I'm trying to update a one to many relationship in Prisma. My schema looks like this
model A_User {
id Int #id
username String
age Int
bio String #db.VarChar(1000)
createdOn DateTime #default(now())
features A_Features[]
}
model A_Features {
id Int #id #default(autoincrement())
description String
A_User A_User? #relation(fields: [a_UserId], references: [id])
a_UserId Int?
}
I'm trying to add a couple of new features to user with id: 1, or update them if they are already there.
I'm trying doing something like
const post = await prisma.a_User.update({
where: { id: 1},
data: {
features: {
upsert: [
{ description: 'first feature'},
{ description: 'second feature'}
]
}
}
})
The compiler isn't happy, it tells me
Type '{ features: { upsert: { description: string; }[]; }; }' is not assignable to type '(Without<A_UserUpdateInput, A_UserUncheckedUpdateInput> & A_UserUncheckedUpdateInput) | (Without<...> & A_UserUpdateInput)'.
Object literal may only specify known properties, and 'features' does not exist in type '(Without<A_UserUpdateInput, A_UserUncheckedUpdateInput> & A_UserUncheckedUpdateInput) | (Without<...> & A_UserUpdateInput)'.ts(2322)
index.d.ts(1572, 5): The expected type comes from property 'data' which is declared here on type '{ select?: A_UserSelect; include?: A_UserInclude; data: (Without<A_UserUpdateInput, A_UserUncheckedUpdateInput> & A_UserUncheckedUpdateInput) | (Without<...> & A_UserUpdateInput); where: A_UserWhereUniqueInput; }'
(property) features: {
upsert: {
description: string;
}[];
}
I can't work out how to do it nor I can find clear help in the documentation. Any idea on how to implement it or where I can find some examples?
I'm providing my solution based on the clarifications you provided in the comments. First I would make the following changes to your Schema.
Changing the schema
model A_User {
id Int #id
username String
age Int
bio String #db.VarChar(1000)
createdOn DateTime #default(now())
features A_Features[]
}
model A_Features {
id Int #id #default(autoincrement())
description String #unique
users A_User[]
}
Notably, the relationship between A_User and A_Features is now many-to-many. So a single A_Features record can be connected to many A_User records (as well as the opposite).
Additionally, A_Features.description is now unique, so it's possible to uniquely search for a certain feature using just it's description.
You can read the Prisma Guide on Relations to learn more about many-to-many relations.
Writing the update query
Again, based on the clarification you provided in the comments, the update operation will do the following:
Overwrite existing features in a A_User record. So any previous features will be disconnected and replaced with the newly provided ones. Note that the previous features will not be deleted from A_Features table, but they will simply be disconnected from the A_User.features relation.
Create the newly provided features that do not yet exist in the A_Features table, and Connect the provided features that already exist in the A_Features table.
You can perform this operation using two separate update queries. The first update will Disconnect all previously connected features for the provided A_User. The second query will Connect or Create the newly provided features in the A_Features table. Finally, you can use the transactions API to ensure that both operations happen in order and together. The transactions API will ensure that if there is an error in any one of the two updates, then both will fail and be rolled back by the database.
//inside async function
const disconnectPreviouslyConnectedFeatures = prisma.a_User.update({
where: {id: 1},
data: {
features: {
set: [] // disconnecting all previous features
}
}
})
const connectOrCreateNewFeatures = prisma.a_User.update({
where: {id: 1},
data: {
features: {
// connect or create the new features
connectOrCreate: [
{
where: {
description: "'first feature'"
}, create: {
description: "'first feature'"
}
},
{
where: {
description: "second feature"
}, create: {
description: "second feature"
}
}
]
}
}
})
// transaction to ensure either BOTH operations happen or NONE of them happen.
await prisma.$transaction([disconnectPreviouslyConnectedFeatures, connectOrCreateNewFeatures ])
If you want a better idea of how connect, disconnect and connectOrCreate works, read the Nested Writes section of the Prisma Relation queries article in the docs.
The TypeScript definitions of prisma.a_User.update can tell you exactly what options it takes. That will tell you why the 'features' does not exist in type error is occurring. I imagine the object you're passing to data takes a different set of options than you are specifying; if you can inspect the TypeScript types, Prisma will tell you exactly what options are available.
If you're trying to add new features, and update specific ones, you would need to specify how Prisma can find an old feature (if it exists) to update that one. Upsert won't work in the way that you're currently using it; you need to provide some kind of identifier to the upsert call in order to figure out if the feature you're adding already exists.
https://www.prisma.io/docs/reference/api-reference/prisma-client-reference/#upsert
You need at least create (what data to pass if the feature does NOT exist), update (what data to pass if the feature DOES exist), and where (how Prisma can find the feature that you want to update or create.)
