NOT IN vs IN Do Not Return Complimentary Results

Question

Hi I am working through example #7 from the sql zoo tutorial: SELECT within SELECT. In the following question

\"Find each country that belongs to a continent where all p

Answer 1

Show the table DECLARATION. It seems you use CONTINENT as the continent number. Then you should check it is marked with PRIMARY KEY and NOT NULL options. I realyl suspect you just forgot about very special meaning NULL has in SQL.

I make an example in Firebird 2.5.1 SQL server.

CREATE TABLE WORLD (
    CONTINENT   INTEGER,
    NAME        VARCHAR(20),
    POPULATION  INTEGER
);


INSERT INTO WORLD (CONTINENT, NAME, POPULATION) VALUES (NULL, 'null-id', 100);
INSERT INTO WORLD (CONTINENT, NAME, POPULATION) VALUES (1, 'normal 1', 10);
INSERT INTO WORLD (CONTINENT, NAME, POPULATION) VALUES (2, 'normal 2', 200);
INSERT INTO WORLD (CONTINENT, NAME, POPULATION) VALUES (3, 'null-pop', NULL);
INSERT INTO WORLD (CONTINENT, NAME, POPULATION) VALUES (4, 'normal 4', 110);

COMMIT WORK;

Now let's try your requests and see if the 1st row, having CONTINENT IS NULL would be present anywhere:

SELECT continent, population FROM world
WHERE continent IN (
    SELECT continent FROM world
    WHERE population > 100)

CONTINENT   POPULATION
2           200
4           110

and then

SELECT continent, population FROM world
WHERE continent NOT IN (
    SELECT continent FROM world
    WHERE population > 100)

CONTINENT   POPULATION
1           10
3           

By the logic of the request you suppose CONTINENT to be the row ID, then you should make it NOT-NULL and then there would not be the line, that is not seen by [NOT] IN condition.

---

Now, let re-phrase this into flat query:

SELECT continent, population FROM world
    WHERE NOT (population > 100)

CONTINENT   POPULATION
      100
1           10

SELECT continent, population FROM world
    WHERE population > 100

CONTINENT   POPULATION
2           200
4           110

This time the missed row was the one having NULL for Population column.

---

Then FreshPrinceOfSO suggested using EXISTS clause. While potentially it may end with most slow (non-effective) query plan, it at least masks away the special meaning of NULL value in SQL.

SELECT continent, population FROM world w_ext
WHERE EXISTS (
   SELECT continent FROM world w_int
   WHERE (w_int.population > 100) and (w_int.continent = w_ext.continent)
)

CONTINENT   POPULATION
2   200
4   110

SELECT continent, population FROM world w_ext
WHERE NOT EXISTS (
   SELECT continent FROM world w_int
   WHERE (w_int.population > 100) and (w_int.continent = w_ext.continent)
)

CONTINENT   POPULATION
  100
1   10
3