Sort alphanumeric column

Question

I have a column in database:

Serial Number
-------------
S1
S10
...
S2
S11
..
S13

I want to sort and return the result as follows for serial number <= 10 :

S1
S2
S10

One way I tried was:

select Serial_number form table where Serial_Number IN ('S1', 'S2',... 'S10');

This solves the purpose but looking for a better way

Answer 1

Here is an easy way for this format:

order by length(Serial_Number),
         Serial_Number

This works because the prefix ( 'S' ) is the same length on all the values.

Answer 2

For Postgres you can use something like this:

select serial_number
from the_table
order by regexp_replace(serial_number, '[^0-9]', '', 'g')::integer;

The regexp_replace will remove all non-numeric characters and the result is treated as a number which is suited for a "proper" sorting.

Edit 1:

You can use the new "number" to limit the result of the query:

select serial_number
from (
  select serial_number, 
         regexp_replace(serial_number, '[^0-9]', '', 'g')::integer as snum
  from the_table
) t 
where snum <= 10
order by snum;

Edit 2

If you receive the error ERROR: invalid input syntax for integer: "" then apparently you have values in the serial_number column which do no follow the format you posted in your question. It means that regexp_replace() remove all characters from the string, so a string like S would cause that.

To prevent that, you need to either exclude those rows from the result using:

where length(regexp_replace(serial_number, '[^0-9]', '', 'g')) > 0

in the inner select. Or, if you need those rows for some reason, deal with that in the select list:

select serial_number
from (
  select serial_number, 
         case
            when length(regexp_replace(serial_number, '[^0-9]', '', 'g')) > 0 then regexp_replace(serial_number, '[^0-9]', '', 'g')::integer as snum
            else null -- or 0 whatever you need
         end as snum
  from the_table
) t 
where snum <= 10
order by snum;

---

This is a really nice example on why you should never mix two different things in a single column. If all your serial numbers have a prefix S you shouldn't store it and put the real number in a real integer (or bigint ) column.

Using something like NOT_SET to indicate a missing value is also a bad choice. The NULL value was precisely invented for that reason: to indicate the absence of data.

Answer 3

Since only the first character spoils your numeric fun, just trim it with right() and sort by the numeric value:

SELECT *
FROM   tbl
WHERE  right(serial_number, -1)::int < 11
ORDER  BY right(serial_number, -1)::int;

Requires Postgres 9.1 or later. In older versions substitute with substring (x, 10000) .