スフィンクスの検索時間を短縮するためのアイデア

Question

私は思考のsphinx gemを使用しています。私のクエリは約45秒で完了しました（インデックスを含むフォルダは1.1GBです）。私は何かが間違って設定されていると推測しています（初めてSphinxのユーザ）。とにかく、あなたが気になるものがあれば教えてください。ここに私の構成は次のとおりです。ここで

define_index do 
    indexes :name 
    indexes :summary 
    indexes :tag_list 

    indexes categories.name, :as => :category_name 

    has "RADIANS(lat)", :as => :latitude, :type => :float 
    has "RADIANS(lng)", :as => :longitude, :type => :float 

    set_property :field_weights => { 
    :name   => 8, 
    :summary  => 6, 
    :category_name => 5, 
    :tag_list  => 3 
    } 
    set_property :delta => ThinkingSphinx::Deltas::ResqueDelta 
    set_property :ignore_chars => %w(' -) 
end 

はサンプルクエリです：

Location.search('Restaurant', 
       :geo => [0.5837843098436726,-1.9560609568879357], 
       :latitude_attr => "latitude", 
       :longitude_attr => "longitude", 
       :with => {"@geodist" => 0.0..4000.0}, 
       :include => :categories, 
       :page => 1, 
       :per_page => 100) 

私のログは示しています

Sphinx Query (43066.3ms) restaurant 
Sphinx Found 467 results 

私はドキュメントを掘りやものをしようとしておこう！

UPDATE：私のdevelopment.sphinx.conf

indexer 
{ 
} 

searchd 
{ 
    listen = 127.0.0.1:9312 
    log = /project_path/log/searchd.log 
    query_log = /project_path/log/searchd.query.log 
    pid_file = /project_path/log/searchd.development.pid 
} 

source location_core_0 
{ 
    type = pgsql 
    sql_host = localhost 
    sql_user = user 
    sql_pass = pass 
    sql_db = db_name 
    sql_query_pre = UPDATE "business_entities" SET "delta" = FALSE WHERE "delta" = TRUE 
    sql_query_pre = SET TIME ZONE 'UTC' 
    sql_query = SELECT "business_entities"."id" * 1::INT8 + 0 AS "id" , "business_entities"."name" AS "name", "business_entities"."summary" AS "summary", "business_entities"."tag_list" AS "tag_list", "business_entities"."id" AS "sphinx_internal_id", 0 AS "sphinx_deleted", CASE COALESCE("business_entities"."type", '') WHEN 'Location' THEN 2817059741 WHEN 'Group' THEN 2885774273 WHEN 'BraintreeBusiness' THEN 28779289 WHEN 'InvoicedBusiness' THEN 1440117572 ELSE 2817059741 END AS "class_crc", COALESCE("business_entities"."type", '') AS "sphinx_internal_class", RADIANS(lat) AS "latitude", RADIANS(lng) AS "longitude" FROM "business_entities" WHERE ("business_entities"."type" = 'Location') AND ("business_entities"."id" >= $start AND "business_entities"."id" <= $end AND "business_entities"."delta" = FALSE AND "business_entities"."type" = 'Location') GROUP BY "business_entities"."id", "business_entities"."name", "business_entities"."summary", "business_entities"."tag_list", "business_entities"."id", "business_entities"."type" 
    sql_query_range = SELECT COALESCE(MIN("id"), 1::bigint), COALESCE(MAX("id"), 1::bigint) FROM "business_entities" WHERE "business_entities"."delta" = FALSE 
    sql_attr_uint = sphinx_internal_id 
    sql_attr_uint = sphinx_deleted 
    sql_attr_uint = class_crc 
    sql_attr_float = latitude 
    sql_attr_float = longitude 
    sql_attr_string = sphinx_internal_class 
    sql_query_info = SELECT * FROM "business_entities" WHERE "id" = (($id - 0)/1) 
} 

index location_core 
{ 
    source = location_core_0 
    path = /project_path/db/sphinx/development/location_core 
    morphology = stem_en 
    charset_type = utf-8 
    ignore_chars = ', - 
    enable_star = 1 
} 