You also need to call upsert multiple times; one for each feature you're looking to update or create. You can batch the calls together with Promise.all in that case.
const upsertFeature1Promise = prisma.a_User.update({
data: {
// upsert call goes here, with "create", "update", and "where"
}
});
const upsertFeature2Promise = prisma.a_User.update({
data: {
// upsert call goes here, with "create", "update", and "where"
}
});
const [results1, results2] = await Promise.all([
upsertFeaturePromise1,
upsertFeaturePromise2
]);
Related
I am using prisma + mysql (on planetscale). When I link two items that are in different tables, I normally use connect or disconnect:
const getUser = await prisma.user.update({
where: {
id: 9
},
data: {
posts: {
| connect: {
| id: 11
| },
create: {
title: "My new post title"
}
}
}
})
I am wondering whether that's necessary or why that's necessary?
I also noticed that I can just update records in my database by updating the id (as a plain string), and it will still work. e.g.:
// example for updating a one-to-many relationship:
const getUser = await prisma.user.update({
where: {
id: 9
},
data: {
postId: "123192312i39123123"
}
}
})
... or if it's an explicit many-to-many relation, I can just edit the row in the relation-table & update the id.
Is this a bad way of doing things? Am I going to break something later down the line in doing it this way?
Your cloud provider is not relevant in the context of the question. It will not affect how your framework(prisma) behaves in updates.
I am wondering whether that's necessary or why that's necessary?
You have a user with a one to many relation: user => n posts.
You have an existing post in the db, and you want to add that post to the posts collection of a user.
That posts relation can be either explicit or implicit. The connect clause handles the addition of relation:
{
posts: {
connect: { id: 11 }
}
}
Without using the connect you'd have to create a new post:
{
posts: {
create: {
title: "My new post title"
}
}
}
update records in my database by updating the id (as a plain string)
Not sure what you mean here, mind sharing the schema?
or if it's an explicit many-to-many relation, I can just edit the row in the relation-table & update the id
If it's explicit many-to-many then it's OK to manually edit the id fields. As long as the ids are found and the relation makes sense, there's no problem with manual updates.
I'm trying to create a friendship mechanic for my app using Prisma among other tools. In the docs it shows the following example for how to create a many-to-many self relation:
model User {
id Int #id #default(autoincrement())
name String?
followedBy Follows[] #relation("following")
following Follows[] #relation("follower")
}
model Follows {
follower User #relation("follower", fields: [followerId], references: [id])
followerId Int
following User #relation("following", fields: [followingId], references: [id])
followingId Int
##id([followerId, followingId])
}
I have implemented this and it works, however the issue is that for friendships, there is no 'following' and 'followedBy', you're just friends. At the moment, when I query, I have to query both fields in order to find all of a user's friends. Is there any way to define this type of relationship with only one field? Whereby we just have a single list of friends on a user?
I agree that it would be nice if Prisma could more natively support this sort of self-relation where the relationship is expected to be symmetric (e.g. userA is friends with userB if and only if userB is friends with userA).
However, as far as I can tell Prisma insists on having two "sides" of the relationship. (If someone knows better, I would love to hear it!) So what follows is the approach I am taking, which avoids having to query both relations to find a user's full set of friends.
Concept
We'll use one "side" of the relation to contain the complete set of friends. The other "side" exists solely to meet Prisma's requirement, and we'll never query it directly.
When adding or removing a friend relationship, we'll make two prisma calls, one to update each object.
Code
Schema file:
model User {
id Int #id #default(autoincrement())
name String?
friends User[] #relation("UserFriends")
// This second "side" of the UserFriends relation exists solely
// to satisfy prisma's requirements; we won't access it directly.
symmetricFriends User[] #relation("UserFriends")
}
Methods to add and remove friendships (there's plenty of redundant code in here that could be abstracted out, but I think it's clearer to read this way):
const addFriendship = async (userIdA: string, userIdB: string) => {
await prisma.user.update({
where: {id: userIdA},
data: {friends: {connect: [{id: userIdB}]}},
});
await prisma.user.update({
where: {id: userIdB},
data: {friends: {connect: [{id: userIdA}]}},
});
};
const removeFriendship = async (userIdA: string, userIdB: string) => {
await prisma.user.update({
where: {id: userIdA},
data: {friends: {disconnect: [{id: userIdB}]}},
});
await prisma.user.update({
where: {id: userIdB},
data: {friends: {disconnect: [{id: userIdA}]}},
});
}
With this approach, one can load a user and get all their friends in the expected manner, e.g.
const getUserWithFriends = async (userId) =>
await prisma.user.find({
where: {id: userId},
include: {friends: true},
});
I have the following scenario, my application has two entities: box and items with N to N relationship. I am using sequelize with MySQL.
