在导入数据之前在 MySQL 数据库列上使用 php function

Question

I have been working with mySQL and PHP for a few months now and am building a search function for my website. This search function has a filter and a search bar. The filter is currently implemented by using the "WHERE" function of mySQL, by which I exclude any rows that don't fit within the filter requirements to avoid a large data transfer from mySQL. I would like to do this for the search bar as well, but want to use PHP's levenshtein function between the title of the page and search term.

Would anyone have any idea A. whether this is even possible and B. how this would be accomplished? Also maybe C.: should I even worry about the size of the database I'm transferring from mySQL? At what point does it cause serious latency for the user?

An example of how I've done the filter implementation, which is automatically generated in a PHP function that is called by using AJAX.

SELECT title, coverphoto, highteavegatag, pagename, totaltimemin, totaltimemax, preptimemin, preptimemax, date FROM recipes WHERE ((LOCATE('french',cuisine) > 0) AND (LOCATE('dutch',cuisine) > 0)) AND ((dietglnpv & 2 = 2) AND (dietGLNPV & 4 = 4)) AND ((LOCATE('sweet',pagecategory) > 0) AND (LOCATE('diner',pagecategory) > 0)) AND ((totaltimemin <= 331) AND (totaltimemax >= 85)) AND ((preptimemin <= 45) AND (preptimemax >= 28))

And an idea of what I would like to add (I know this won't work, but maybe there's a way).

SELECT (...) FROM recipes WHERE (...) AND (levenshtein(searchstr,title) < 10)

Answer 1

I think this quickly will become very inefficient. Remember, you probably don't want to do a levenshtein() between the whole title and the search string, but between the words in the title and the words in the search string. The PHP manual isn't clear on this but levenshtein() computes the minimum number of single-character edits, not just edits of chunks of characters.

Most general search engines use a more sophisticated approach. You could use the full-text searching functionality in MySQL, or you could build your own system. This usually means finding all the words that occur in the columns to be searched and indexing them. This indexes is done before any searches occur, that way each individual search doesn't have to look through all the texts. Then you match the search string with these words and from them you find the matching rows.