This must be a niche scenario since I have not been able to find a similar question around and in my brief testing in my SQL workbench just using the string in place of the column name did not work.
eg:
SELECT MATCH ('fork') AGAINST ('user entered text about forks' IN NATURAL LANGUAGE MODE);
Doesn't work...
I have a query that returns matches on a full text index with the relevance score as one of the columns returned. In this app, I am looking for "search suggestions" in a suggestions table that is built off the websites search index content. The user side also stores everything they search for in their local browser storage.
Currently, I have front end code that uses regex to pull matches from their local storage search history (up to 5) and then sends what they typed (as they type) to the back end to get the best matches from the suggestions table.
The way it works now, is the (up to 5) history matches are shown first, then the rest are filled in up to 10 total matches from the back end. What I would prefer, is that I send the history matches to the back end and include them in the FT match query in some way so that the result set contains all matched suggestions from the table + the history matches sent from the front end, but all sorted by the full text match relevance score to get them all in order of relevance. The new way may result in no history matches showing or it might result in more than 5 history matches showing, it would all boil down the releveance score.
Is something like this possible? The only other way I could image doing this is somehow creating a temporary table with a full text index, on the fly, and then joining that table in my current query, then removing the temp table when its done. The problem with that, in my mind, is that this is all happening in real time as the user types so I don't want to add something like that if its going to bog down the response time. Is there a fast/optimal way of doing this? Is there a way that would also remove the temporary table when the query ends?
Or is there some other command that can just give me a score based on string value against what the user typed in like what I tried above?
EDIT: It looks like my temporary table idea could work: https://dev.mysql.com/doc/refman/8.0/en/create-temporary-table.html
I'll just have to see what kind of perforamce impact this has. Im still interested to hear thoughts on if this is the best / only way or if there is a better one.
The CREATE TEMPORARY TABLE route was the way to go here. I tested it out and its working.
Worthy of note to future travelers. I had to switch my main table from innodb to myisam for this to work. I was able to mix/match the myisam temp table with the innodb main table, but the scoring algorithms are different so the innodb matches were taking priority due to higher scores. This was not an issue for me as I really did not need / use transactions for the primary suggestions table so I just made them both MyISAM engines.
Another item of note, is that I had to switch to splitting the user's query into "words" and ecapsulating them in "*" and running the match as a boolean search instead of natural language becausae in the case of the temp table, a user would likely have entered similar searches which would mean most of the words were in more than 50% of the rows so no matches were returning. Boolean search works around this. Again, not a big deal for my particular use case.
Had I needed to stay in innodb for this, it would have been a problem because from what I can tell, there is no way to set a full text index on an innodb temporary table.