I am using pseudocode to represent the tables:
Box {
id: Integer primary key
name: String
}
Item {
id: Integer primary key
name: String
}
I have set up the schemas with relations hasMany in both directions using the following through relation:
Box.hasMany(Item, { through: Box_Item });
Item.hasMany(Box, { through: Box_Item });
Box_Item {
id_box: Integer,
id_item: Integer,
item_order: Integer
}
With primary_key(id_box, id_item).
I tested it and I can call myBox.getItems() on my instance object myBox and easily get all the items it has.
I can make calls as
BoxModel.findOne({
where: { id: 1 },
include: [{ model: ItemModel }]
});
And it automatically understands there is a relation between the models through Box_Item and get everything correctly, except that I'm not getting the results sorted by item_order field. This field is a number from 1 to N that represents the item order inside that box.
I tried
BoxModel.findOne({
where: { id: 1 },
include: [
{
model: ItemModel,
order: 'item_order'
}
]
});
But it seems sequelizejs does not support order inside include yet (checked on their github repo).
I tried to force
BoxModel.findOne({
where: { id: 1 },
order: '`box_model`.`item_order`'
include: [ { model: ItemModel } ]
})
looking through the query sequelize creates but it just put the ORDER BY in two different places (inside INNER JOIN and at the end of the query, don't know why...) and I got an error.
So I searched for this on stackoverflow (1), found a few questions but I don't get a good way for doing that using the ORM.
How could I get the items sorted by item_order field when asking for specific box items?
After a few days trying to get it done I found an answer on stackoverflow that helped me.
After creating the relationships between Box and Item I can easily call on an instance:
myBox.getItems({
order: '`box_model`.`item_order`'
});
And then I get the result I'm expecting. But I had to look through the query sequelize is creating based on the models and get the correct field based on their renaming rules.
If you want you can pass the as parameter and rename your tables.
I have a query I'm trying to perform based on a one to many relationship.
As an example there is a model called Users and one called Projects.
Users hasMany Projects
Projects have many types which are stored in a type (enum) column. There are 4 different types that potentially a user may have that I want to load. The catch is I want to include the most recent project record (createdAt column) for all networks that potentially will be there. I have not found a way to structure the query for it to work as an include. I have however found a way to do a raw query which does what I want.
I am looking for a way without having to do a raw query. By doing the raw query I have to map the returned results to users I've returned from the other method, or I have to do a simple include and then trim off all the results that are not the most recent. The latter is fine, but I see this getting slower as a user will have many projects and it will keep growing steadily.
This allow serialize a json for anywhere action about a model. Read it, very well
sequelize-virtual-fields
// define models
var Person = sequelize.define('Person', { name: Sequelize.STRING });
var Task = sequelize.define('Task', {
name: Sequelize.STRING,
nameWithPerson: {
type: Sequelize.VIRTUAL,
get: function() { return this.name + ' (' + this.Person.name + ')' }
attributes: [ 'name' ],
include: [ { model: Person, attributes: [ 'name' ] } ],
order: [ ['name'], [ Person, 'name' ] ]
}
});
// define associations
Task.belongsTo(Person);
Person.hasMany(Task);
// activate virtual fields functionality
sequelize.initVirtualFields();
Using sails 0.10.5/waterline 0.10.15:
I cannot find an answer to a simple question: how to count the elements of an association without using populate() (which would load all data).
Let take a simple many2many relation with via:
User:
attributes: {
following: {
collection: 'user',
via: 'follower',
dominant: true
},
follower: {
collection: 'user',
via: 'following'
}
Now I need the size of the collections.
Currently I try
User.findById(1).populateAll().exec(function(err, user) {
// count of followings -> user.following.length;
// count of followers-> user.follower.length;
}
which leads to loading the collections.
I'm missing a count function at collection level to avoid population/loading of data.
Is there a possibility to access the (auto generated) join tables to run a count-query directly on the join?
Something like:
User.findById(1).count({'followings'}).exec(function(err, followings) {
...}
or
UserFollowingFollow_FollowFollowing.countByUserFollowingFollowId(1).
exec(function(err, followings) {
...}
Waterline does offer the count query method and it can be used like this to solve your problem:
User.count().where({follower: followerId})
.exec(function(err, numberOfFollowings) {
//numberOfFollowings will be the integer that you need
})
followerId is the id that you are passing to User.findOne() in your example.
You can also read the Waterline documentation about this.