I've run into an issue while trying to extend my API to include a GraphQL endpoint. The application I'm working on is a kind of forum with Messages. A message can contain comments of type Message. If a message is a comment it has a parent of type Message. Simplified, the schema looks like this:
type Message {
id: String
content: String
comments: [Message]
parent: Message
}
type RootQuery {
message(id: String): Message
messages: [Message]
}
The problem with this schema is that it allows for queries like this:
{
messages {
comments {
parent {
comments {
parent {
comments {
parent {
id
content
}
}
}
}
}
}
}
}
Keep in mind that I may want to allow for arbitrarily deep nesting of comments. In that case the following query should be allowed:
{
messages {
comments {
comments {
comments {
id
content
}
}
}
}
}
So, my question is this: Should I introduce a new type - Comment - to the API that do not know of its parent? Or are there any other ways of restricting this kind of unwanted behaviour?
Also, would the use of a Comment-type prohibit me from using the fragment messageFields on Message syntax in my queries? Perhaps this is the time to introduce interfaces to the schema?
Suggestion to a solution if I introduce the type Comment (I have not tried this):
interface Message {
id: String
content: String
comments: [Message]
}
type DefaultMessage : Message {
id: String
content: String
comments: [Comment]
parent: Message
}
type Comment : Message {
id: String
content: String
comments: [Message]
}
type RootQuery {
message(id: String): Message
messages: [Message]
}
Just in case anyone else ends up here wondering how to do recursive types in graphql-js, there's a useful hint in graphql-js's code:
* When two types need to refer to each other, or a type needs to refer to
* itself in a field, you can use a function expression (aka a closure or a
* thunk) to supply the fields lazily.
*
* Example:
*
* var PersonType = new GraphQLObjectType({
* name: 'Person',
* fields: () => ({
* name: { type: GraphQLString },
* bestFriend: { type: PersonType },
* })
* });
*
*/
https://github.com/graphql/graphql-js/blob/master/src/type/definition.js#L274
If a message is a comment it has a parent of type Message.
Looks like the parent field should be under type Comment, not DefaultMessage. That still wouldn't prevent a parent - comments - parent query but if you're worried about this for a DDOS reason, there are many other types of requests that are difficult to compute, even with REST APIs, and you should have other measures to detect such an attack.
Recursive nodes
However, you pose a very interesting question with the nested comments. How would you know how many times you need to nest the comment in the query to get all nested responses? I don't think it's currently possible with GraphQL to specify recursive objects.
I'd probably go around this limitation by fetching each nested comment one by one (or by X levels at a time) starting from the last comment as the node
{
messages {
comments {
id
content
}
}
}
followed by
{
node(commendId) {
comment {
id
content
}
}
}
I suppose you have a depth attribute of the Comment data structure, which should be pretty useful, for example, to limit the max nested depth when the users are posting comments.
So that your problem could be solved like this: in the resolver of the comments attribute, check the depth, return nothing if the depth is going illegal, otherwise fetch the comments and return.
Related
I have a simple mutation editPerson. It changes the name and/or description of a person specified by an id.
I use this little snippet to call the mutator from React components:
function useEditPerson(variables) {
const gqlClient = useGQLClient();
const personFragment = gql`fragment useEditPerson__person on Person {
id
name
description
}`;
return useMutation(gql`
${personFragment}
mutation editPerson($id: ID!, $description: String, $name: String) {
editPerson(id: $id, description: $description, name: $name) {
...useEditPerson__person
}
}
`, {
variables,
optimisticResponse: vars => {
const person = gqlClient.readFragment({
id: vars.id,
fragment: personFragment,
});
return {
editPerson: {
__typename: "Person",
description: "",
name: "",
...person,
...vars,
},
};
},
});
}
This works well enough unless either the name or description for the indicated person hasn't yet been queried and does not exist in the cache; in this case person is null. This is expected from readFragment - any incomplete fragment does this.
The thing is I really need that data to avoid invariant errors - if they're not in the cache I'm totally okay using empty strings as default values, those values aren't displayed anywhere in the UI anyway.
Is there any way to read partial fragments from the cache? Is there a better way to get that data for the optimistic response?
I guess you use the snippet in the form that has all the data you need. So, you can pass the needed data to your useEditPerson hook through the arguments and then use in optimistic response, and then you won't need to use gqlClient.
For a mutation addVoucher there are a limited list of potential errors that can occur.
Voucher code invalid
Voucher has expired
Voucher has already been redeemed
At the moment I'm throwing a custom error when one of these occurs.
// On the server:
const addVoucherResolver = () => {
if(checkIfInvalid) {
throw new Error('Voucher code invalid')
}
return {
// data
}
}
Then on the client I search the message description so I can alert the user. However this feels brittle and also the GraphQL API doesn't automatically document the potential errors. Is there a way to define the potential errors in the GraphQL schema?
