执行以下命令时:
ALTER TABLE `mytable` ADD UNIQUE (
`column1` ,
`column2`
);
我得到了这个错误信息:
#1071 - Specified key was too long; max key length is 767 bytes
columnn1和column2的信息:
column1 varchar(20) utf8_general_ci
column2 varchar(500) utf8_general_ci
我认为varchar(20)只需要21个字节，而varchar(500)只需要501个字节。所以总字节数是522，小于767。为什么我得到了错误消息?
#1071 - Specified key was too long; max key length is 767 bytes
当前回答
我从varchar变成了nvarchar，对我有用。
其他回答
只是在创建表时将utf8mb4更改为utf8就解决了我的问题。例如:CREATE TABLE…DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_unicode_ci;创建表…DEFAULT CHARSET=utf8 COLLATE=utf8_unicode_ci;
请检查sql_mode是否为like
sql_mode=NO_ENGINE_SUBSTITUTION,STRICT_TRANS_TABLES
如果是，则更改为
sql_mode=NO_ENGINE_SUBSTITUTION
OR
重新启动服务器，修改my.cnf文件(如下所示)
innodb_large_prefix=on
您使用的是什么字符编码?有些字符集(如UTF-16等等)每个字符使用一个以上的字节。
我发现这个查询在检测哪些列的索引违反了最大长度方面很有用:
SELECT
c.TABLE_NAME As TableName,
c.COLUMN_NAME AS ColumnName,
c.DATA_TYPE AS DataType,
c.CHARACTER_MAXIMUM_LENGTH AS ColumnLength,
s.INDEX_NAME AS IndexName
FROM information_schema.COLUMNS AS c
INNER JOIN information_schema.statistics AS s
ON s.table_name = c.TABLE_NAME
AND s.COLUMN_NAME = c.COLUMN_NAME
WHERE c.TABLE_SCHEMA = DATABASE()
AND c.CHARACTER_MAXIMUM_LENGTH > 191
AND c.DATA_TYPE IN ('char', 'varchar', 'text')
我认为varchar(20)只需要21个字节，而varchar(500)只需要 需要501字节。所以总字节数是522，小于767。那么,为什么 我收到错误信息了吗?
UTF8每个字符需要3个字节来存储字符串，因此在您的情况下，20 +500字符= 20*3+500*3 = 1560字节，这超过了允许的767字节。
UTF8的限制是767/3 = 255个字符，对于每个字符使用4个字节的UTF8mb4，它是767/4 = 191个字符。
如果您需要使用比限制更长的列，有两种解决方案:
Use "cheaper" encoding (the one that requires less bytes per character) In my case, I needed to add Unique index on column containing SEO string of article, as I use only [A-z0-9\-] characters for SEO, I used latin1_general_ci which uses only one byte per character and so column can have 767 bytes length. Create hash from your column and use unique index only on that The other option for me was to create another column which would store hash of SEO, this column would have UNIQUE key to ensure SEO values are unique. I would also add KEY index to original SEO column to speed up look up.
