Code Monkey home page Code Monkey logo

nullability-wg's Introduction

GraphQL

GraphQL Logo

The GraphQL specification is edited in the markdown files found in /spec the latest release of which is published at https://graphql.github.io/graphql-spec/.

The latest draft specification can be found at https://graphql.github.io/graphql-spec/draft/ which tracks the latest commit to the main branch in this repository.

Previous releases of the GraphQL specification can be found at permalinks that match their release tag. For example, https://graphql.github.io/graphql-spec/October2016/. If you are linking directly to the GraphQL specification, it's best to link to a tagged permalink for the particular referenced version.

Overview

This is a Working Draft of the Specification for GraphQL, a query language for APIs created by Facebook.

The target audience for this specification is not the client developer, but those who have, or are actively interested in, building their own GraphQL implementations and tools.

In order to be broadly adopted, GraphQL will have to target a wide variety of backend environments, frameworks, and languages, which will necessitate a collaborative effort across projects and organizations. This specification serves as a point of coordination for this effort.

Looking for help? Find resources from the community.

Getting Started

GraphQL consists of a type system, query language and execution semantics, static validation, and type introspection, each outlined below. To guide you through each of these components, we've written an example designed to illustrate the various pieces of GraphQL.

This example is not comprehensive, but it is designed to quickly introduce the core concepts of GraphQL, to provide some context before diving into the more detailed specification or the GraphQL.js reference implementation.

The premise of the example is that we want to use GraphQL to query for information about characters and locations in the original Star Wars trilogy.

Type System

At the heart of any GraphQL implementation is a description of what types of objects it can return, described in a GraphQL type system and returned in the GraphQL Schema.

For our Star Wars example, the starWarsSchema.ts file in GraphQL.js defines this type system.

The most basic type in the system will be Human, representing characters like Luke, Leia, and Han. All humans in our type system will have a name, so we define the Human type to have a field called "name". This returns a String, and we know that it is not null (since all Humans have a name), so we will define the "name" field to be a non-nullable String. Using a shorthand notation that we will use throughout the spec and documentation, we would describe the human type as:

type Human {
  name: String
}

This shorthand is convenient for describing the basic shape of a type system; the JavaScript implementation is more full-featured, and allows types and fields to be documented. It also sets up the mapping between the type system and the underlying data; for a test case in GraphQL.js, the underlying data is a set of JavaScript objects, but in most cases the backing data will be accessed through some service, and this type system layer will be responsible for mapping from types and fields to that service.

A common pattern in many APIs, and indeed in GraphQL is to give objects an ID that can be used to refetch the object. So let's add that to our Human type. We'll also add a string for their home planet.

type Human {
  id: String
  name: String
  homePlanet: String
}

Since we're talking about the Star Wars trilogy, it would be useful to describe the episodes in which each character appears. To do so, we'll first define an enum, which lists the three episodes in the trilogy:

enum Episode {
  NEWHOPE
  EMPIRE
  JEDI
}

Now we want to add a field to Human describing what episodes they were in. This will return a list of Episodes:

type Human {
  id: String
  name: String
  appearsIn: [Episode]
  homePlanet: String
}

Now, let's introduce another type, Droid:

type Droid {
  id: String
  name: String
  appearsIn: [Episode]
  primaryFunction: String
}

Now we have two types! Let's add a way of going between them: humans and droids both have friends. But humans can be friends with both humans and droids. How do we refer to either a human or a droid?

If we look, we note that there's common functionality between humans and droids; they both have IDs, names, and episodes in which they appear. So we'll add an interface, Character, and make both Human and Droid implement it. Once we have that, we can add the friends field, that returns a list of Characters.

Our type system so far is:

enum Episode {
  NEWHOPE
  EMPIRE
  JEDI
}

interface Character {
  id: String
  name: String
  friends: [Character]
  appearsIn: [Episode]
}

type Human implements Character {
  id: String
  name: String
  friends: [Character]
  appearsIn: [Episode]
  homePlanet: String
}

type Droid implements Character {
  id: String
  name: String
  friends: [Character]
  appearsIn: [Episode]
  primaryFunction: String
}

One question we might ask, though, is whether any of those fields can return null. By default, null is a permitted value for any type in GraphQL, since fetching data to fulfill a GraphQL query often requires talking to different services that may or may not be available. However, if the type system can guarantee that a type is never null, then we can mark it as Non Null in the type system. We indicate that in our shorthand by adding an "!" after the type. We can update our type system to note that the id is never null.

Note that while in our current implementation, we can guarantee that more fields are non-null (since our current implementation has hard-coded data), we didn't mark them as non-null. One can imagine we would eventually replace our hardcoded data with a backend service, which might not be perfectly reliable; by leaving these fields as nullable, we allow ourselves the flexibility to eventually return null to indicate a backend error, while also telling the client that the error occurred.

enum Episode {
  NEWHOPE
  EMPIRE
  JEDI
}

interface Character {
  id: String!
  name: String
  friends: [Character]
  appearsIn: [Episode]
}

type Human implements Character {
  id: String!
  name: String
  friends: [Character]
  appearsIn: [Episode]
  homePlanet: String
}

type Droid implements Character {
  id: String!
  name: String
  friends: [Character]
  appearsIn: [Episode]
  primaryFunction: String
}

We're missing one last piece: an entry point into the type system.

When we define a schema, we define an object type that is the basis for all query operations. The name of this type is Query by convention, and it describes our public, top-level API. Our Query type for this example will look like this:

type Query {
  hero(episode: Episode): Character
  human(id: String!): Human
  droid(id: String!): Droid
}

In this example, there are three top-level operations that can be done on our schema:

  • hero returns the Character who is the hero of the Star Wars trilogy; it takes an optional argument that allows us to fetch the hero of a specific episode instead.
  • human accepts a non-null string as a query argument, a human's ID, and returns the human with that ID.
  • droid does the same for droids.

These fields demonstrate another feature of the type system, the ability for a field to specify arguments that configure their behavior.

When we package the whole type system together, defining the Query type above as our entry point for queries, this creates a GraphQL Schema.

This example just scratched the surface of the type system. The specification goes into more detail about this topic in the "Type System" section, and the type directory in GraphQL.js contains code implementing a specification-compliant GraphQL type system.

Query Syntax

GraphQL queries declaratively describe what data the issuer wishes to fetch from whoever is fulfilling the GraphQL query.

For our Star Wars example, the starWarsQueryTests.js file in the GraphQL.js repository contains a number of queries and responses. That file is a test file that uses the schema discussed above and a set of sample data, located in starWarsData.js. This test file can be run to exercise the reference implementation.

An example query on the above schema would be:

query HeroNameQuery {
  hero {
    name
  }
}

The initial line, query HeroNameQuery, defines a query with the operation name HeroNameQuery that starts with the schema's root query type; in this case, Query. As defined above, Query has a hero field that returns a Character, so we'll query for that. Character then has a name field that returns a String, so we query for that, completing our query. The result of this query would then be:

{
  "hero": {
    "name": "R2-D2"
  }
}

Specifying the query keyword and an operation name is only required when a GraphQL document defines multiple operations. We therefore could have written the previous query with the query shorthand:

{
  hero {
    name
  }
}

Assuming that the backing data for the GraphQL server identified R2-D2 as the hero. The response continues to vary based on the request; if we asked for R2-D2's ID and friends with this query:

query HeroNameAndFriendsQuery {
  hero {
    id
    name
    friends {
      id
      name
    }
  }
}

then we'll get back a response like this:

{
  "hero": {
    "id": "2001",
    "name": "R2-D2",
    "friends": [
      {
        "id": "1000",
        "name": "Luke Skywalker"
      },
      {
        "id": "1002",
        "name": "Han Solo"
      },
      {
        "id": "1003",
        "name": "Leia Organa"
      }
    ]
  }
}

One of the key aspects of GraphQL is its ability to nest queries. In the above query, we asked for R2-D2's friends, but we can ask for more information about each of those objects. So let's construct a query that asks for R2-D2's friends, gets their name and episode appearances, then asks for each of their friends.

query NestedQuery {
  hero {
    name
    friends {
      name
      appearsIn
      friends {
        name
      }
    }
  }
}

which will give us the nested response

{
  "hero": {
    "name": "R2-D2",
    "friends": [
      {
        "name": "Luke Skywalker",
        "appearsIn": ["NEWHOPE", "EMPIRE", "JEDI"],
        "friends": [
          { "name": "Han Solo" },
          { "name": "Leia Organa" },
          { "name": "C-3PO" },
          { "name": "R2-D2" }
        ]
      },
      {
        "name": "Han Solo",
        "appearsIn": ["NEWHOPE", "EMPIRE", "JEDI"],
        "friends": [
          { "name": "Luke Skywalker" },
          { "name": "Leia Organa" },
          { "name": "R2-D2" }
        ]
      },
      {
        "name": "Leia Organa",
        "appearsIn": ["NEWHOPE", "EMPIRE", "JEDI"],
        "friends": [
          { "name": "Luke Skywalker" },
          { "name": "Han Solo" },
          { "name": "C-3PO" },
          { "name": "R2-D2" }
        ]
      }
    ]
  }
}

The Query type above defined a way to fetch a human given their ID. We can use it by hard-coding the ID in the query:

query FetchLukeQuery {
  human(id: "1000") {
    name
  }
}

to get

{
  "human": {
    "name": "Luke Skywalker"
  }
}

Alternately, we could have defined the query to have a query parameter:

query FetchSomeIDQuery($someId: String!) {
  human(id: $someId) {
    name
  }
}

This query is now parameterized by $someId; to run it, we must provide that ID. If we ran it with $someId set to "1000", we would get Luke; set to "1002", we would get Han. If we passed an invalid ID here, we would get null back for the human, indicating that no such object exists.

Notice that the key in the response is the name of the field, by default. It is sometimes useful to change this key, for clarity or to avoid key collisions when fetching the same field with different arguments.

We can do that with field aliases, as demonstrated in this query:

query FetchLukeAliased {
  luke: human(id: "1000") {
    name
  }
}

We aliased the result of the human field to the key luke. Now the response is:

{
  "luke": {
    "name": "Luke Skywalker"
  }
}

Notice the key is "luke" and not "human", as it was in our previous example where we did not use the alias.

This is particularly useful if we want to use the same field twice with different arguments, as in the following query:

query FetchLukeAndLeiaAliased {
  luke: human(id: "1000") {
    name
  }
  leia: human(id: "1003") {
    name
  }
}

We aliased the result of the first human field to the key luke, and the second to leia. So the result will be:

{
  "luke": {
    "name": "Luke Skywalker"
  },
  "leia": {
    "name": "Leia Organa"
  }
}

Now imagine we wanted to ask for Luke and Leia's home planets. We could do so with this query:

query DuplicateFields {
  luke: human(id: "1000") {
    name
    homePlanet
  }
  leia: human(id: "1003") {
    name
    homePlanet
  }
}

but we can already see that this could get unwieldy, since we have to add new fields to both parts of the query. Instead, we can extract out the common fields into a fragment, and include the fragment in the query, like this:

query UseFragment {
  luke: human(id: "1000") {
    ...HumanFragment
  }
  leia: human(id: "1003") {
    ...HumanFragment
  }
}

fragment HumanFragment on Human {
  name
  homePlanet
}

Both of those queries give this result:

{
  "luke": {
    "name": "Luke Skywalker",
    "homePlanet": "Tatooine"
  },
  "leia": {
    "name": "Leia Organa",
    "homePlanet": "Alderaan"
  }
}

The UseFragment and DuplicateFields queries will both get the same result, but UseFragment is less verbose; if we wanted to add more fields, we could add it to the common fragment rather than copying it into multiple places.

We defined the type system above, so we know the type of each object in the output; the query can ask for that type using the special field __typename, defined on every object.

query CheckTypeOfR2 {
  hero {
    __typename
    name
  }
}

Since R2-D2 is a droid, this will return

{
  "hero": {
    "__typename": "Droid",
    "name": "R2-D2"
  }
}

This was particularly useful because hero was defined to return a Character, which is an interface; we might want to know what concrete type was actually returned. If we instead asked for the hero of Episode V:

query CheckTypeOfLuke {
  hero(episode: EMPIRE) {
    __typename
    name
  }
}

We would find that it was Luke, who is a Human:

{
  "hero": {
    "__typename": "Human",
    "name": "Luke Skywalker"
  }
}

As with the type system, this example just scratched the surface of the query language. The specification goes into more detail about this topic in the "Language" section, and the language directory in GraphQL.js contains code implementing a specification-compliant GraphQL query language parser and lexer.

Validation

By using the type system, it can be predetermined whether a GraphQL query is valid or not. This allows servers and clients to effectively inform developers when an invalid query has been created, without having to rely on runtime checks.

For our Star Wars example, the file starWarsValidationTests.js contains a number of demonstrations of invalid operations, and is a test file that can be run to exercise the reference implementation's validator.

To start, let's take a complex valid query. This is the NestedQuery example from the above section, but with the duplicated fields factored out into a fragment:

query NestedQueryWithFragment {
  hero {
    ...NameAndAppearances
    friends {
      ...NameAndAppearances
      friends {
        ...NameAndAppearances
      }
    }
  }
}

fragment NameAndAppearances on Character {
  name
  appearsIn
}

And this query is valid. Let's take a look at some invalid queries!

When we query for fields, we have to query for a field that exists on the given type. So as hero returns a Character, we have to query for a field on Character. That type does not have a favoriteSpaceship field, so this query:

# INVALID: favoriteSpaceship does not exist on Character
query HeroSpaceshipQuery {
  hero {
    favoriteSpaceship
  }
}

is invalid.

Whenever we query for a field and it returns something other than a scalar or an enum, we need to specify what data we want to get back from the field. Hero returns a Character, and we've been requesting fields like name and appearsIn on it; if we omit that, the query will not be valid:

# INVALID: hero is not a scalar, so fields are needed
query HeroNoFieldsQuery {
  hero
}

Similarly, if a field is a scalar, it doesn't make sense to query for additional fields on it, and doing so will make the query invalid:

# INVALID: name is a scalar, so fields are not permitted
query HeroFieldsOnScalarQuery {
  hero {
    name {
      firstCharacterOfName
    }
  }
}

Earlier, it was noted that a query can only query for fields on the type in question; when we query for hero which returns a Character, we can only query for fields that exist on Character. What happens if we want to query for R2-D2s primary function, though?

# INVALID: primaryFunction does not exist on Character
query DroidFieldOnCharacter {
  hero {
    name
    primaryFunction
  }
}

That query is invalid, because primaryFunction is not a field on Character. We want some way of indicating that we wish to fetch primaryFunction if the Character is a Droid, and to ignore that field otherwise. We can use the fragments we introduced earlier to do this. By setting up a fragment defined on Droid and including it, we ensure that we only query for primaryFunction where it is defined.

query DroidFieldInFragment {
  hero {
    name
    ...DroidFields
  }
}

fragment DroidFields on Droid {
  primaryFunction
}

This query is valid, but it's a bit verbose; named fragments were valuable above when we used them multiple times, but we're only using this one once. Instead of using a named fragment, we can use an inline fragment; this still allows us to indicate the type we are querying on, but without naming a separate fragment:

query DroidFieldInInlineFragment {
  hero {
    name
    ... on Droid {
      primaryFunction
    }
  }
}

This has just scratched the surface of the validation system; there are a number of validation rules in place to ensure that a GraphQL query is semantically meaningful. The specification goes into more detail about this topic in the "Validation" section, and the validation directory in GraphQL.js contains code implementing a specification-compliant GraphQL validator.

Introspection

It's often useful to ask a GraphQL schema for information about what queries it supports. GraphQL allows us to do so using the introspection system!

For our Star Wars example, the file starWarsIntrospectionTests.js contains a number of queries demonstrating the introspection system, and is a test file that can be run to exercise the reference implementation's introspection system.

We designed the type system, so we know what types are available, but if we didn't, we can ask GraphQL, by querying the __schema field, always available on the root type of a Query. Let's do so now, and ask what types are available.

query IntrospectionTypeQuery {
  __schema {
    types {
      name
    }
  }
}

and we get back:

{
  "__schema": {
    "types": [
      {
        "name": "Query"
      },
      {
        "name": "Character"
      },
      {
        "name": "Human"
      },
      {
        "name": "String"
      },
      {
        "name": "Episode"
      },
      {
        "name": "Droid"
      },
      {
        "name": "__Schema"
      },
      {
        "name": "__Type"
      },
      {
        "name": "__TypeKind"
      },
      {
        "name": "Boolean"
      },
      {
        "name": "__Field"
      },
      {
        "name": "__InputValue"
      },
      {
        "name": "__EnumValue"
      },
      {
        "name": "__Directive"
      }
    ]
  }
}

Wow, that's a lot of types! What are they? Let's group them:

  • Query, Character, Human, Episode, Droid - These are the ones that we defined in our type system.
  • String, Boolean - These are built-in scalars that the type system provided.
  • __Schema, __Type, __TypeKind, __Field, __InputValue, __EnumValue, __Directive - These all are preceded with a double underscore, indicating that they are part of the introspection system.

Now, let's try and figure out a good place to start exploring what queries are available. When we designed our type system, we specified what type all queries would start at; let's ask the introspection system about that!

query IntrospectionQueryTypeQuery {
  __schema {
    queryType {
      name
    }
  }
}

and we get back:

{
  "__schema": {
    "queryType": {
      "name": "Query"
    }
  }
}

And that matches what we said in the type system section, that the Query type is where we will start! Note that the naming here was just by convention; we could have named our Query type anything else, and it still would have been returned here if we had specified it as the starting type for queries. Naming it Query, though, is a useful convention.

It is often useful to examine one specific type. Let's take a look at the Droid type:

query IntrospectionDroidTypeQuery {
  __type(name: "Droid") {
    name
  }
}

and we get back:

{
  "__type": {
    "name": "Droid"
  }
}

What if we want to know more about Droid, though? For example, is it an interface or an object?

query IntrospectionDroidKindQuery {
  __type(name: "Droid") {
    name
    kind
  }
}

and we get back:

{
  "__type": {
    "name": "Droid",
    "kind": "OBJECT"
  }
}

kind returns a __TypeKind enum, one of whose values is OBJECT. If we asked about Character instead:

query IntrospectionCharacterKindQuery {
  __type(name: "Character") {
    name
    kind
  }
}

and we get back:

{
  "__type": {
    "name": "Character",
    "kind": "INTERFACE"
  }
}

We'd find that it is an interface.

It's useful for an object to know what fields are available, so let's ask the introspection system about Droid:

query IntrospectionDroidFieldsQuery {
  __type(name: "Droid") {
    name
    fields {
      name
      type {
        name
        kind
      }
    }
  }
}

and we get back:

{
  "__type": {
    "name": "Droid",
    "fields": [
      {
        "name": "id",
        "type": {
          "name": null,
          "kind": "NON_NULL"
        }
      },
      {
        "name": "name",
        "type": {
          "name": "String",
          "kind": "SCALAR"
        }
      },
      {
        "name": "friends",
        "type": {
          "name": null,
          "kind": "LIST"
        }
      },
      {
        "name": "appearsIn",
        "type": {
          "name": null,
          "kind": "LIST"
        }
      },
      {
        "name": "primaryFunction",
        "type": {
          "name": "String",
          "kind": "SCALAR"
        }
      }
    ]
  }
}

Those are our fields that we defined on Droid!

id looks a bit weird there, it has no name for the type. That's because it's a "wrapper" type of kind NON_NULL. If we queried for ofType on that field's type, we would find the String type there, telling us that this is a non-null String.

Similarly, both friends and appearsIn have no name, since they are the LIST wrapper type. We can query for ofType on those types, which will tell us what these are lists of.

query IntrospectionDroidWrappedFieldsQuery {
  __type(name: "Droid") {
    name
    fields {
      name
      type {
        name
        kind
        ofType {
          name
          kind
        }
      }
    }
  }
}

and we get back:

{
  "__type": {
    "name": "Droid",
    "fields": [
      {
        "name": "id",
        "type": {
          "name": null,
          "kind": "NON_NULL",
          "ofType": {
            "name": "String",
            "kind": "SCALAR"
          }
        }
      },
      {
        "name": "name",
        "type": {
          "name": "String",
          "kind": "SCALAR",
          "ofType": null
        }
      },
      {
        "name": "friends",
        "type": {
          "name": null,
          "kind": "LIST",
          "ofType": {
            "name": "Character",
            "kind": "INTERFACE"
          }
        }
      },
      {
        "name": "appearsIn",
        "type": {
          "name": null,
          "kind": "LIST",
          "ofType": {
            "name": "Episode",
            "kind": "ENUM"
          }
        }
      },
      {
        "name": "primaryFunction",
        "type": {
          "name": "String",
          "kind": "SCALAR",
          "ofType": null
        }
      }
    ]
  }
}

Let's end with a feature of the introspection system particularly useful for tooling; let's ask the system for documentation!

query IntrospectionDroidDescriptionQuery {
  __type(name: "Droid") {
    name
    description
  }
}

yields

{
  "__type": {
    "name": "Droid",
    "description": "A mechanical creature in the Star Wars universe."
  }
}

So we can access the documentation about the type system using introspection, and create documentation browsers, or rich IDE experiences.

This has just scratched the surface of the introspection system; we can query for enum values, what interfaces a type implements, and more. We can even introspect on the introspection system itself. The specification goes into more detail about this topic in the "Introspection" section, and the introspection file in GraphQL.js contains code implementing a specification-compliant GraphQL query introspection system.

Additional Content

This README walked through the GraphQL.js reference implementation's type system, query execution, validation, and introspection systems. There's more in both GraphQL.js and specification, including a description and implementation for executing queries, how to format a response, explaining how a type system maps to an underlying implementation, and how to format a GraphQL response, as well as the grammar for GraphQL.

Contributing to this repo

This repository is managed by EasyCLA. Project participants must sign the free (GraphQL Specification Membership agreement before making a contribution. You only need to do this one time, and it can be signed by individual contributors or their employers.

To initiate the signature process please open a PR against this repo. The EasyCLA bot will block the merge if we still need a membership agreement from you.

You can find detailed information here. If you have issues, please email [email protected].

If your company benefits from GraphQL and you would like to provide essential financial support for the systems and people that power our community, please also consider membership in the GraphQL Foundation.

nullability-wg's People

Contributors

anthonymdev avatar aprilrd avatar benjie avatar bod avatar calvincestari avatar captbaritone avatar ernieturner avatar fotoetienne avatar martinbonnin avatar mjmahone avatar twof avatar

Stargazers

 avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar

Watchers

 avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar

nullability-wg's Issues

Rename Working Group to "Nullability Working Group"

Context:
This working group set out to solve the issues discussed in the original Client Controlled Nullability RFC however, the solution to the problem has evolved past the solution described in that RFC to include the solutions described in the the True Schema Nullability discussions. For the time being CCN is being put on pause so that efforts can be redirected to True Schema Nullability.


Note: Action Item issues are reviewed and closed during Working Group
meetings.

Write up concerns regarding the try/catch method proposed for CCN

Benjie expressed concerns about the try/catch interpretation of CCN and would prefer to just use the mechanisms that GraphQL already has. He intends to write up some examples to demonstrate the issues that he foresees.


Note: Action Item issues are reviewed and closed during Working Group
meetings.

CCN: A path to True Nullability Schema

TL;DR I see a path where CCN’s ? could be leveraged by smart clients to safely expose the true resolver-nullability of fields directly to product code.

Prelude: CCN Behavior Definition

Since Client Controlled Nullability (CCN) may have a different meaning to different people, I’ll start by specifying my hope for how CCN will work. The rest of this post assumes this behavior:

Under CCN, the ! and ? annotations would allow the query to override the schema nullability of a field within a selection for the purposes of the execution of that selection.

  • ! means: Treat this field, in this selection, as if it were non-nullable
  • ? means: Treat this field, in this selection, as nullable

In other words, for the purposes of executing a query selection, every place the spec refers a field’s schema nullability, it would instead refer to the field’s nullability within the selection, which may or may not have been modified by CCN annotations in the query. Beyond that, all error handling and null bubbling behaviors of the current spec would be unchanged. Note that this includes the fact that errors thrown by a ? field would still be included in the response errors metadata.


Client-defined resiliency

GraphQL’s current recommended approach to providing response resiliency in the face of resolver errors is to make fields in the schema nullable by default. Unfortunately, this has the effect of obscuring the true nullability of fields. Clients, and even users, can’t tell from the schema alone if null is expected as a possible value, or if the field will only return null in exceptional (error) cases.

In this world of nullable-by-default schemas, the Client Controlled Nullability (CCN) proposal is primarily a tool to add assertions via !. While this is a marked ergonomic improvement, these assertions must be added blindly, without knowing if null is an expected value or not. This is at best awkward and at worst dangerous.

