count characters matching in CONCAT MySQL REGEXP - mysql

Question

I have the following MySQL query which works
SELECT *,
CONCAT( office, ' ', contactperson ) AS bigDataField
FROM webcms_mod_references
HAVING bigDataField REGEXP "one|two"
Now there is no ORDER BY and if:
- bigDataField contains "one" this field is shown
- bigDataField contains "one two" this field is shown aswell
now it depends on the id which one of those is shown first, but I want the one with the more matches to be shown first!
I tried with
SUM(
CASE WHEN bigDataField REGEXP "one|two"
THEN 1
ELSE 0 END
) AS matches
But that does not work. Can anyone help me. I think the best would be as the title says to count the matching charachters from the REGEXP. If there are other ways please explain.
The REGEXP is a user input, so, I'm trying to implement a small search over a small Database.

Answer 1

This is theoretical whilst sqlfiddle is down but you might have to split the REGEXP into two so you can count the matches. REGEXP will return either a 0 or 1. Either it matched or didn't. There's no support for finding how many times it was matched in a string.
SELECT *,
CONCAT( office, ' ', contactperson ) AS bigDataField
FROM webcms_mod_references
HAVING bigDataField REGEXP "one|two"
ORDER BY (bigDataField REGEXP "one" + bigDataField REGEXP "two") DESC

Answer 2

There is no way to count the amount of matches on a regex. What you can do is match them separately and order by each of those matches. EG:
SELECT *,
CONCAT( office, ' ', contactperson ) AS bigDataField
FROM webcms_mod_references
HAVING bigDataField REGEXP "one|two"
ORDER BY
CASE WHEN bigDataField REGEXP "one" AND bigDataField REGEXP "two" THEN 0
ELSE 1 -- The else should catch the "two" alone or the "one" alone because of the filtering
END
Of course, you can use a LIKE here too but maybe your regex are more complex than that :)

Answer 3

When I want to count some substring I do replace and "-" the length, example:
SELECT (
LENGTH('longstringlongtextlongfile') -
LENGTH(REPLACE('longstringlongtextlongfile', 'long', ''))
) / LENGTH('long') AS `occurrences`
I think this is an elegant solution for a problem of counting how many times 'long' appears inside provided 'string'

Answer 4

This is not especially the answer to this question, but I think strongly attached to it... (And I hope, will help someone, who cames from google, etc)
So if you use PHP (if not, may dont keep reading ...), you can build the query with that, and in this case, you can do this (about #Moob great answer):
function buildSearchOrderBy(string $regex, string $columName, string $alternateOrderByColumName): string
{
$keywords = explode ('|', $regex);
if (empty ($keywords)) {
return $alternateOrderByColumName;
}
$orderBy = '(';
$i = 0;
foreach ($keywords as $keyword) {
$i++;
if ($i > 1) $orderBy .= " + ";
$orderBy .= "IF((" . $columName . " REGEXP '" . $keyword . "')>0, " . (100 + strlen($keyword)) . ", 0)";
}
$orderBy .= ')';
return $orderBy;
}
So in this case every match worth 100 + so many scores, what the numbers of the characters in the current keyword. Every match starting from 100, because this ensure the base, that the first results will be these, where the total score originate from the more matches, but in proportionally worth more a longer keyword in any case.
Builded to one column check, but I think you can update easy.
If copied to your project, just use like this (just an example):
$orderBy = buildSearchOrderBy($regex, 'article.title', 'article.created');
$statement = "SELECT *
FROM article
WHERE article.title REGEXP '(" . $regex . ")'
ORDER BY " . $orderBy . " DESC"
;