Currently my schema looks like this:
type Mutation {
addVoucherResolver(id: ID!): Order
}
type Order {
cost: Int!
}
It would be nice to be able to do something like this:
type Mutation {
addVoucherResolver(id: ID!): Order || VoucherError
}
type Order {
cost: Int!
}
enum ErrorType {
INVALID
EXPIRED
REDEEMED
}
type VoucherError {
status: ErrorType!
}
Then anyone consuming the API would know all the potential errors. This feels like a standard requirement to me but from reading up there doesn't seem to be a standardises GraphQL approach.
It's possible to use a Union or Interface to do what you're trying to accomplish:
type Mutation {
addVoucher(id: ID!): AddVoucherPayload
}
union AddVoucherPayload = Order | VoucherError
You're right that there isn't a standardized way to handle user-visible errors. With certain implementations, like apollo-server, it is possible to expose additional properties on the errors returned in the response, as described here. This does make parsing the errors easier, but is still not ideal.
A "Payload" pattern has emerged fairly recently for handling these errors as part of the schema. You see can see it in public API's like Shopify's. Instead of a Union like in the example above, we just utilize an Object Type:
type Mutation {
addVoucher(id: ID!): AddVoucherPayload
otherMutation: OtherMutationPayload
}
type AddVoucherPayload {
order: Order
errors: [Error!]!
}
type OtherMutationPayload {
something: Something
errors: [Error!]!
}
type Error {
message: String!
code: ErrorCode! # or a String if you like
}
enum ErrorCode {
INVALID_VOUCHER
EXPIRED_VOUCHER
REDEEMED_VOUCHER
# etc
}
Some implementations add a status or success field as well, although I find that making the actual data field (order is our example) nullable and then returning null when the mutation fails is also sufficient. We can even take this one step further and add an interface to help ensure consistency across our payload types:
interface Payload {
errors: [Error!]!
}
Of course, if you want to be more granular and distinguish between different types of errors to better document which mutation can return what set of errors, you won't be able to use an interface.
I've had success with this sort of approach, as it not only documents possible errors, but also makes it easier for clients to deal with them. It also means that any other errors that are returned with a response should serve as an immediately red flag that something has gone wrong with either the client or the server. YMMV.
You can use scalar type present in graphql
just write scalar JSON and return any JSON type where you want to return it.
`
scalar JSON
type Response {
status: Boolean
message: String
data: [JSON]
}
`
Here is Mutation which return Response
`
type Mutation {
addVoucherResolver(id: ID!): Response
}
`
You can return from resolver
return {
status: false,
message: 'Voucher code invalid(or any error based on condition)',
data: null
}
or
return {
status: true,
message: 'Order fetch successfully.',
data: [{
object of order
}]
}
on Front end you can use status key to identify response is fetch or error occurs.
I am working on the following types, where the "content" of a "Comment" is a union type:
type TextContent {
text: String
}
type RichContent {
participants: [String]
startTime: String
}
union Content = TextContent | RichContent
type Comment {
id: ID
sender: String
content: Content
}
type Review {
id: ID
title: String
lastComment: Comment
}
In my Apollo query, I was trying to use conditional fragments on the 2 Content types:
query listOfReviews {
reviews {
...reviewFields
}
}
fragment reviewFields on Review {
id
title
lastComment {
content {
... on TextContent {
text
}
... on RichContent {
participants
startTime
}
}
}
}
I received a runtime error where Apollo seems trying to access "participants" field of "undefined", where the actual content object is:
{
__typename: "TextContent:,
text: "abc"
}
It looks the two types of the union Content are merged together.
My question is: is it allowed to use type conditions on nested fields in Apollo queries? Or type conditions have to be used on the top level types returned by the queries? If it's allowed, how should I fix my types/queries?
Thanks a lot!
#const86 helped point out that this is due to this bug: https://github.com/apollographql/apollo-link-state/pull/258.
Using facebook's reference library, I found a way to hack generic types like this:
type PagedResource<Query, Item> = (pagedQuery: PagedQuery<Query>) => PagedResponse<Item>
interface PagedQuery<Query> {
query: Query;
take: number;
skip: number;
}
interface PagedResponse<Item> {
items: Array<Item>;
total: number;
}
function pagedResource({type, resolve, args}) {
return {
type: pagedType(type),
args: Object.assign(args, {
page: { type: new GraphQLNonNull(pageQueryType()) }
}),
resolve
};
function pageQueryType() {
return new GraphQLInputObjectType({
name: 'PageQuery',
fields: {
skip: { type: new GraphQLNonNull(GraphQLInt) },
take: { type: new GraphQLNonNull(GraphQLInt) }
}
});
}
function pagedType(type) {
return new GraphQLObjectType({
name: 'Paged' + type.toString(),
fields: {
items: { type: new GraphQLNonNull(new GraphQLList(type)) },
total: { type: new GraphQLNonNull(GraphQLInt) }
}
});
}
}
But I like how with Apollo Server I can declaratively create the schema. So question is, how do you guys go about creating generic-like types with the schema language?
