in sequelize (javascript) it's pretty straightforward to either find a record, or create it if it doesn't exist. I don't see anything similar with sea-orm. There's a 'save' method that seems to insert or update, but I need to know details about the record ahead of time :/
Any ideas?
https://sequelize.org/docs/v6/core-concepts/model-querying-finders/
I managed to get this working, but there has to be a better way. How else could I write this?
pub async fn insert_or_return_user( db: &DbConn, partial_user: Auth0UserPart, ) -> Result { let user = users::ActiveModel { email: Set(partial_user.email.to_owned()), email_verified: Set(partial_user.email_verified.to_owned()), auth0_sub: Set(partial_user.sub.to_owned()), ..Default::default() }; let result = user.clone().insert(db).await; match result { Ok(u) => { println!("{u:#?}"); Ok(u.try_into_model().unwrap() as UsersModel) } Err(error) => { let user = Users::find() .filter(users::Column::Auth0Sub.eq(&partial_user.sub)) .one(db) .await?; Ok(user.unwrap() as UsersModel) } } }
I have never used sea-orm, but I wonder if
.on_conflict
could be used to simplify the code above?This is likely what OP will have to do. It actually looks like ANSI SQL now has merge, but you can scroll down a bit and see how each DB handles it slightly differently if you don’t use merge.
I think I had that in a few attempts, I can’t remember why I removed it. Thanks for pointing this out.
I would likely do it in reverse. Try and find the object and if it doesn’t exist create it.
That will always be prune to race conditions, where you check if someting exists (then some other thread creates it) and then you try to create it. You should always try to create first, then if it fails due to it already existing, fetch it. That is a good general rule for anything from hashmaps to databases.
That’s true, but also trying to create first has a race condition too. The above code will panic at the unwrap if the record is deleted after the failed insert, and before the select
So, a loop it is…
I’m new to multithreaded programming. How would some other thread create it? Like what’s the real-world scenario?
That is all dependent on the program, but the simplest scenario is by an API with two requests at the same time. But it may also be like if you scan for new files, and use inotify, then you may also have a scanning loop as a fallback. Then the scan and inotify may trigger at almost the same time, so if that then results in a db create or insert you can get in to this problem. So, there are multiple ways to get in to trouble, and life always find new ways 😀
Indeed it does haha, thanks