MySQL为何不建议使用null列

 

Preface

 

    Null is a special constraint of columns.The columns in table will be added null constrain if you do not define the column with "not null" key words explicitly when creating the table.Many programmers like to define columns by default because of the conveniences(reducing the judgement code of nullibility) what consequently cause some uncertainty of query and poor performance of database.

 

Introduce

    Null is null means it is not anything at all,we cannot think of null is equal to '' and they are totally different.MySQL provides three operators to handle null value:"IS NULL","IS NOT NULL","<=>" and a function ifnull().

    IS NULL: It returns true,if the column value is null.

    IS NOT NULL: It returns true,if the columns value is not null.

    <=>: It's a compare operator similar with "=" but not the same.It returns true even for the two null values.(eg. null <=> null is legal)

    IFNULL(): Specify two input parameters,if the first is null value then returns the second one.It's similar with Oracle's NVL() function.

 

Example

 

Null never returns true when comparing with any other values except null with "<=>".

 (root@localhost mysql3306.sock)[zlm]>create table test_null(
     -> id int not null,
     -> name varchar()
     -> );
 Query OK,  rows affected (0.02 sec)

 (root@localhost mysql3306.sock)[zlm]>insert into test_null values(,'zlm');
 Query OK,  row affected (0.00 sec)

 (root@localhost mysql3306.sock)[zlm]>insert into test_null values(,null);
 Query OK,  row affected (0.00 sec)

 (root@localhost mysql3306.sock)[zlm]>select * from test_null;
 +----+------+
 | id | name |
 +----+------+
 |   | zlm  |
 |   | NULL |
 +----+------+
  rows in set (0.00 sec)

 (root@localhost mysql3306.sock)[zlm]>select * from test_null where name=null;
 Empty set (0.00 sec)

 (root@localhost mysql3306.sock)[zlm]>select * from test_null where name is null;
 +----+------+
 | id | name |
 +----+------+
 |   | NULL |
 +----+------+
  row in set (0.00 sec)

 (root@localhost mysql3306.sock)[zlm]>select * from test_null where name is not null;
 +----+------+
 | id | name |
 +----+------+
 |   | zlm  |
 +----+------+
  row in set (0.00 sec)

 (root@localhost mysql3306.sock)[zlm]>select * from test_null where null=null;
 Empty set (0.00 sec)

 (root@localhost mysql3306.sock)[zlm]>select * from test_null where null<>null;
 Empty set (0.00 sec)

 (root@localhost mysql3306.sock)[zlm]>select * from test_null where null<=>null;
 +----+------+
 | id | name |
 +----+------+
 |   | zlm  |
 |   | NULL |
 +----+------+
  rows in set (0.00 sec)

 //null<=>null always return true,it's equal to "where 1=1".

Null means "a missing and unknown value".Let's see details below.

 (root@localhost mysql3306.sock)[zlm]>SELECT  IS NULL,  IS NOT NULL, '' IS NULL, '' IS NOT NULL;
 +-----------+---------------+------------+----------------+
 |  IS NULL |  IS NOT NULL | '' IS NULL | '' IS NOT NULL |
 +-----------+---------------+------------+----------------+
 |          |              |           |               |
 +-----------+---------------+------------+----------------+
  row in set (0.00 sec)

 //It's not equal to zero number or vacant string.
 //In MySQL,0 means fasle,1 means true.

 (root@localhost mysql3306.sock)[zlm]>SELECT  = NULL,  <> NULL,  < NULL,  > NULL;
 +----------+-----------+----------+----------+
 |  = NULL |  <> NULL |  < NULL |  > NULL |
 +----------+-----------+----------+----------+
 |     NULL |      NULL |     NULL |     NULL |
 +----------+-----------+----------+----------+
  row in set (0.00 sec)

 //It cannot be compared with number.
 //In MySQL,null means false,too.

It truns null as a result if any expression contains null value.

 (root@localhost mysql3306.sock)[zlm]>select ifnull(null,'First is null'),ifnull(null+,'First is null'),ifnull(concat('abc',null),'First is null');
 +------------------------------+---------------------------------+--------------------------------------------+
 | ifnull(null,'First is null') | ifnull(null+,'First is null') | ifnull(concat('abc',null),'First is null') |
 +------------------------------+---------------------------------+--------------------------------------------+
 | First is null                | First is null                   | First is null                              |
 +------------------------------+---------------------------------+--------------------------------------------+
  row in set (0.00 sec)

 //null value needs to be disposed with ifnull() function,what usually causes sql statement more complex.
 //As we all know,MySQL does not support funcion index.Therefore,indexes on the column may not be used.That's really worse.

It's diffrent when using count(*) & count(null column).

 (root@localhost mysql3306.sock)[zlm]>select count(*),count(name) from test_null;
 +----------+-------------+
 | count(*) | count(name) |
 +----------+-------------+
 |         |            |
 +----------+-------------+
  row in set (0.00 sec)

 //count(*) returns all rows ignore the null while count(name) returns the non-null rows in column "name".
 //This will also leads to uncertainty if someone is unaware of the details above.

