Very often, I want to run a query on one of my users where I want a row stored and associated with that user, in a 1-to-1 relationship. So let's say (this is just an arbitrary example), that I have a table that keeps track of a user's car, along with some info about the car. Each user can have either 0 or 1 cars. If the user has no car, there is no entry in the table for that user.
cars table (again, just an example):
id, user_id, car_make, car_model
So, when I update this table, I always end up doing something like this (pseudo-code):
result = SELECT * FROM cars WHERE user_id=5
if (num_rows(result)>0){
UPDATE cars SET car_make='toyota', car_model='prius' WHERE user_id=5
}else{
INSERT INTO cars (user_id, car_make, car_model) VALUES (5, 'toyota', 'prius')
}
How can I make this into one elegant statement that works "atomically"? What happens if, in another process, the row is REMOVED between the SELECT and UPDATE statements? My UPDATE statement will fail where the INSERT statement should have run. And I feel like I need to do two similar (but different) statements to accomplish the same thing! What I need is some statement that will assure me that the data I want exists in the table, especially when I only want 1 row that satisfies my requirement. For example, it might be something like (this is totally made-up of course):
MAKE SURE A ROW IN cars WHERE user_id=5 IS SET WITH car_make='toyota', car_model='prius'
That way, if user_id of 5 exists already, it will be updated, otherwise, it will be inserted. Also, if I changed the requirements, for example to say that each user can have zero or one cars of a given car_make, then I could further specify that:
MAKE SURE A ROW IN cars WHERE user_id=5 AND car_make='toyota' IS SET WITH car_model='prius'
I hope my question makes sense! How can I improve this basic insert-if-not-found or update-if-found operation that comes up so often? Thanks for any help!
See Question&Answers more detail:
os