Question

I need to make a synonym search function, using MySQL.

Here is the tables structure :

NAMES_TABLE           SYNONYMS_TABLE
------------          --------------
ID | NAME              ID_1 | ID_2
---+--------          ------+-------
1  | NAME_A              1  |  2
2  | NAME_B              2  |  1
3  | NAME_C              1  |  3
4  | NAME_D              3  |  1
5  | NAME_E              4  |  5
                         5  |  4  

I want it to work like this : entering a word in a form that is in the sends me all the words corresponding, where their indexes are in the synonyms table, as ID_1 or ID_2.

But I can’t figure out to make the request…

Thanks for your help.

Answer 1

One directional: ID_1 is a name, ID_2 is a synonym:

select
  sn.NAME
from
  NAMES_TABLE n
  inner join SYNONYMS_TABLE s on s.ID_1 = n.ID
  inner join NAMES_TABLE sn on sn.ID2 = s.ID_2
where
  n.NAME = :NAME

Two directional: either ID_1 or ID_2 can contain a name, while the other is its synonym.

select
  sn1.NAME
from
  NAMES_TABLE n1
  inner join SYNONYMS_TABLE s1 on s1.ID_1 = n1.ID
  inner join NAMES_TABLE sn1 on sn1.ID = s1.ID_2
where
  n1.NAME = :NAME
union
select
  sn2.NAME
from
  NAMES_TABLE n2
  inner join SYNONYMS_TABLE s2 on s2.ID_2 = n2.ID
  inner join NAMES_TABLE sn2 on sn2.ID = s2.ID_1
where
  n2.NAME = :NAME

Note ID_2 and ID_1 being reversed in the second part of the second query.

Which one you need, depends on the data. If you want ‘lane’ to be a synonym for ‘road’, but not the other way around, you should use the first method. But in that case you’ll have to remember that synonyms in both directions should be added twice, so you’ll get:

NAMES_TABLE           SYNONYMS_TABLE
------------          --------------
ID | NAME              ID_1 | ID_2
---+--------          ------+-------
1  | Road              1  |  3
2  | Lane              3  |  1
3  | Street            2  |  1

Now Road and Street are each others synonyms, but while Lane is a synonym for Road, Road is not for Lane.
If you don’t want that behavour, you’d better use the second, although you may risk entering double records by swapping ID_1 and ID_2. You should be able to prevent this by checking that ID_2 is always higher than ID_1, never the same or lower.