Also, Federico’s article goes nicely with it. Thanks for the reminder @Federico.

The sentence I disagree with is “In PostgreSQL, NULL means no value”. I agree that this is PostgreSQL intention (because this is what SQL standard says), but in practice all DBMSs I know treat it too inconsistently.

First, you’re comparing an RDBMS engine (PostgreSQL) to programming languages (C & Java). That doesn’t even make sense. These are not the same thing.

Second, you need to understand it in the historical context. In C, NULL is 0. Why? In Unix systems, the first page of memory (which includes address 0) is typically set as inaccessible to a process. When pointers are initialized to 0, they point to this inaccessible piece of memory. This is one of the best ways to catch an uninitialized pointer, as using it will result in your program crashing immediately. So the uninitialized pointer is a null pointer. We don’t use NULL in other contexts – we just use 0. I wouldn’t put “for (x=NULL; …”. There’s a reason we have this, and it’s not to make typing “0” more difficult.

Third, your database examples are surprisingly good given the rest of this, but you missed one of the qualities of NULL values that is useful. They aren’t used in aggregate functions. For instance, taking the students_mark example, “select count(marks) from students_mark” will give you “2”, and “select average(marks) from students_mark” will give you “4.5”. In a database context, NULL means “missing or unknown”. Unfortunately, those are different concepts and you have to use your programming to make up for it.

Please also see the remarks from Federico Razzoli.

I would also add info about “null” in javascript.

To sum up, my view is that some earlier versions of relational algebra defined two markers (I-marker and A-marker) that defined, respectively, an absent value and an unknown value. SQL standard merged these idea into a single NULL marker in an inconsistent way. The result is that, in practice, sometimes NULLs semantics only make sense if it means absent, sometimes they make sense if it means unknown, and sometimes they don’t make sense in both cases.

In what databases?

NULL meaning is dictated by SQL standard, and Postgres follows it.

It would be interesting to see if some databases treat it differently compared to the standard.