But doesn’t the other_table retrieval occur just once per query :
SELECT some_stuff FROM some_table
WHERE a_field IN
( SELECT other_stuff FROM other_table WHERE logic_goes_here )

While my contrasting concern is that the WHERE speaks_english is issued once per row, isn’t it? as :
SELECT sql_no_cache sum(ci.Population) FROM City AS ci WHERE speaks_english(CountryCode)> 0;

Am I missing something, am I correct?

I’ll actually mimic what the statement was trying to do – find the population of countries that speak a given language, use a cursor to get the rows for the countries that speak the language (the inner query), and then loop over it:

CREATE PROCEDURE speaks_language (IN in_language VARCHAR(30))
BEGIN

DECLARE v_countryCode CHAR(3);
DECLARE v_countrySpeaksEnglish BIGINT DEFAULT 0;
DECLARE v_sumSpeaksEnglish BIGINT DEFAULT 0;
DECLARE done INT DEFAULT 0;

DECLARE countries CURSOR FOR
SELECT DISTINCT co.Code FROM Country AS co
INNER JOIN CountryLanguage AS cl ON cl.CountryCode = co.Code
WHERE cl.Language = in_language;
DECLARE CONTINUE HANDLER FOR NOT FOUND SET done = 1;

OPEN countries;

REPEAT
FETCH countries INTO v_countryCode;
IF NOT done THEN
SELECT sql_no_cache sum(ci.Population) INTO v_countrySpeaksEnglish
FROM City AS ci WHERE CountryCode = v_countryCode;
SET v_sumSpeaksEnglish = v_sumSpeaksEnglish + IFNULL(v_countrySpeaksEnglish, 0);
END IF;
UNTIL done END REPEAT;

CLOSE countries;

SELECT v_sumSpeaksEnglish;

END//

Here’s the results on my system, including the original query:

mysql> CALL speaks_language(‘English’)//
+——————–+
| v_sumSpeaksEnglish |
+——————–+
| 237134840 |
+——————–+
1 row in set (0.04 sec)

Query OK, 0 rows affected (0.04 sec)

mysql> CALL speaks_language(‘English’)//
+——————–+
| v_sumSpeaksEnglish |
+——————–+
| 237134840 |
+——————–+
1 row in set (0.04 sec)

Query OK, 0 rows affected (0.04 sec)

mysql> CALL speaks_language(‘English’)//
+——————–+
| v_sumSpeaksEnglish |
+——————–+
| 237134840 |
+——————–+
1 row in set (0.05 sec)

Query OK, 0 rows affected (0.05 sec)

mysql> SELECT sql_no_cache sum(ci.Population) FROM City AS ci WHERE CountryCode IN ( SELECT DISTINCT co.Code FROM Country AS co INNER JOIN CountryLanguage AS cl ON cl.CountryCode = co.Code WHERE lower(cl.LANGUAGE) = ‘English’);//+——————–+| sum(ci.Population) |+——————–+
| 237134840 |
+——————–+
1 row in set (0.05 sec)

mysql> SELECT sql_no_cache sum(ci.Population) FROM City AS ci WHERE CountryCode IN ( SELECT DISTINCT co.Code FROM Country AS co INNER JOIN CountryLanguage AS cl ON cl.CountryCode = co.Code WHERE lower(cl.LANGUAGE) = ‘English’);//
+——————–+
| sum(ci.Population) |
+——————–+
| 237134840 |
+——————–+
1 row in set (0.06 sec)

mysql> SELECT sql_no_cache sum(ci.Population) FROM City AS ci WHERE CountryCode IN ( SELECT DISTINCT co.Code FROM Country AS co INNER JOIN CountryLanguage AS cl ON cl.CountryCode = co.Code WHERE lower(cl.LANGUAGE) = ‘English’);//
+——————–+
| sum(ci.Population) |
+——————–+
| 237134840 |
+——————–+
1 row in set (0.05 sec)

DELIMITER //
DROP FUNCTION IF EXISTS speaks_english//
CREATE FUNCTION speaks_english(c char(3)) returns integer deterministic
begin
RETURN (SELECT 1 FROM Country AS co INNER JOIN CountryLanguage AS cl ON cl.CountryCode = co.Code WHERE lower(cl.LANGUAGE) = ‘English’ AND co.Code = c LIMIT 1);
end//
DELIMITER ;

On my machine it’s still 10 times slower than the subquery, though.

My test wasn’t apples to apples either, but I found that writing stuff in PHP was about ten times faster. If you add in that MySQL stored procedures don’t have a debugger, it’s a tough sell to write pages and pages of business logic in them.