Home/ Questions/Q 3356870
In Process

The Archive Base Latest Questions

Editorial Team
  • 0
Asked: 2026-05-18T02:33:56+00:00

I have a MySQL table in which a column contains string prefixes. For instance

  • 0

I have a MySQL table in which a column contains string prefixes. For instance these prefixes could be top-level directories on an Unix file system:

my_table:    
+---------+
| prefix  |
+---------+
|  /usr/  |
|  /bin/  |
|  /var/  |
|  /lib/  |
+---------+

How can I write a query that efficiently finds all rows in this table where the value of the prefix column is the beginning of a given string?

For instance given the string ‘/usr/bin/cat’ how can I write a query that finds the row containing ‘/usr/’ which is the beginning of ‘/usr/bin/cat’.

My first guess is to use LIKE this way:

SELECT * FROM my_table
WHERE '/usr/bin/cat' LIKE CONCAT(prefix, '%')

But I’m afraid this query won’t be using the index I have on the prefix column.

I also came up with the following:

SELECT * FROM my_table
WHERE prefix <= '/usr/bin/cat' ORDER BY prefix DESC LIMIT 1

Which retrieves the prefix equal to or immediately preceding ‘/usr/bin/cat’ in lexicographical order. I can then verify whether that prefix actually begins with ‘/usr/bin/cat’ or not.

But that only works with a single row and I wonder if that’s the optimal solution.

Edit: I used root directories as an example but I’d like to know if there’s a way to deal with arbitrary strings as well. Perhaps these strings won’t contain path separators or the prefix could be several level deep. Say: ‘/usr/lib’.

Edit: It seems that my second query is bogus. ‘/usr/’ is smaller than ‘/usr/bin/cat’ but so is ‘/usr/a’. That query is still much faster than a full table scan on a large table but to make it work I have to fetch more rows and go through them until I find the first actual prefix.

So it seems an index can help in this kind of prefix search but I still don’t know the best way to take advantage of it.

Share

Leave an answer

You must login to add an answer.

Need An Account,

1 Answer

  1. Editorial Team

    — situation: We do not know where the string can be cut.

    — But we must know maximal length of the prefix.

    — EDIT: It would also help to know the minimal length of prefix – to eliminate lots of false positives that we do not want to find. (min = 2 characters).

    — This will definitely use the index: in this example it is max.8 characters. x = 8
    — in your application, just try to generate such SQL query:
    — No full table scan,just (x – min +1) times uses the index. Hopefully this will be FAST enough! 🙂

    SELECT * FROM my_table WHERE prefix = '/u'

 UNION

SELECT * FROM my_table WHERE prefix = '/us'

 UNION

SELECT * FROM my_table WHERE prefix = '/usr'

 UNION

SELECT * FROM my_table WHERE prefix = '/usr/'

 UNION

SELECT * FROM my_table WHERE prefix = '/usr/b'

 UNION

SELECT * FROM my_table WHERE prefix = '/usr/bi'

 UNION

SELECT * FROM my_table WHERE prefix = '/usr/bin';
    • 0