Should this SELECT use a join or a subquery?

Question

SELECT u.users_username, u.givenname, u.familyname, u.studentassent, u.parentconsent, u.birthdate, u.gender
FROM users AS u
JOIN classes_users as c
ON c.users_username = u.users_username
JOIN classes_users as x
ON c.classes_id = x.classes_id
WHERE x.users_username = "johnny" AND x.role = "teacher"

Or

SELECT u.users_username, u.givenname, u.familyname, u.studentassent, u.parentconsent, u.birthdate, u.gender
FROM users AS u
WHERE u.users_username
IN (
    SELECT c.users_username
    FROM classes_users as c
    JOIN classes_users as x
    ON c.classes_id = x.classes_id
    WHERE x.users_username = "johnny" AND x.role = "teacher"
)

I'm thinking the first one is better, but I'm still learning how to write better SQL statements am not clear on all the internals of what happens that makes one statement better than the other in this case.

If there is a better way to write them than the two ways I've written, please let me know. Thanks.

EDIT: There are teachers and students. Their position as student or teacher in any given class is found by looking at the classes_users table. What I want to do is when given a user, find the classes in which he is a teacher, then return all students in those classes.

Here is my DB schema:

-- -----------------------------------------------------
-- Table `kcptech`.`users`
-- -----------------------------------------------------
CREATE  TABLE IF NOT EXISTS `kcptech`.`users` (
`users_username` VARCHAR(63) NOT NULL ,
`password` VARCHAR(255) NULL DEFAULT NULL ,
`salt` VARCHAR(127) NULL DEFAULT NULL ,
`givenname` VARCHAR(96) NULL DEFAULT NULL ,
`familyname` VARCHAR(128) NULL DEFAULT NULL ,
`privileges` TINYINT NULL DEFAULT NULL ,
`studentassent` TINYINT(1) UNSIGNED NULL DEFAULT NULL ,
`parentconsent` TINYINT(1) UNSIGNED NULL DEFAULT NULL ,
`birthdate` DATE NULL DEFAULT NULL ,
`gender` VARCHAR(1) NULL DEFAULT NULL ,
`registration` TIMESTAMP NULL DEFAULT NULL ,
PRIMARY KEY (`users_username`) ,
UNIQUE INDEX `uname_UNIQUE` (`users_username` ASC) )
ENGINE = InnoDB;




-- -----------------------------------------------------
-- Table `kcptech`.`classes`
-- -----------------------------------------------------
CREATE  TABLE IF NOT EXISTS `kcptech`.`classes` (
`classes_id` INT UNSIGNED NOT NULL AUTO_INCREMENT ,
`course` VARCHAR(127) NULL ,
`period` VARCHAR(31) NULL DEFAULT '' ,
`organization` VARCHAR(127) NULL DEFAULT '' ,
PRIMARY KEY (`classes_id`) ,
UNIQUE INDEX `id_UNIQUE` (`classes_id` ASC) )
ENGINE = InnoDB;


-- -----------------------------------------------------
-- Table `kcptech`.`classes_users`
-- -----------------------------------------------------
CREATE  TABLE IF NOT EXISTS `kcptech`.`classes_users` (
`classes_id` INT UNSIGNED NOT NULL ,
`users_username` VARCHAR(64) NOT NULL ,
`role` VARCHAR(12) NOT NULL ,
PRIMARY KEY (`classes_id`, `users_username`) ,
INDEX `fk_class_users__users_users_username` (`users_username` ASC) ,
INDEX `fk_class_users__class_class_id` (`classes_id` ASC) ,
CONSTRAINT `fk_class_users__users_users_username`
  FOREIGN KEY (`users_username` )
  REFERENCES `kcptech`.`users` (`users_username` )
  ON DELETE NO ACTION
  ON UPDATE NO ACTION,
CONSTRAINT `fk_class_users__class_class_id`
  FOREIGN KEY (`classes_id` )
  REFERENCES `kcptech`.`classes` (`classes_id` )
  ON DELETE NO ACTION
  ON UPDATE NO ACTION)
ENGINE = InnoDB;

Answer 1

The first one is better, assuming that (classes_id, users_username) is unique.

MySQL cannot do semi-joins (IN constructs) with the inner query leading. So the IN query will always use users as a leading table, while for the JOIN query, the optimizer can choose the leading table.

If (classes_id, users_username) is not unique, the queries are semantically not equivalent. You would need to add DISTINCT to the join query.