I’m curious to know how NULLs are stored into a database ?
It surely depends on the database server but I would like to have an general idea about it.
First try:
Suppose that the server put a undefined value (could be anything) into the field for a NULL value.
Could you be very lucky and retrieve the NULL value with
...WHERE field = 'the undefined value (remember, could be anything...)'
Second try:
Does the server have a flag or any meta-data somewhere to indicate this field is NULL ?
Then the server must read this meta data to verify the field.
If the meta-data indicates a NULL value and if the query doesn’t have ‘field IS NULL’, then the record is ignored.
It seems too easy…
On PostgreSQL, it uses an optional bitmap with one bit per column (0 is null, 1 is not null). If the bitmap is not present, all columns are not null.
This is completely separate from the storage of the data itself, but is on the same page as the row (so both the row and the bitmap are read together).
References: