Confused by the upsert.

[teenjuna]

Hey! I'm really confused by upsert. It looks like I should use it when I want to save some record, by not sure if it already exists or not. Usually this happens when I have some struct with optional id field (e.g. id: Option<i32>). If id.is_some(), I want to update, otherwise - create. Seems like the ideal case for upsert, but it requires id to be specified! That's... confusing. Though I never used the original Prisma client and don't know if it's meant to be like that.

Also, the syntax for arguments is a bit confusing, too. Let's look at the example provided in docs:

let post: post::Data = client
    .post()
    .upsert( // First argument is a unique filter
        post::id::equals("upsert".to_string()),
        ( // Second argument is a tuple with the same values
          // as an equivalent call to create()
            true,
            "title".to_string(),
            "upsert".to_string(),
            vec![]
        ),
        vec![ // Final argument is a vec of updates
            post::content::set(Some("new content".to_string())),
            post::views::increment(1)
        ]
    )
    .exec()
    .await
    .unwrap();

I have two questions:

1. Why is the content::set(...) in the vec of updates and not in the vec in the tuple, like this?

let post: post::Data = client
    .post()
    .upsert( // First argument is a unique filter
        post::id::equals("upsert".to_string()),
        ( // Second argument is a tuple with the same values
          // as an equivalent call to create()
            true,
            "title".to_string(),
            "upsert".to_string(),
            vec![
                post::content::set(Some("new content".to_string())),
            ]
        ),
        vec![ // Final argument is a vec of updates
            post::views::increment(1)
        ]
    )
    .exec()
    .await
    .unwrap();

2. Why is the second argument must be in a tuple? Maybe the idea was in a mental separation, or is it clippy's too_many_arguments lint? If the former is true, I want to say that if feels unnecessary because the vec of updates is always the last argument. If the latter is true, I would suggest just attaching #[allow(clippy::too_many_arguments)] to the method in codegen part.

[Brendonovich]

While admittedly the example in the docs may not be the easiest to understand, the structure is correct. To address your points individually:

• Why provide the ID? Upsert exists to create a record if it doesn't exist, and if the record does exist then update it with the provided values. Prisma needs a unique field to determine whether the record exists or not and decide whether to create or update.

1. It could be, but content in the create section is already specified as "upsert".to_string(). The example is meant to demonstrate assigning different values depending on whether the record exists or not.
2. The structure of the arguments is upsert(where, create, update). where is the same argument as accepted as a call to find_unique, create is the same arguments accepted by a call to create, and update is the same arguments as accepted by a call to update. While I could flatten the create arguments into the arguments of upsert, I prefer having a distinct separation between which data is for creation and which is for updating. Either extra arguments to upsert or that tuple are required, as I have to be able to enforce that values for required fields are provided by the user.

[teenjuna]

Considering the second point: so if I just left the update vec empty, this operation won't modify existing record at all?

Considering the ID, is it possible to integrate Option in the future? Something like this:

1. ID is None -> upsert = create
2. ID is Some(id), but id doesn't exist -> upsert = create
3. ID is Some(id) and id exists -> upsert = update

[Brendonovich]

Considering the second point: so if I just left the update vec empty, this operation won't modify existing record at all?

Correct

Considering the ID, is it possible to integrate Option in the future? Something like this:

1. ID is None -> upsert = create
2. ID is Some(id), but id doesn't exist -> upsert = create
3. ID is Some(id) and id exists -> upsert = update

Most likely no. Prisma requires that the where argument is unique, which does not include None values. If you think about it, that doesn't even make sense. If id is None and there are multiple records with a NULL id then how should Prisma choose which record to upsert?

[teenjuna]

Fair enough! Thank for your answer :)