Eliminate rows with names that are slightly different - sql

Question

I have in POSTGRESQL a database with a UUID, firstname (fname) and phone
uuid fname phone
1 JOHN 111
2 john 111
3 John 111
4 JOHN JAMES 111
5 Charles 222
6 Peter 222
7 James 222
8 Jimmy 222
9 Fred 333
10 Fred 333
11 Greg 333
I would like to keep only the group phone + firstname that have a similarity between at least two names. So, for example, in this case I would like to keep the phone 111 and one of the names and the phone 333 keeping the name that repeats (Fred). The phone 222 woud be eliminate as all names are not similar.
The result data would be
fname phone
John 111
Fred 333
The problem I am having is when the name is similar but it has more names (as in John and John James or when the name was mistyped, as in John and Jonh). I have tried to do the following
SELECT
m1.phone,
m1.fname,
m1.uuid
FROM
master as m1
JOIN master as m2 on m1.uuid = m2.uuid
WHERE
m1.phone = m2.phone
and m1.fname ILIKE m2.fname
ORDER BY 1

Answer 1

The definition of similarity is a bit vague, but this works for the data you have in the question:
select m.*
from master m
where exists (select 1
from master m2
where m2.phone = m.phone and m2.uuid <> m.uuid and
(m.fname ilike '%' || m2.fname || '%' or
m2.fname ilike '%' || m.fname || '%'
)
);
Here is a rextester.
Name matching is a complicated task and not well suited to SQL. However, you might want to look into Levenshtein distance and other string similarity metrics if this is a problem that you are facing.
Note: This keeps all names that match. If you want only one row per phone, you can use distinct on.