问题
I have a database dump from the geonames website for Great Britain. It consists of approx 60000 records. example data is as follows:
id | name | admin1 | admin2 | admin3 | feature_class | feature_code
-------------------------------------------------------------------------------------------
2652355 | Cornwall | ENG | C6 | | A | ADM2
11609029 | Cornwall | ENG | | | L | RGN
6269131 | England | ENG | | | A | ADM1
The first record with feature code ADM2 means it is administrative level 2 The secord record with feature code RGN means it is a region.
I want to search for records by place names to build an autocomplete feature. Where records have the same name and if one of those records is a region i.e. has feature_code RGN then I want to return only that record otherwise I want to return the record which matches that name that has the lowest id.
I have tried the following but it doesn't work:
SELECT IF(t0.feature_code = 'RGN', MAX(t0.id), MIN(t0.id)) as id
, CONCAT_WS(', ', t0.name,
IF(t3.name != t0.name, t3.name, NULL),
IF(t2.name != t0.name, t2.name, NULL),
IF(t1.name != t0.name, t1.name, NULL)) AS name
FROM locations t0
LEFT JOIN locations t1 ON t1.admin1 = t0.admin1 AND t1.feature_code = 'ADM1'
LEFT JOIN locations t2 ON t2.admin2 = t0.admin2 AND t2.feature_code = 'ADM2'
LEFT JOIN locations t3 ON t3.admin3 = t0.admin3 AND t3.feature_code = 'ADM3'
WHERE
(t0.feature_class IN ('P', 'A') OR (t0.feature_class = 'L' AND t0.feature_code = 'RGN' ) )
AND t0.name like 'Cornwall%'
GROUP BY CONCAT_WS(', ', t0.name,
IF(t3.name != t0.name, t3.name, NULL),
IF(t2.name != t0.name, t2.name, NULL),
IF(t1.name != t0.name, t1.name, NULL))
ORDER BY t0.name
It returns the incorrect record:
id | name
---------------------------
2652355 | Cornwall, England
回答1:
I think that conditional aggregation should do the trick. You can filter records by name, then apply the logic within aggregate functions. If a record exists with feature_code = 'RGN' then you want to select it, else you would pick the minimum id in matching record.
SELECT IFNULL(MAX(CASE WHEN feature_code = 'RGN' THEN id END), MIN(id)) id_found
FROM mytable
WHERE name = @name;
Demo on DB Fiddle when searching for 'Cornwall':
| id_found |
| -------- |
| 11609029 |
NB : if you want the whole matching record, one solution is to simply JOIN the above resultset with the original table:
SELECT t.*
FROM mytable t
INNER JOIN (
SELECT IFNULL(MAX(CASE WHEN feature_code = 'RGN' THEN id END), MIN(id)) id_found
FROM mytable
WHERE name = @name
) x ON x.id_found = t.id;
Demo:
| id | name | admin1 | admin2 | admin3 | feature_class | feature_code |
| -------- | -------- | ------ | ------ | ------ | ------------- | ------------ |
| 11609029 | Cornwall | ENG | | | L | RGN |
回答2:
In MySQL, you can use a correlated subquery:
select l.*
from locations l
where l.id = (select l2.id
from locations l2
where l2.name = l.name
order by (feature_code = 'RGN') desc, -- put regions first
id asc
);
In MySQL 8+, you can also use row_number():
select l.*
from (select l.*,
row_number() over (partition by name
order by (feature_code = 'RGN') desc, id
) as seqnum
from locations l
) l
where seqnum = 1;
回答3:
one approach could be exists and union all
select t1.* from location t1
where exists ( select 1 from location t2 where t2.name=t1.name and t2.feature_code='RGN'
)
and t1.feature_code='RGN'
union all
select t1.* from location t1
where not exists ( select 1 from location t2 where t2.name=t1.name and
t2.feature_code='RGN'
)
and t1.id=(select min(id) from location t2 where t2.name=t1.name)
来源：https://stackoverflow.com/questions/54960930/mysql-conditional-min-max-to-return-distinct-record