Question

I’ve seen this on a lot of fields on a DB from a project I’ve been working on, where a column will be defined not null, but will have an empty string as the default. what’s the point of doing this? If you will allow an empty string, why not just allow a field to be null?

Answer 1

NULLs have special behavior: comparing anything with a NULL gives you back a NULL, which is something else than false or 0. It means “unknown”.

For example, take this table:

 user_id | gender
------------------
 1       | NULL
 2       | 'M'
 3       | 'F'
 4       | 'F'

SELECT * FROM mytable WHERE gender = 'M' will return 1 row, as expected

SELECT * FROM mytable WHERE gender != 'M' will return 2 rows, NOT 3 rows.

SELECT * FROM mytable WHERE gender != 'M' OR gender IS NULL will return the expected 3 rows.

---

Edit: For some applications, using 0 (or, God forbid, another “magic number”) instead of NULL is not even advisable (units or exact values are not relevant in this example):

 Date       | Temperature 
--------------------------
 2010-01-01 | 10          
 2010-01-02 | 4
 2010-01-03 | 0
 2010-01-04 | -22
 2010-01-05 | -45
 2010-01-06 | NULL
 2010-01-07 | -34

Here, the NULL on Jan 6th means “value unknown” – maybe because the temperature was so low that the thermometer probe stopped responding. However, it’s a completely different meaning than Jan 3rd, when the temperature was 0, that is, 0 degrees.

Also, as @Bill Karwin mentions, NULLs behave specially in aggregate functions (COUNT,SUM,AVG etc.): calculating AVG(Temperature) on the above data would give you -14.5, as the NULL row is ignored.