Why is ' 2' > '10'? | 易学教程

问题

Why is ' 2' with an initial space bigger than '10'?

select ' 2' > '10';
 ?column? 
----------
 t
(1 row)

I tried it with both latin1 and utf8 english collations:

                                    List of databases
   Name    |   Owner    | Encoding  |    Collation     |      Ctype       |   Access privileges   
-----------+------------+-----------+------------------+------------------+-----------------------
 cpn       | cpn        | UTF8      | en_US.UTF-8      | en_US.UTF-8      | 
 teste     | cpn        | LATIN1    | en_US.ISO-8859-1 | en_US.ISO-8859-1 |

I know it has to do with the type because when it is cast it works as expected:

teste=> select ' 2'::char > '10';
 ?column? 
----------
 f
(1 row)

What exactly is happening here?

EDIT:

All the above was done with 8.4.8 in Fedora 13. But I just tested with 9.04 in Centos 6 with the same result:

select ' 2' > '10';
 ?column? 
----------
 t
(1 row)

List of databases

   Name    |   Owner    | Encoding  |  Collation  |    Ctype    |   Access privileges   
-----------+------------+-----------+-------------+-------------+-----------------------
 cpn       | postgres   | UTF8      | en_US.UTF-8 | en_US.UTF-8 |

New Edit:

This is to further confuse:

select ' ' > '1';
 ?column? 
----------
 f
(1 row)

回答1:

I think PostgreSQL automatically tries to figure out the type behind the scenes and in Linux it tries to get rid of the ' ', some of the comparisons are also based on locale.

Thus, ' 2' > '10' becomes '2'>'10' and the comparison is '2'>'1'; they are not equal, so no need to continue with the rest of the string, and ascii('2') is greater than ascii('1'), so it evaluates to true.
If it were an equality operation (e.g. ' 22' = '22 ') it would result to false because Postgres does a byte by byte comparison. This is important because the engine uses two different algorithms when doing comparisons.
If you specify the type via typecasting, then it won't override the space rules (' '=>'').

Also credit goes to: RhodiumToad and Peerce in #postgresql

回答2:

I think this has to do with locale settings.

According to PostgreSQL docs: Locale Support:

The locale settings influence the following SQL features:

Sort order in queries using ORDER BY on textual data

The ability to use indexes with LIKE clauses

The upper, lower, and initcap functions

The to_char family of functions

来源：https://stackoverflow.com/questions/7053817/why-is-2-10

标签

sql

postgresql

collation