freelancing-god-thinking-sphinx 0.9.5

data/lib/thinking_sphinx/configuration.rb

@@ -0,0 +1,275 @@
+module ThinkingSphinx
+  # This class both keeps track of the configuration settings for Sphinx and
+  # also generates the resulting file for Sphinx to use.
+  # 
+  # Here are the default settings, relative to RAILS_ROOT where relevant:
+  #
+  # config file::      config/#{environment}.sphinx.conf
+  # searchd log file:: log/searchd.log
+  # query log file::   log/searchd.query.log
+  # pid file::         log/searchd.#{environment}.pid
+  # searchd files::    db/sphinx/#{environment}/
+  # address::          0.0.0.0 (all)
+  # port::             3312
+  # allow star::       false
+  # mem limit::        64M
+  # max matches::      1000
+  # morphology::       stem_en
+  # charset type::     utf-8
+  # charset table::    nil
+  # ignore chars::     nil
+  #
+  # If you want to change these settings, create a YAML file at
+  # config/sphinx.yml with settings for each environment, in a similar
+  # fashion to database.yml - using the following keys: config_file,
+  # searchd_log_file, query_log_file, pid_file, searchd_file_path, port,
+  # allow_star, mem_limit, max_matches, morphology, charset_type,
+  # charset_table, ignore_chars. I think you've got the idea.
+  # 
+  # Each setting in the YAML file is optional - so only put in the ones you
+  # want to change.
+  #
+  # Keep in mind, if for some particular reason you're using a version of
+  # Sphinx older than 0.9.8 r871 (that's prior to the proper 0.9.8 release),
+  # don't set allow_star to true.
+  # 
+  class Configuration
+    attr_accessor :config_file, :searchd_log_file, :query_log_file,
+      :pid_file, :searchd_file_path, :address, :port, :allow_star, :mem_limit,
+      :max_matches, :morphology, :charset_type, :charset_table, :ignore_chars,
+      :app_root
+    
+    attr_reader :environment
+    
+    # Load in the configuration settings - this will look for config/sphinx.yml
+    # and parse it according to the current environment.
+    # 
+    def initialize(app_root = Dir.pwd)
+      self.app_root          = RAILS_ROOT if defined?(RAILS_ROOT)
+      self.app_root          = Merb.root  if defined?(Merb)
+      self.app_root        ||= app_root
+      
+      self.config_file       = "#{app_root}/config/#{environment}.sphinx.conf"
+      self.searchd_log_file  = "#{app_root}/log/searchd.log"
+      self.query_log_file    = "#{app_root}/log/searchd.query.log"
+      self.pid_file          = "#{app_root}/log/searchd.#{environment}.pid"
+      self.searchd_file_path = "#{app_root}/db/sphinx/#{environment}"
+      self.port              = 3312
+      self.allow_star        = false
+      self.mem_limit         = "64M"
+      self.max_matches       = 1000
+      self.morphology        = "stem_en"
+      self.charset_type      = "utf-8"
+      self.charset_table     = nil
+      self.ignore_chars      = nil
+      
+      parse_config
+    end
+    
+    def self.environment
+      @@environment ||= (
+        defined?(Merb) ? ENV['MERB_ENV'] : ENV['RAILS_ENV']
+      ) || "development"
+    end
+    
+    def environment
+      self.class.environment
+    end
+    
+    # Generate the config file for Sphinx by using all the settings defined and
+    # looping through all the models with indexes to build the relevant
+    # indexer and searchd configuration, and sources and indexes details.
+    #
+    def build(file_path=nil)
+      load_models
+      file_path ||= "#{self.config_file}"
+      database_confs = YAML.load(File.open("#{app_root}/config/database.yml"))
+      database_confs.symbolize_keys!
+      database_conf  = database_confs[environment.to_sym]
+      database_conf.symbolize_keys!
+      
+      open(file_path, "w") do |file|
+        file.write <<-CONFIG
+indexer
+{
+  mem_limit = #{self.mem_limit}
+}
+
+searchd
+{
+  port = #{self.port}
+  log = #{self.searchd_log_file}
+  query_log = #{self.query_log_file}
+  read_timeout = 5
+  max_children = 30
+  pid_file = #{self.pid_file}
+  max_matches = #{self.max_matches}
+}
+        CONFIG
+        
+        ThinkingSphinx.indexed_models.each do |model|
+          model           = model.constantize
+          sources         = []
+          prefixed_fields = []
+          infixed_fields  = []
+          
+          model.indexes.each_with_index do |index, i|
+            # Set up associations and joins
+            index.link!
