HAYASHI Kentaro
null+****@clear*****
Fri Dec 14 14:58:02 JST 2012
HAYASHI Kentaro 2012-12-14 14:58:02 +0900 (Fri, 14 Dec 2012) New Revision: 37e8779e419b5f7ec3bacbd268b65ef8405ce823 https://github.com/groonga/groonga.github.com/commit/37e8779e419b5f7ec3bacbd268b65ef8405ce823 Log: doc: fix incorrect description and examples about range search Modified files: en/_posts/2012-11-29-release.textile ja/_posts/2012-11-29-release.textile Modified: en/_posts/2012-11-29-release.textile (+69 -24) =================================================================== --- en/_posts/2012-11-29-release.textile 2012-12-13 18:09:22 +0900 (d14e786) +++ en/_posts/2012-11-29-release.textile 2012-12-14 14:58:02 +0900 (b0326fe) @@ -161,59 +161,104 @@ Now, you can search articles which contains specific keywords as a comment. h3. Supported range search by using index This release began to support range search by using index. -For example, you can search date after the specified one for the column which stores the value of Time type. +As a result, you can search in a short time by contrast to previous release. Here is the sample how to use this feature. Schema definition: <pre> - table_create Users TABLE_HASH_KEY ShortText - column_create Users birthday COLUMN_SCALAR Time - table_create Birthdays TABLE_PAT_KEY Time - column_create Birthdays users_birthday COLUMN_INDEX Users birthday + table_create Shops TABLE_HASH_KEY ShortText + column_create Shops ranking COLUMN_SCALAR UInt32 + + table_create Rankings TABLE_PAT_KEY UInt32 + column_create Rankings shops_ranking COLUMN_INDEX Shops ranking </pre> -Sample data: +Sample data (ranking data about 10,000,000 shops): <pre> - load --table Users + load --table Shops [ - {"_key": "Alice", "birthday": "1992-02-09 00:00:00"}, - {"_key": "Bob", "birthday": "1988-01-04 00:00:00"}, - {"_key": "Carlos", "birthday": "1982-12-29 00:00:00"} + ... + {"_key": "Shop98", "ranking": 98}, + {"_key": "Shop99", "ranking": 99}, + {"_key": "Carlos's shop", "ranking": 100}, + {"_key": "Alice's shop", "ranking": 101}, + {"_key": "Bob's shop", "ranking": 102}, + {"_key": "Shop103", "ranking": 103}, + {"_key": "Shop104", "ranking": 104}, + {"_key": "Shop105", "ranking": 105}, + ... ] </pre> -Now, registered birthdays, you can search date after the specified one. +Now, registered shop name as a key, the value of ranking. -Here is the sample query to search person who is younger than Bob. +Here is the sample query to search three shops which is higher value of ranking than Alice's shop. -In range search, you can specify 'younger' expression as 'birthday > "1988-01-04 00:00:00"' in this case. +In range search, you can specify 'higher' expression as @'ranking >= 98 && ranking < 101'@ in this case. Here is concrete example. <pre> - select Users --filter 'birthday > "1988-01-04 00:00:00"' + select Shops --filter 'ranking >= 98 && ranking < 101' [ - [0,1354069642.52512,0.000299692153930664], + [0,1355461536.3541,3.55056285858154], [ [ - [1], - [ - ["_id","UInt32"], - ["_key","ShortText"], - ["birthday","Time"] - ], - [1,"Alice",697561200.0] + [3], + [["_id","UInt32"],["_key","ShortText"],["ranking","UInt32"]], + [98,"Shop98",98], + [99,"Shop99",99], + [100,"Carlos's shop",100] + ] + ] + ] +</pre> + +<pre> + select Shops --filter 'ranking >= 98 && ranking < 101' + [ + [0,1355461395.8912,2.43025946617126], + [ + [ + [3], + [["_id","UInt32"],["_key","ShortText"],["ranking","UInt32"]], + [98,"Shop98",98], + [99,"Shop99",99], + [100,"Carlos's shop",100] ] ] ] </pre> -Alice is the only person who is younger than Bob, you see that search results is valid. +The search result is same, but the execution time is different. + +<pre> + [0,1355461536.3541,3.55056285858154], +</pre> + +In groonga 2.0.8, it takes 3.55056285858154 seconds. + +<pre> + [0,1355461395.8912,2.43025946617126], +</pre> + +In groonga 2.0.9, it takes 2.43025946617126 seconds. + +See "Output Format":http://groonga.org/docs/reference/command/output_format.html about the output of groonga command