Re: problem with sorting using 'ORDER BY' when character

On Thu, 20 May 2004, Dragan Matic wrote:

> If I have a table t with column c which is defined as char(5) and fill
> it with following values:
>
> insert into t (c) values (' 1')
> insert into t (c) values (' 2')
> insert into t (c) values (' 3')
> insert into t (c) values (' 4')
> insert into t (c) values (' 11')
> insert into t (c) values (' 12')
> insert into t (c) values (' 14')
> insert into t (c) values (' 24')
> insert into t (c) values (' 21')
> insert into t (c) values (' 31')
> insert into t (c) values (' 333')
>
> and then do the following: SELECT C FROM T ORDER BY C
> Postgres gives me the following
>
> 1
> 11
> 12
> 14
> 2
> 21
> 24
> 3
> 31
> 333
> 4
>
> the same thing done with MS SQL server gives this as a result:
>
> 1
> 2
> 3
> 4
> 11
> 12
> 14
> 21
> 24
> 31
> 333
>
> which is the result I find more logical, meaning the user would expect
> data sorted this way. Is there some way to make Postgres sort elements
> in this way (setting sort order or collation, I suppose)? Tnx in advance

You are probably running in a collation that doesn't treat spaces as
particularly significant (for example with a locale of en_US). If you
want collation by byte order you can use "C" locale (although you need to
re-initdb to change it).

You could also sort it as numbers by converting to a numeric type first
presumably.