When using distinct,group by,order by,all null values are considered as the same value.

 (root@localhost mysql3306.sock)[zlm]>insert into test_null values(,null);
 Query OK,  row affected (0.00 sec)

 (root@localhost mysql3306.sock)[zlm]>select distinct name from test_null;
 +------+
 | name |
 +------+
 | zlm  |
 | NULL |
 +------+
  rows in set (0.00 sec)

 //Two rows of null value returned one and the result became two.

 (root@localhost mysql3306.sock)[zlm]>select name from test_null group by name;
 +------+
 | name |
 +------+
 | NULL |
 | zlm  |
 +------+
  rows in set (0.00 sec)

 //Two rows of null value were put into the same group.
 //By default,group by will also sort the result(null row showed first).

 (root@localhost mysql3306.sock)[zlm]>select id,name from test_null order by name;
 +----+------+
 | id | name |
 +----+------+
 |   | NULL |
 |   | NULL |
 |   | zlm  |
 +----+------+
  rows in set (0.00 sec)

 //Three rows were sorted(two null rows showed first).

MySQL supports to use index on column which contains null value(what's different from oracle).

 (root@localhost mysql3306.sock)[sysbench]>show tables;
 +--------------------+
 | Tables_in_sysbench |
 +--------------------+
 | sbtest1            |
 | sbtest10           |
 | sbtest2            |
 | sbtest3            |
 | sbtest4            |
 | sbtest5            |
 | sbtest6            |
 | sbtest7            |
 | sbtest8            |
 | sbtest9            |
 +--------------------+
  rows in set (0.00 sec)

 (root@localhost mysql3306.sock)[sysbench]>show create table sbtest1\G
 *************************** . row ***************************
        Table: sbtest1
 Create Table: CREATE TABLE `sbtest1` (
   `id` int() NOT NULL AUTO_INCREMENT,
   `k` int() NOT NULL DEFAULT '',
   `c` char() NOT NULL DEFAULT '',
   `pad` char() NOT NULL DEFAULT '',
   PRIMARY KEY (`id`),
   KEY `k_1` (`k`)
 ) ENGINE=InnoDB AUTO_INCREMENT= DEFAULT CHARSET=utf8
  row in set (0.00 sec)

 (root@localhost mysql3306.sock)[sysbench]>alter table sbtest1 modify k int null,modify c char() null,modify pad char() null;
 Query OK,  rows affected (4.14 sec)
 Records:   Duplicates:   Warnings: 

 (root@localhost mysql3306.sock)[sysbench]>insert into sbtest1 values(,null,null,null);
 Query OK,  row affected (0.00 sec)

 (root@localhost mysql3306.sock)[sysbench]>explain select id,k from sbtest1 where id=;
 +----+-------------+---------+------------+-------+---------------+---------+---------+-------+------+----------+-------+
 | id | select_type | table   | partitions | type  | possible_keys | key     | key_len | ref   | rows | filtered | Extra |
 +----+-------------+---------+------------+-------+---------------+---------+---------+-------+------+----------+-------+
 |   | SIMPLE      | sbtest1 | NULL       | const | PRIMARY       | PRIMARY |        | const |     |   100.00 | NULL  |
 +----+-------------+---------+------------+-------+---------------+---------+---------+-------+------+----------+-------+
  row in set,  warning (0.00 sec)

 (root@localhost mysql3306.sock)[sysbench]>explain select id,k from sbtest1 where k is null;
 +----+-------------+---------+------------+------+---------------+------+---------+-------+------+----------+--------------------------+
 | id | select_type | table   | partitions | type | possible_keys | key  | key_len | ref   | rows | filtered | Extra                    |
 +----+-------------+---------+------------+------+---------------+------+---------+-------+------+----------+--------------------------+
 |   | SIMPLE      | sbtest1 | NULL       | ref  | k_1           | k_1  |        | const |     |   100.00 | Using where; Using index |
 +----+-------------+---------+------------+------+---------------+------+---------+-------+------+----------+--------------------------+
  row in set,  warning (0.00 sec)

 //In the first query,the newly added row is retrieved by primary key.
 //In the second query,the newly added row is retrieved by secondary key "k_1"
 //It has been proved that indexes can be used on the columns which contain null value.
 //column "k" is int datatype which occupies 4 bytes,but the value of "key_len" turn out to be 5.what's happed?Because null value needs 1 byte to store the null flag in the rows.

Summary

• null value always leads to many uncertainties when disposing sql statement.It may cause bad performance accidentally.
  
• null value will not be estimated in aggregate function(eg. count(),max(),min()) which may cause inaccurate results.
  
• null value will influence the behavior of the operations such as "distinct","group by","order by" which causes wrong sort.
  
• null value needs ifnull() function to do judgement which makes the program code more complex.
  
• null value needs a extra 1 byte to store the null information in the rows.
  
• As these above drawbacks,it's not recommended to define columns with default null.
  
• We recommand to define "not null" on all columns and use zero number & vacant string to substitute relevant data type of null.