details. + +|Version of groonga|groonga 2.0.8|groonga 2.0.9| +|Execution time(seconds)|3.55056285858154|2.43025946617126| + +By upgrading 2.0.8 to 2.0.9, you can see the execution time is clipped to about 70%. + +Here is the measurement environment: + +|CPU|Intel(R) Core i7-2640M CPU****@2*****| +|Memory|8GB| -In range search, you can specify @>@, @<@, @>=@, @<=@ operators. h3. Supported calculation across meridian, equator, the date line by geo_distance() function Modified: ja/_posts/2012-11-29-release.textile (+73 -25) =================================================================== --- ja/_posts/2012-11-29-release.textile 2012-12-13 18:09:22 +0900 (be55612) +++ ja/_posts/2012-11-29-release.textile 2012-12-14 14:58:02 +0900 (6a4f54f) @@ -159,55 +159,103 @@ h3. インデックスを利用した高速な指定範囲検索のサポート 今回のリリースでは、範囲を指定した検索をインデックスを使って高速に行えるようになりました。 -例えば、Time型のカラムで指定した日以前の結果を高速に検索することができるようになりました。 - -どう使うかのサンプルを誕生日を例に示します。 +どれくらい高速になったかという例を簡単なサンプルで示します。 サンプルのスキーマ定義は以下の通りです。 <pre> - table_create Users TABLE_HASH_KEY ShortText - column_create Users birthday COLUMN_SCALAR Time - table_create Birthdays TABLE_PAT_KEY Time - column_create Birthdays users_birthday COLUMN_INDEX Users birthday + table_create Shops TABLE_HASH_KEY ShortText + column_create Shops ranking COLUMN_SCALAR UInt32 + + table_create Rankings TABLE_PAT_KEY UInt32 + column_create Rankings shops_ranking COLUMN_INDEX Shops ranking </pre> -以下のように個人の誕生日を登録します。 +以下のようにしてお店とそのランキングの値を登録します。高速化した結果をよりわかりやすく示すために、1000万件ほどのデータを登録します。 <pre> - load --table Users + load --table Shops [ - {"_key": "Alice", "birthday": "1992-02-09 00:00:00"}, - {"_key": "Bob", "birthday": "1988-01-04 00:00:00"}, - {"_key": "Carlos", "birthday": "1982-12-29 00:00:00"} + ... + {"_key": "Shop98", "ranking": 98}, + {"_key": "Shop99", "ranking": 99}, + {"_key": "Carlos's shop", "ranking": 100}, + {"_key": "Alice's shop", "ranking": 101}, + {"_key": "Bob's shop", "ranking": 102}, + {"_key": "Shop103", "ranking": 103}, + {"_key": "Shop104", "ranking": 104}, + {"_key": "Shop105", "ranking": 105}, + ... ] </pre> -誕生日のデータを登録できたので、実際に検索してみます。ここでは、Bobの誕生日より後の誕生日の人を検索してみます。 +お店とランキングの値が登録できたので、実際に検索してみます。ここでは、Aliceのお店よりランキングが上のお店3件を検索してみます。 -誕生日はTime型のカラムに登録してあるので、範囲の指定にはbirthdayに対してBobの誕生日より後というのを @'birthday > "1988-01-04 00:00:00"'@ として表現します。 +範囲の指定にはrankingカラムに対して @'ranking >= 98 && ranking < 101'@ として表現します。 -実際の検索結果は以下のようになります。 +実際の検索結果は以下のようになります。(groonga 2.0.8の場合) <pre> - select Users --filter 'birthday > "1988-01-04 00:00:00"' + select Shops --filter 'ranking >= 98 && ranking < 101' [ - [0,1354069642.52512,0.000299692153930664], + [0,1355461536.3541,3.55056285858154], [ [ - [1], - [ - ["_id","UInt32"], - ["_key","ShortText"], - ["birthday","Time"] - ], - [1,"Alice",697561200.0] + [3], + [["_id","UInt32"],["_key","ShortText"],["ranking","UInt32"]], + [98,"Shop98",98], + [99,"Shop99",99], + [100,"Carlos's shop",100] + ] + ] + ] +</pre> + +丁度3件検索結果が取得できていることがわかります。 + +同じようにしてgroonga 2.0.9でも検索してみます。検索結果は以下のようになります。 + +<pre> + select Shops --filter 'ranking >= 98 && ranking < 101' + [ + [0,1355461395.8912,2.43025946617126], + [ + [ + [3], + [["_id","UInt32"],["_key","ShortText"],["ranking","UInt32"]], + [98,"Shop98",98], + [99,"Shop99",99], + [100,"Carlos's shop",100] ] ] ] </pre> -Bobより後の誕生日は、Aliceだけなので正しく検索できていることがわかります。範囲検索では、不等号( @>,<,<=,>=@ )が使えます。 +検索結果は当然同じですが、実行時間が違います。実行時間は以下の部分が該当します。 + +<pre> + [0,1355461536.3541,3.55056285858154], +</pre> + +groonga 2.0.8の場合は3.55056285858154秒でした。 + +<pre> + [0,1355461395.8912,2.43025946617126], +</pre> + +一方、groonga 2.0.9の場合は2.43025946617126秒でした。 + +groongaが出力する結果のフォーマット詳細については "出力形式":http://groonga.org/ja/docs/reference/command/output_format.html を参照してください。 + +|groongaのバージョン|groonga 2.0.8|groonga 2.0.9| +|実行時間(秒)|3.55056285858154|2.43025946617126| + +2.0.9にするだけで実行時間が7割程度に短縮できていることがわかります。 + +測定は以下の環境で実施しました。 + +|CPU|Intel(R) Core i7-2640M CPU****@2*****| +|メモリ|8GB| h3. geo_distance()関数で矩形近似により境界をまたぐ二点間の距離計算をサポート -------------- next part -------------- HTML����������������������������...Télécharger
