)}" + 
+          ".#{quote_column(column.__name)}" :
+          nil
+        }.compact.join(', ')
+      end
+    end
+    
+    # Could there be more than one value related to the parent record? If so,
+    # then this will return true. If not, false. It's that simple.
+    # 
+    def is_many?
+      associations.values.flatten.any? { |assoc| assoc.is_many? }
+    end
+    
+    # Returns true if any of the columns are string values, instead of database
+    # column references.
+    def is_string?
+      columns.all? { |col| col.is_string? }
+    end
+    
+    def all_ints?
+      @columns.all? { |col|
+        klasses = @associations[col].empty? ? [@model] :
+          @associations[col].collect { |assoc| assoc.reflection.klass }
+        klasses.all? { |klass|
+          column = klass.columns.detect { |column| column.name == col.__name.to_s }
+          !column.nil? && column.type == :integer
+        }
+      }
+    end
+    
+    def type_from_database
+      klass = @associations.values.flatten.first ? 
+        @associations.values.flatten.first.reflection.klass : @model
+      
+      klass.columns.detect { |col|
+        @columns.collect { |c| c.__name.to_s }.include? col.name
+      }.type
+    end
+    
+    def translated_type_from_database
+      case type_from_db = type_from_database
+      when :datetime, :string, :float, :boolean, :integer
+        type_from_db
+      when :decimal
+        :float
+      when :timestamp, :date
+        :datetime
+      else
+        raise <<-MESSAGE
+
+Cannot automatically map column type #{type_from_db} to an equivalent Sphinx
+type (integer, float, boolean, datetime, string as ordinal). You could try to
+explicitly convert the column's value in your define_index block:
+  has "CAST(column AS INT)", :type => :integer, :as => :column
+        MESSAGE
+      end
+    end
+  end
+end

data/lib/thinking_sphinx/class_facet.rb

@@ -0,0 +1,20 @@
+module ThinkingSphinx
+  class ClassFacet < ThinkingSphinx::Facet
+    def name
+      :class
+    end
+    
+    def attribute_name
+      "class_crc"
+    end
+    
+    def value(object, attribute_value)
+      object.class.name
+      # ThinkingSphinx.indexed_models.each do |i|
+      #   return i if i.to_crc32 == attribute_value
+      # end
+      # 
+      # raise "Unknown class crc"
+    end
+  end
+end

data/lib/thinking_sphinx/collection.rb

@@ -0,0 +1,142 @@
+module ThinkingSphinx
+  class Collection < ::Array
+    attr_reader :total_entries, :total_pages, :current_page, :per_page
+    attr_accessor :results
+
+    # Compatibility with older versions of will_paginate
+    alias_method :page_count, :total_pages
+    
+    def initialize(page, per_page, entries, total_entries)
+      @current_page, @per_page, @total_entries = page, per_page, total_entries
+      
+      @total_pages = (entries / @per_page.to_f).ceil
+    end
+    
+    def self.ids_from_results(results, page, limit, options)
+      collection = self.new(page, limit,
+        results[:total] || 0, results[:total_found] || 0
+      )
+      collection.results = results
+      collection.replace results[:matches].collect { |match|
+        match[:attributes]["sphinx_internal_id"]
+      }
+      return collection
+    end
+    
+    def self.create_from_results(results, page, limit, options)
+      collection = self.new(page, limit,
+        results[:total] || 0, results[:total_found] || 0
+      )
+      collection.results = results
+      collection.replace instances_from_matches(results[:matches], options)
+      return collection
+    end
+    
+    def self.instances_from_matches(matches, options = {})
+      if klass = options[:class]
+        instances_from_class klass, matches, options
+      else
+        instances_from_classes matches, options
+      end
+    end
+    
+    def self.instances_from_class(klass, matches, options = {})
+      index_options = klass.sphinx_index_options
+
+      ids = matches.collect { |match| match[:attributes]["sphinx_internal_id"] }
+      instances = ids.length > 0 ? klass.find(
+        :all,
+        :conditions => {klass.primary_key.to_sym => ids},
+        :include    => (options[:include] || index_options[:include]),
+        :select     => (options[:select]  || index_options[:select])
+      ) : []
+
+      # Raise an exception if we find records in Sphinx but not in the DB, so
+      # the search method can retry without them. See 
+      # ThinkingSphinx::Search.retry_search_on_stale_index.
+      if options[:raise_on_stale] && instances.length < ids.length
+        stale_ids = ids - instances.map {|i| i.id }
+        raise StaleIdsException, stale_ids
+      end
+
+      ids.collect { |obj_id|
+        instances.detect { |obj| obj.id == obj_id }
+      }
+    end
+    
+    # Group results by class and call #find(:all) once for each group to reduce
+    # the number of #find's in multi-model searches.
+    # 
+    def self.instances_from_classes(matches, options = {})
+      groups = matches.group_by { |match| match[:attributes]["class_crc"] }
+      groups.each do |crc, group|
+        group.replace(
+          instances_from_class(class_from_crc(crc), group, options)
+        )
+      end
+      
+      matches.collect do |match|
+        groups.detect { |crc, group|
+          crc == match[:attributes]["class_crc"]
+        }[1].detect { |obj|
+          obj.id == match[:attributes]["sphinx_internal_id"]
+        }
+      end
+    end
+    
+    def self.class_from_crc(crc)
+      @@models_by_crc ||= ThinkingSphinx.indexed_models.inject({}) do |hash, model|
+        hash[model.constantize.to_crc32] = model
+        model.constantize.subclasses.each { |subclass|
+          hash[subclass.to_crc32] = subclass.name
+        }
+        hash
+      end
+      @@models_by_crc[crc].constantize
+    end
+    
+    def previous_page
+      current_page > 1 ? (current_page - 1) : nil
+    end
+    
+    def next_page
+      current_page < total_pages ? (current_page + 1): nil
+    end
+    
+    def offset
+      (current_page - 1) * @per_page
+    end
+    
+    def method_missing(method, *args, &block)
+      super unless method.to_s[/^each_with_.*/]
+      
+      each_with_attribute method.to_s.gsub(/^each_with_/, ''), &block
+    end
+    
+    def each_with_groupby_and_count(&block)
+      results[:matches].each_with_index do |match, index|
+        yield self[index], match[:attributes]["@groupby"], match[:attributes]["@count"]
+      end
+    end
+    
+    def each_with_attribute(attribute, &block)
+      results[:matches].each_with_index do |match, index|
+        yield self[index], (match[:attributes][attribute] || match[:attributes]["@#{attribute}"])
+      end
+    end
+    
+    def each_with_weighting(&block)
+      results[:matches].each_with_index do |match, index|
+        yield self[index], match[:weight]
+      end
+    end
+    
+    def inject_with_groupby_and_count(initial = nil, &block)
+      index = -1
+      results[:matches].inject(initial) do |memo, match|
+        index += 1
+        yield memo, self[index], match[:attributes]["@groupby"], match[:attributes]["@count"]
+      end
+    end
+  end
+end