Add skip attribute to input objects

[mwilliammyers]

What do you think about adding a #[graphql(skip)] attribute for deriving GraphQLInputObject like GraphQLObject?

I often find myself wanting to skip fields for input objects (e.g. a database id that should be auto generated and not set by users).

Describe the solution you'd like It would be awesome to be able to do something like:

#[derive(juniper::GraphQLScalarValue)]
struct Id(String); // or `uuid::Uuid` etc.

impl Default for Id {
    fn default() -> Self {
        Id::generate()
    }
}

#[derive(serde::Serialize, juniper::GraphQLInputObject)]
struct UserInput {
    #[graphql(skip)]
    #[serde(default)]
    id: Id,
    
    name: String,
}

Describe alternatives you've considered

• Creating a separate struct with the same fields + the additional skipped fields and using that for the db. This is what I am currently doing and it is rather cumbersome and brittle. I would prefer to not have to use proc-macros to keep the fields of the two in sync (and to auto-generate From impls because that seems overkill and my project already takes way too long to compile on very fast hardware.
• Preprocessing the documents in the database to auto generate ids and set the id field. I would love to do this but Elasticsearch doesn't support this (see [this] issue (https://github.com/elastic/elasticsearch/issues/41163) and this issue).

Additional context In reality I would use it more like this with the elastic crate:

#[derive(serde::Serialize, juniper::GraphQLInputObject, elastic_derive::ElasticType)]
struct UserInput {
    #[graphql(skip)]
    #[elastic(id)]
    #[serde(default)]
    id: Id,
    
    name: String,
}

#[derive(serde::Deserialize, juniper::GraphQLObject, elastic_derive::ElasticType)]
struct User {
    // GraphQL API consumers should be able to *view* but not set the id field.
    #[elastic(id)]
    id: Id,
    
    name: String,
}

This would tell elastic to use the id field as the document ID and would create a document that looks like:

{
  "_index": "users",
  "_type": "_doc",
  "_id": "31cfad4a-814d-4e0b-bea9-52ceaefa8f96",
  "_source": {
    "id": "31cfad4a-814d-4e0b-bea9-52ceaefa8f96",
    "name": "Bob"
  }
}

[LegNeato]

Sure, this seems like something we should do for consistency.

[mwilliammyers]

Ok. I’ll see if I can open a PR for it.

[LegNeato]

Sounds good! You should probably hold off until we merge the async branch onto master (extremely soon)

[nebnes]

@mwilliammyers I am looking for add #[graphql(skip)] on GraphQLInputObject too. Have you news about this possible feature ?

[ilslv]

@tyranron should we add this feature as a part of #[derive(GraphQLInputObject)] or push users to create another struct with From impl? I'm in favour of this, mainly because we already allow skip on #[derive(GraphQLObject)], so Rust structs are already may not be describing pure GraphQL schema. Also approach with From impl isn't very refactoring friendly, as adding a field to a input object wouldn't raise error for missing field on From target.

[tyranron]

@ilslv I don't really like the idea of pushing users to something. We have already possibility to provide From for another struct. Having #[graphql(ignore)] in addition to that will be quite handy (imagine PhantomData or PhantomPinned field for whatever purpose).

Of course, we should require the Default impl for the skipped fields.