Mysql - Search by tokenized string - mysql

Question

I have a simple request: search by a user name.
Nevertheless, imagine I have this data on database:
1 - John Amazing Doe
2 - John Stupid Doe
3 - John Anthony
When I search by "John Doe", is there an easy way to search by results that have "john" or "doe"? Even better, if we could rank by who has more matches?
Thanks

Answer 1

Prepare the data prior to building the query
This is something that would be best handled a bit programmatically prior to sending to MySQL.
Turn the name into a list, break on space:
"John Doe" --> "John|Doe"
SELECT username FROM users WHERE username RLIKE 'John|Doe'
Be careful about sql injection here.
php Example
$db = new PDO([conn info]);
$searchArg = 'John Doe';
$nameRegex = str_replace(' ','|', strtolower($searchArg));
$userSearch = $db->prepare('SELECT username FROM users WHERE LOWER(username) REGEXP :nameRegex');
$userSearch->bindValue(':nameRegex', $nameRegex, PDO::PARAM::STR);
$userSearch->execute();
foreach ( $userSearch->fetchAll(PDO::FETCH_ASSOC) as $user ) {
echo $user['username'] . '\n<br>';
}

Answer 2

A full text index and MATCH() is your best option. That said, here is another way:
You can search for names that have 'John' or 'Doe' using LIKE:
SELECT user_name
FROM users
WHERE user_name LIKE 'John%'
OR user_name LIKE '%Doe'
Use leading and trailing % only if names will not always start with 'John' or end with 'Doe', otherwise use one. Double % will not make use of any indexes you have on user_name
You can then rank these using UNION and sorting column:
SELECT user_name, 1 as Sort_Col
FROM users
WHERE user_name LIKE 'John%'
AND user_name LIKE '%Doe'
UNION
SELECT user_name, 2 as Sort_Col
FROM users
WHERE (user_name LIKE 'John%'
OR user_name LIKE '%Doe')
AND user_name NOT IN (
SELECT user_name
FROM users
WHERE user_name LIKE 'John%'
AND user_name LIKE '%Doe')
ORDER BY Sort_Col ASC
You'll have to use an annoying NOT IN for the second query to be sure that you don't return rows that have both 'John' and 'Doe'. Normally the UNION would suffice, but since we added Sort_Col it will not. Just another reason to add a full text index if at all possible.