MySql 在单个查询中多次选择

Question

I have a restful web service that has an endpoint for getting all sub-resources of a single resource (say, for a contrived example, selecting all pets for a certain pet store by making a GET request to http://www.url.com/stores/1234/pets ).

This web service is highly concurrent, and gets hit with many requests at once. I'd like to write a single query (for a MySQL database) that allows me to retrieve the number of pets associated with a store, and to be able to distinguish if a store with that id doesn't exist.

That is, if a store with that id exists, but there are no pets from that store, it should return one thing (an empty list); but if that store id does not exist, then raise an exception. Can someone please help me write this SQL query?

Here is some pseudo-sql

CREATE TABLE STORE(
STORE_ID <- AUTO GENERATED PK
.....
)

CREATE TABLE PET(
PET_ID <- AUTO GENERATED PK
STORE_ID <-FOREIGN KEY POINTING TO STORE TABLE
)

STORE|PET
1    | 1,2
2    | NO PETS

I would want the query to be able to give different outputs for querying for store id 2 (where the store exists, but no pets associated with it) and querying for pets for store 3 (where no store with id 3 exists).

Please avoid multiple queries (even using locks/transactions) or stored procedures if possible.

Additional notes:
For the web service, at first I had made a query to check that a store exists with that id, if it doesn't, then throw an exception (that turns into a 404). However, in a highly concurrent application, we could check that the store exists, and find that it does. However, there could be another request that deletes that store before our next query (to retrieve all pets) occurs. In this case, I would like to be true to RESTful principles and return a 404, rather than an empty list. Hence, I would need to distinguish between the two cases. I know how to do this with a transaction and a lock (select for update), but I'd prefer to do this another way if possible.

Answer 1

select s.store_id, count(p.store_id) 
from store as s 
left join pet as p -- get a row if the store has no row in PETS
  on s.store_id = p.store_id
where s.store_id = 9
group by s.store_id -- get no row even if store doesn't exist.

The LEFT join assures that the store is returned even if it has no rows in PETS, the count(p.store_id) returns 0 (caution: don't use count(s.store_id) , this would result in 1).

And the GROUP BY utilizes the fact that a COUNT with Group By returns no result if there's no input row (caution: without Group By the Count will return a row).