You can create an interface or union to achieve a similar result. I think this article does a good job explaining how to implement interfaces and unions correctly. Your schema would look something like this:
type Query {
pagedQuery(page: PageInput!): PagedResult
}
input PageInput {
skip: Int!
take: Int!
}
type PagedResult {
items: [Pageable!]!
total: Int
}
# Regular type definitions for Bar, Foo, Baz types...
union Pageable = Bar | Foo | Baz
You also need to define a resolveType method for the union. With graphql-tools, this is done through the resolvers:
const resolvers = {
Query: { ... },
Pageable {
__resolveType: (obj) => {
// resolve logic here, needs to return a string specifying type
// i.e. if (obj.__typename == 'Foo') return 'Foo'
}
}
}
__resolveType takes the business object being resolved as its first argument (typically your raw DB result that you give GraphQL to resolve). You need to apply some logic here to figure out of all the different Pageable types, which one we're handling. With most ORMs, you can just add some kind of typename field to the model instance you're working with and just have resolveType return that.
Edit: As you pointed out, the downside to this approach is that the returned type in items is no longer transparent to the client -- the client would have to know what type is being returned and specify the fields for items within an inline fragment like ... on Foo. Of course, your clients will still have to have some idea about what type is being returned, otherwise they won't know what fields to request.
I imagine creating generics the way you want is impossible when generating a schema declaratively. To get your schema to work the same way it currently does, you would have to bite the bullet and define PagedFoo when you define Foo, define PagedBar when you define Bar and so on.
The only other alternative I can think of is to combine the two approaches. Create your "base" schema programatically. You would only need to define the paginated queries under the Root Query using your pagedResource function. You can then use printSchema from graphql/utilities to convert it to a String that can be concatenated with the rest of your type definitions. Within your type definitions, you can use the extend keyword to build on any of the types already declared in the base schema, like this:
extend Query {
nonPaginatedQuery: Result
}
If you go this route, you can skip passing a resolve function to pagedResource, or defining any resolvers on your programatically-defined types, and just utilize the resolvers object you normally pass to buildExecutableSchema.
I've been playing with GraphQL recently, and am currently learning about mutations. I'm a bit confused with something. I have a model Post with relation Comments. I have a mutation that looks like this:
mutation addCommentToPost {
updatePost(
id: "POST-1",
comments: [{
body: "Hello!"
}]
) {
id,
comments {
id,
body
}
}
}
The problem is, whenever I run this, it seems to remove all the comments and sets the comments to only the one I just added. To be more specific, how do I write a mutation that pushes to the comments array rather than replacing it?
You are using a mutation called updatePosts, which I assume (based on the name) simply updates a post by replacing the fields that are passed. If you want to use the updatePosts mutation to add a comment, you will first have to query for the post to get the current list of comments, add your comment to the end, and then call updateComment with the entire list of comments (including the one that you just added to the end).
However, this isn't really a good solution, especially if the list of comments is potentially very long. If you have the ability to change the GraphQL server, you should create a new mutation on the server with a signature like addComment(postId: ID, comment: CommentInput). In the resolve function for that mutation, simply add the comment that is passed to the end of the list of current comments.
// resolver for addComment:
addComment(root, args) {
// validate inputs here ...
const post = db.getPost(args.postId);
post.comments.append(args.comment);
db.writePost(post.id, post);
}
db.getPost and db.writePost are functions you have to define yourself to retrieve/write a post from/to wherever you store it.
It's important to note that unlike a SQL or Mongo query, a GraphQL mutation itself doesn't have any meaning without the resolve functions. What the mutation does is defined entirely inside its resolve function. Mutation names and arguments only gain meaning together with the resolve function. It's up to you (or the GraphQL server developers in your company) to write the resolve functions.
The way this situation is currently solved in the Graphcool API is to use a create mutation for the Comment that links to the Post. This is called a nested connect mutation.
This is how it would look like:
mutation {
createComment(
text: "Hello!"
postId: "POST-1"
) {
id
text
post {
comments {
id
}
}
}
}
In the future, other nested arguments like comments_set or comments_push could be introduced, then pushing would be possible like this:
mutation addCommentToPost {
updatePost(
id: "POST-1",
comments_push: [{
body: "Hello!"
}]
) {
id,
comments {
id,
body
}
}
}
Disclosure: I work at Graphcool.
You can use those code as an example for mutation.
module.exports = (refs) => ({
type: refs.commentType,
args: {
id: {
type: GraphQLString
},
body: {
type: GraphQLString
}
},
resolve: (parent, args, root) => {
return createUser(args);
}
});