Mysql 在LONGTEXT欄位上作like操作的消耗;

: trac_apsara 17:19:04> select count(*) from wiki;
+----------+
| count(*) |
+----------+
| 76514 |
+----------+
1 row in set (0.03 sec)

: trac_apsara 17:19:08> select count(distinct name ) from wiki;
+-----------------------+
| count(distinct name ) |
+-----------------------+
| 40369 |
+-----------------------+
1 row in set (0.59 sec)

: trac_apsara 17:19:21> show variables like 'innodb_buffer%';
+-------------------------+------------+
| Variable_name | Value |
+-------------------------+------------+
| innodb_buffer_pool_size | 1073741824 |
+-------------------------+------------+
1 row in set (0.00 sec)

: trac_apsara 17:21:08> show table status like 'wiki' G
*************************** 1. row ***************************
Name: wiki
Engine: InnoDB
Version: 10
Row_format: Compact
Rows: 336009
Avg_row_length: 4458
Data_length: 1498120192
Max_data_length: 0
Index_length: 10551296
Data_free: 7340032
Auto_increment: NULL
Create_time: 2010-09-29 14:49:20
Update_time: NULL
Check_time: NULL
Collation: utf8_bin
Checksum: NULL
Create_options:
Comment:
1 row in set (0.01 sec)
===================================================================

#下面我們來看一下SQL和資料：

## SQL1 :
SELECT w1.name,w1.time,w1.author,w1.text
FROM wiki w1,
(SELECT name,max(version) AS ver FROM wiki GROUP BY name) w2
WHERE w1.version = w2.ver AND w1.name = w2.name
AND (w1.name LIKE '%RpcRequestPtr%' ESCAPE '/'
OR w1.author LIKE '%RpcRequestPtr%' ESCAPE '/'
OR w1.text LIKE '%RpcRequestPtr%' ESCAPE '/'
);

## SQL2 :
SELECT w1.name,w1.time,w1.author,w1.text
FROM wiki w1,
(SELECT name,max(version) AS ver FROM wiki GROUP BY name) w2
WHERE w1.version = w2.ver AND w1.name = w2.name
AND (w1.name LIKE '%RpcRequestPtr%' ESCAPE '/'
OR w1.author LIKE '%RpcRequestPtr%' ESCAPE '/'
###### OR w1.text LIKE '%RpcRequestPtr%' ESCAPE '/'
);

兩個SQL僅一個WHERE條件之差。

: trac_apsara 17:24:08> explain SELECT w1.name,w1.time,w1.author,w1.text
-> FROM wiki w1,
-> (SELECT name,max(version) AS ver FROM wiki GROUP BY name) w2
-> WHERE w1.version = w2.ver AND w1.name = w2.name
-> AND (w1.name LIKE '%RpcRequestPtr%' ESCAPE '/'
-> OR w1.author LIKE '%RpcRequestPtr%' ESCAPE '/'
-> OR w1.text LIKE '%RpcRequestPtr%' ESCAPE '/'
-> );
+----+-------------+------------+------+---------------+--------------+---------+----------------+--------+---------------------------------+
| id | select_type | table | type | possible_keys | key | key_len | ref | rows | Extra |
+----+-------------+------------+------+---------------+--------------+---------+----------------+--------+---------------------------------+
| 1 | PRIMARY | | ALL | NULL | NULL | NULL | NULL | 40369 | |
| 1 | PRIMARY | w1 | ref | name_ver_ind | name_ver_ind | 608 | w2.name,w2.ver | 3 | Using where |
| 2 | DERIVED | wiki | ALL | NULL | NULL | NULL | NULL | 445724 | Using temporary; Using filesort |
+----+-------------+------------+------+---------------+--------------+---------+----------------+--------+---------------------------------+
3 rows in set (1.04 sec)

: trac_apsara 17:22:26> explain SELECT w1.name,w1.time,w1.author,w1.text
-> FROM wiki w1,
-> (SELECT name,max(version) AS ver FROM wiki GROUP BY name) w2
-> WHERE w1.version = w2.ver AND w1.name = w2.name
-> AND (w1.name LIKE '%RpcRequestPtr%' ESCAPE '/'
-> OR w1.author LIKE '%RpcRequestPtr%' ESCAPE '/'
-> # OR w1.text LIKE '%RpcRequestPtr%' ESCAPE '/'
-> );
+----+-------------+------------+------+---------------+--------------+---------+----------------+--------+---------------------------------+
| id | select_type | table | type | possible_keys | key | key_len | ref | rows | Extra |
+----+-------------+------------+------+---------------+--------------+---------+----------------+--------+---------------------------------+
| 1 | PRIMARY | | ALL | NULL | NULL | NULL | NULL | 40369 | |
| 1 | PRIMARY | w1 | ref | name_ver_ind | name_ver_ind | 608 | w2.name,w2.ver | 3 | Using where |
| 2 | DERIVED | wiki | ALL | NULL | NULL | NULL | NULL | 445724 | Using temporary; Using filesort |
+----+-------------+------------+------+---------------+--------------+---------+----------------+--------+---------------------------------+
3 rows in set (1.03 sec)

### 從執行計劃來看，兩個SQL一模一樣； 處理的行數也是一樣的；

: trac_apsara 17:25:39> reset query cache ;
Query OK, 0 rows affected (0.00 sec)

: trac_apsara 17:25:52> SELECT w1.name,w1.time,w1.author,w1.text
-> FROM wiki w1,
-> (SELECT name,max(version) AS ver FROM wiki GROUP BY name) w2
-> WHERE w1.version = w2.ver AND w1.name = w2.name
-> AND (w1.name LIKE '%RpcRequestPtr%' ESCAPE '/'
-> OR w1.author LIKE '%RpcRequestPtr%' ESCAPE '/'
-> # OR w1.text LIKE '%RpcRequestPtr%' ESCAPE '/'
-> );

Empty set (1.31 sec)

: trac_apsara 17:26:12> reset query cache ;
Query OK, 0 rows affected (0.00 sec)

: trac_apsara 17:26:15> SELECT w1.name,w1.time,w1.author,w1.text
-> FROM wiki w1,
-> (SELECT name,max(version) AS ver FROM wiki GROUP BY name) w2
-> WHERE w1.version = w2.ver AND w1.name = w2.name
-> AND (w1.name LIKE '%RpcRequestPtr%' ESCAPE '/'
-> OR w1.author LIKE '%RpcRequestPtr%' ESCAPE '/'
-> OR w1.text LIKE '%RpcRequestPtr%' ESCAPE '/'
-> );
13 rows in set (3.50 sec)

## 從執行時間來看,
## SQL1 : 3.50 sec , SQL2: 1.31 sec

## 從這裡我們基本可以判斷出來，MYSQL 用了 2.19 sec 在記憶體中處理40369次TEXT欄位的LIKE模糊查詢操作；
## 而從WIKI表INDEX查詢40369次，卻只用了1.31秒（可能更少），當然資料已經在CACHE裡。

## 我們以後做SQL可要注意了。不光是讀硬碟會消耗時間，在記憶體中的LIKE模糊查詢操作，也很費時間；