+            
+            attr_sources = index.attributes.collect { |attrib|
+              attrib.to_sphinx_clause
+            }.join("\n  ")
+            
+            adapter = case index.adapter
+            when :postgres
+              create_array_accum
+              "pgsql"
+            when :mysql
+              "mysql"
+            else
+              raise "Unsupported Database Adapter: Sphinx only supports MySQL and PosgreSQL"
+            end
+            
+            file.write <<-SOURCE
+
+source #{model.name.downcase}_#{i}_core
+{
+  type     = #{adapter}
+  sql_host = #{database_conf[:host] || "localhost"}
+  sql_user = #{database_conf[:username]}
+  sql_pass = #{database_conf[:password]}
+  sql_db   = #{database_conf[:database]}
+
+  sql_query_pre    = #{charset_type == "utf-8" && adapter == "mysql" ? "SET NAMES utf8" : ""}
+  sql_query_pre    = #{index.to_sql_query_pre}
+  sql_query        = #{index.to_sql.gsub(/\n/, ' ')}
+  sql_query_range  = #{index.to_sql_query_range}
+  sql_query_info   = #{index.to_sql_query_info}
+  #{attr_sources}
+}
+            SOURCE
+            
+            if index.delta?
+              file.write <<-SOURCE
+
+source #{model.name.downcase}_#{i}_delta : #{model.name.downcase}_#{i}_core
+{
+  sql_query_pre    = #{charset_type == "utf-8" && adapter == "mysql" ? "SET NAMES utf8" : ""}
+  sql_query        = #{index.to_sql(:delta => true).gsub(/\n/, ' ')}
+  sql_query_range  = #{index.to_sql_query_range :delta => true}
+}
+              SOURCE
+            end
+            sources << "#{model.name.downcase}_#{i}_core"
+          end
+          
+          source_list = sources.collect { |s| "source = #{s}" }.join("\n")
+          delta_list  = source_list.gsub(/_core$/, "_delta")
+          file.write <<-INDEX
+
+index #{model.name.downcase}_core
+{
+  #{source_list}
+  path = #{self.searchd_file_path}/#{model.name.downcase}_core
+  charset_type = #{self.charset_type}
+  INDEX
+          file.puts "  morphology = #{self.morphology}"        unless self.morphology.blank?
+          file.puts "  charset_table  = #{self.charset_table}" unless self.charset_table.nil?
+          file.puts "  ignore_chars   = #{self.ignore_chars}"  unless self.ignore_chars.nil?
+          
+          if self.allow_star
+            file.puts "  enable_star    = 1"
+            file.puts "  min_prefix_len = 1"
+          end
+          
+          file.write("}\n")
+          
+          if model.indexes.any? { |index| index.delta? }
+            file.write <<-INDEX
+
+index #{model.name.downcase}_delta : #{model.name.downcase}_core
+{
+  #{delta_list}
+  path = #{self.searchd_file_path}/#{model.name.downcase}_delta
+}
+
+index #{model.name.downcase}
+{
+  type = distributed
+  local = #{model.name.downcase}_core
+  local = #{model.name.downcase}_delta
+  charset_type = #{self.charset_type}
+}
+            INDEX
+          else
+            file.write <<-INDEX
+index #{model.name.downcase}
+{
+  type = distributed
+  local = #{model.name.downcase}_core
+}
+            INDEX
+          end
+        end
+      end
+    end
+    
+    # Make sure all models are loaded - without reloading any that
+    # ActiveRecord::Base is already aware of (otherwise we start to hit some
+    # messy dependencies issues).
+    # 
+    def load_models
+      Dir["#{app_root}/app/models/**/*.rb"].each do |file|
+        model_name = file.gsub(/^.*\/([\w_]+)\.rb/, '\1')
+        
+        next if model_name.nil?
+        next if ::ActiveRecord::Base.send(:subclasses).detect { |model|
+          model.name == model_name
+        }
+        
+        begin
+          model_name.camelize.constantize
+        rescue NameError, LoadError
+          next
+        end
+      end
+    end
+    
+    private
+    
+    # Parse the config/sphinx.yml file - if it exists - then use the attribute
+    # accessors to set the appropriate values. Nothing too clever.
+    # 
+    def parse_config
+      path = "#{app_root}/config/sphinx.yml"
+      return unless File.exists?(path)
+      
+      conf = YAML.load(File.open(path))[environment]
+      
+      conf.each do |key,value|
+        self.send("#{key}=", value) if self.methods.include?("#{key}=")
+      end unless conf.nil?
