在B栏（MySQL）的“文本”中找到准确的单词形式A栏

Question

i have 2 tables, and try to eleminate all entries in table 1 (multiple words per row) wich contain one of the entries in table 2. These words from table 2 can be somewhere in the strings of Table 1.

it should find things like: 'house' in 'big house here' or in 'big house'

it should not find things like this: 'house' in 'houses'

I tried to use the locate function like this:

CREATE TABLE `test`
AS (
  SELECT
    `table1`.`term1`,
    `table2`.`term2`
  FROM `table1`,`table2`
  WHERE
    locate(concat(' ',`table2`.`term2`,' '), concat(' ',`table1`.`term1`,' '))
);

the problem is: it finds some, but not all, and i cannot see the logic behind there why it is not working for everything.

Answer 1

If there is any punctuation surrounding the word you're looking for, your matching won't work.

You could replace all punctuation in the field with spaces .

However, I think a much cleaner solution would be a regular expression :

CREATE TABLE test
AS
SELECT table1.term1, table2.term2
FROM table1, table2
WHERE table1.term1 REGEXP CONCAT('(^|[^A-Za-z]])',table2.term2,'([^A-Za-z]|$)');

(^|[^A-Za-z]) means either start of field or not AZ or az.
([^A-Za-z]|$) means either not AZ or az or end of field.

SQLFiddle .

EDIT:

While the above is pretty and all, it's not particularly efficient. ( 140 ms in a small test)

More efficient: ( 80 ms , could be much better on proper data)

SELECT term1, term2
FROM table1, table2
WHERE term1 LIKE CONCAT('%',term2,'%')
  AND term1 REGEXP CONCAT('(^|[^A-Za-z])',term2,'([^A-Za-z]|$)');

Way more efficient: ( 8 ms ) (for some weird reason, MySQL seemingly can't do regex very well)

SELECT COUNT(*)
FROM table1, table2
WHERE term1 LIKE CONCAT(term2,' %')
   OR term1 LIKE CONCAT(term2,',%')
   OR term1 LIKE CONCAT(term2,'.%')
   OR term1 LIKE CONCAT(term2,';%')
   OR term1 LIKE CONCAT('% ',term2,' %')
   OR term1 LIKE CONCAT('% ',term2,',%')
   OR term1 LIKE CONCAT('% ',term2,'.%')
   OR term1 LIKE CONCAT('% ',term2,';%')
   OR term1 LIKE CONCAT('% ',term2)

Slightly more efficient: ( 4 ms )

SELECT COUNT(*)
FROM table1, table2
WHERE CONCAT(' ', REPLACE(REPLACE(REPLACE(term1, ',', ' '), '.', ' '), ';', ' '), ' ')
        LIKE CONCAT('% ',term2,' %')

You may want to include a few more characters above.

SQLFiddle .

Note that much of the above depends on the data, some may be more efficient in some cases and much worse in others (but regex will probably trail behind).

Even more efficient?

Fulltext indices + searching .