source location_delta_0 : location_core_0 
{ 
    type = pgsql 
    sql_host = localhost 
    sql_user = user 
    sql_pass = pass 
    sql_db = db_name 
    sql_query_pre = 
    sql_query_pre = SET TIME ZONE 'UTC' 
    sql_query = SELECT "business_entities"."id" * 1::INT8 + 0 AS "id" , "business_entities"."name" AS "name", "business_entities"."summary" AS "summary", "business_entities"."tag_list" AS "tag_list", "business_entities"."id" AS "sphinx_internal_id", 0 AS "sphinx_deleted", CASE COALESCE("business_entities"."type", '') WHEN 'Location' THEN 2817059741 WHEN 'Group' THEN 2885774273 WHEN 'BraintreeBusiness' THEN 28779289 WHEN 'InvoicedBusiness' THEN 1440117572 ELSE 2817059741 END AS "class_crc", COALESCE("business_entities"."type", '') AS "sphinx_internal_class", RADIANS(lat) AS "latitude", RADIANS(lng) AS "longitude" FROM "business_entities" WHERE ("business_entities"."type" = 'Location') AND ("business_entities"."id" >= $start AND "business_entities"."id" <= $end AND "business_entities"."delta" = TRUE AND "business_entities"."type" = 'Location') GROUP BY "business_entities"."id", "business_entities"."name", "business_entities"."summary", "business_entities"."tag_list", "business_entities"."id", "business_entities"."type" 
    sql_query_range = SELECT COALESCE(MIN("id"), 1::bigint), COALESCE(MAX("id"), 1::bigint) FROM "business_entities" WHERE "business_entities"."delta" = TRUE 
    sql_attr_uint = sphinx_internal_id 
    sql_attr_uint = sphinx_deleted 
    sql_attr_uint = class_crc 
    sql_attr_float = latitude 
    sql_attr_float = longitude 
    sql_attr_string = sphinx_internal_class 
    sql_query_info = SELECT * FROM "business_entities" WHERE "id" = (($id - 0)/1) 
} 

index location_delta : location_core 
{ 
    source = location_delta_0 
    path = /project_path/db/sphinx/development/location_delta 
} 

index location 
{ 
    type = distributed 
    local = location_delta 
    local = location_core 
} 

Answer 1

私の問題を発見しました - レコードはSTIテーブル内に存在しますが、タイプLocation（Locationには子孫がありません）のインデックスのみにしたいと思います。このテーブルの1,300万レコードのうち、99.99984％（真剣に）はLocationタイプです。 SELECT DISTINCTタイプのFROM business_entitiesクエリが長すぎます（インデックスを使用していても）。

SQL (43647.1ms) SELECT DISTINCT type FROM business_entities 
SQL (39857.7ms) SELECT DISTINCT type FROM business_entities 

Sphinx Query (84173.0ms) restaurant 

だから私猿だけを返すように初期化子でスフィンクスを考えるパッチを適用：ログが84秒続くスフィンクスクエリを報告しましたが、それは本当に問題だった略奪SQLクエリだったので、トリッキーな部分はこれを気づきました。私が気にタイプ：

module ThinkingSphinx 
    class Source 
    module SQL 
     def type_values 
     ['Location'] 
     end 
    end 
    end 
end 

https://gist.github.com/1603565

Answer 2

私はそれが検索のためにとても遅い実行している正確には理由がわからないが、私はあなたのクエリで物事を単純化し、バック複雑さを追加することにより、起動したいです特に何かが原因であるかどうかを調べるために、少しずつ調べてください。だから、最初：おそらくその後

Location.search('Restaurant') 

：

Location.search('Restaurant', :per_page => 100) 

のように。インデックス定義内の:field_weightsにも効果があることを忘れないでください。

私は、あなたがやっていることで特に奇妙なものは見つけていないと思っています。また、43秒間の検索（またはそれに近いもの）は、