+    end
+    
+    def create_array_accum
+      execute "begin"
+      execute "savepoint ts"
+      begin
+        execute <<-SQL
+          CREATE AGGREGATE array_accum (anyelement)
+          (
+              sfunc = array_append,
+              stype = anyarray,
+              initcond = '{}'
+          );
+        SQL
+      rescue
+        raise unless $!.to_s =~ /already exists with same argument types/
+        execute "rollback to savepoint ts"
+      end
+      execute "release savepoint foo"
+      execute "commit"
+    end
+  end
+end

data/lib/thinking_sphinx/field.rb

@@ -0,0 +1,186 @@
+module ThinkingSphinx
+  # Fields - holding the string data which Sphinx indexes for your searches.
+  # 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 field 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 Field
+    attr_accessor :alias, :columns, :sortable, :associations, :model
+    
+    # To create a new field, you'll need to pass in either a single Column
+    # or an array of them, and some (optional) options. The columns are
+    # references to the data that will make up the field.
+    #
+    # Valid options are:
+    # - :as       => :alias_name 
+    # - :sortable => true
+    #
+    # 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.
+    # 
+    # Sortable defaults to false - but is quite useful when set to true, as
+    # it creates an attribute with the same string value (which Sphinx converts
+    # to an integer value), which can be sorted by. Thinking Sphinx is smart
+    # enough to realise that when you specify fields in sort statements, you
+    # mean their respective attributes.
+    #
+    # Here's some examples:
+    #
+    #   Field.new(
+    #     Column.new(:name)
+    #   )
+    #
+    #   Field.new(
+    #     [Column.new(:first_name), Column.new(:last_name)],
+    #     :as => :name, :sortable => true
+    #   )
+    # 
+    #   Field.new(
+    #     [Column.new(:posts, :subject), Column.new(:posts, :content)],
+    #     :as => :posts
+    #   )
+    # 
+    def initialize(columns, options = {})
+      @columns      = Array(columns)
+      @associations = {}
+      
+      @alias        = options[:as]
+      @sortable     = options[:sortable] || false
+    end
+    
+    # Get the part of the SELECT clause related to this field. Don't forget
+    # to set your model and associations first though.
+    #
+    # This will concatenate strings if there's more than one data source or
+    # multiple data values (has_many or has_and_belongs_to_many associations).
+    # 
+    def to_select_sql
+      clause = @columns.collect { |column|
+        column_with_prefix(column)
+      }.join(', ')
+      
+      clause = concatenate(clause) if concat_ws?
+      clause = group_concatenate(clause) if is_many?
+      
+      "#{cast_to_string clause } AS #{quote_column(unique_name)}"
+    end
+    
+    # Get the part of the GROUP BY clause related to this field - if one is
+    # needed. If not, all you'll get back is nil. The latter will happen 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?, ThinkingSphinx.use_group_by_shortcut?
+        nil
+      else
+        @columns.collect { |column|
+          column_with_prefix(column)
+        }
+      end
+    end
+    
+    # Returns the unique name of the field - which is either the alias of
+    # the field, 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
+    
+    private
+    
+    def concatenate(clause)
+      case @model.connection.class.name
+      when "ActiveRecord::ConnectionAdapters::MysqlAdapter"
+        "CONCAT_WS(' ', #{clause})"
+      when "ActiveRecord::ConnectionAdapters::PostgreSQLAdapter"
+        clause.split(', ').join(" || #{separator} || ")
+      else
+        clause
+      end
+    end
+    
+    def group_concatenate(clause)
+      case @model.connection.class.name
+      when "ActiveRecord::ConnectionAdapters::MysqlAdapter"
+        "GROUP_CONCAT(#{clause} SEPARATOR ' ')"
+      when "ActiveRecord::ConnectionAdapters::PostgreSQLAdapter"
+        "array_to_string(array_accum(#{clause}), '#{separator}')"
+      else
+        clause
+      end
+    end
+    
+    def cast_to_string(clause)
+      case @model.connection.class.name
+      when "ActiveRecord::ConnectionAdapters::MysqlAdapter"
+        "CAST(#{clause} AS CHAR)"
+      when "ActiveRecord::ConnectionAdapters::PostgreSQLAdapter"
+        clause
+      else
+        clause
+      end
+    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?
+      @columns.length > 1 || multiple_associations?
+    end
+    
+    # Checks the association tree for each column - if they're all the same,
+    # returns false.
+    # 
+    def multiple_sources?
+      first = associations[@columns.first]
+      
+      !@columns.all? { |col| associations[col] == first }
+    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 associations[column].empty?
+        "#{@model.quoted_table_name}.#{quote_column(column.__name)}"
+      else
+        associations[column].collect { |assoc|
+          "#{@model.connection.quote_table_name(assoc.join.aliased_table_name)}" + 
+          ".#{quote_column(column.__name)}"
+        }.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
+  end
+end