However, CCN’s ?, opens up the up the possibility of a different mechanism to achieve request resiliency. One which avoids obscuring the true nullability of fields. Specifically, an approach where we shift expectation from “it is the server/schema’s responsibility to make requests resilient to errors by typing fields as nullable” to “it is the client/queries responsibility to make requests resilient to errors by annotating non-nullable fields with ?”.

With this approach to resiliency, the schema could specify the “true” nullability of the fields.

For simplistic clients, e.g. Curl, the client/user can now see the true nullability of each field in the schema and add the appropriate amount of resilience for their use case using CCN’s ?. In a sense this is the same as CCN’s ! applied to a fully nullable schema, in that the client is empowered to declare which fields it can manage without, and which fields it requires. We’ve “simply” inverted the default. Of course, defaults are tremendously powerful and this tradeoff should be considered carefully. See “The power of defaults” below.

The opportunity for smart clients

For smart clients, this approach can not only let users “see” the true nullability, it can actually let product code interacted with generate types that model this true nullability. I see this as a fundamental solution of the actual problem that CCN initially set out to solve.

If smart clients can transform errored fields into contained thrown exceptions, that would mean product code should never encounter a null value due to a resolver error. In that case, the types that the smart client generates for its fragments/queries could safely express the true nullability of those fields on the server.

This is something that we are currently, actively exploring for Relay. I’d encourage you to read the linked issue, but in short rather than containing errors with null bubbling, we contain errors with error boundaries. For cases where the user wants to imperatively handle the error case, they may add a @catch directive to the field which behaves very similarly to CCNs ?` and would hopefully some day be subsumed by it.

Note that compiler-based smart clients like Relay transform the queries/fragments defined by the user before sending them to the server. This means Relay can auto-insert ?s on all non-nullable fields, ensuring resiliency is the default behavior and we will always render as much of the UI as possible, given the data that the server was able to send.

So, Relay would use ? in two different was:

  1. As a hidden implementation detail used to ask the server to not apply null bubbling
  2. As a user-facing feature to allow components to locally handle errors instead of relying on error boundaries

A pattern not a feature

One appealing aspect of this vision is that it’s simply composed from existing, or at least proposed, GraphQL spec primitives. It does not require any additional spec changes, and can be optionally adopted by those who find it a good tradeoff.


Appendix/Caveats

This solution is not a silver bullet. It may not be viable for other clients, and even for Relay there are significant challenges that would need to be solved first. I propose it here more as a long-term vision than as an immediate next step. Here are list of concerns/caveats/complicating factors:

Missing Data

In Relay, there are actually two reasons that we type all fields as optional:

  1. The field might return null due to error
  2. The field might be missing due to normalization

To make Relay fields non-nullable by default, we’ll need to first provide a mechanism for Relay to handle refetching (or erroring) in the face of missing data. I believe missing data is a fundamental gap in Relay today and is deserving of a project to resolve that gap.

The power of defaults

Shifting responsibility from the server to the client makes it harder to enforce this best practice of resiliency. Opinionated smart client frameworks may be able take over the role of enforcing resiliency by auto-inserting ?s, but the story for simplistic clients is less clear.

Users will instinctively take the path of least resistance. If adding resiliency is extra work that is not forced upon them by the server or a client framework, it is likely that client code will tend not to go the extra mile to handle potential errors.

Error boundaries

This approach is dependent upon having a client architecture that allows product code to contain errors thrown during render. React Error Boundaries provide this primitive, but client architectures without such a feature may not have a clear path to adding explicit error handling, which is a necessary ingredient for this approach to work.

Even in Relay, explicit error handling has not yet been validated, though we hope to ship it to production soon.

Breaking changes

Another reason that GraphQL recommends that all fields be nullable, even if their current implementation is non-nullable, is that it allows us to turn a non-nullable field into a nullable field as a non-breaking change. This is especially important on mobile where clients live essentially forever. Being able to make a field nullable can be key to being able to delete code.

I don’t have a solution to this problem, but I am curious to learn how well it works in practice. Have users of this approach actually be able to routinely make fields nullable without breaking old clients? Are product engineers really designing apps that gracefully degrade in the face of any field being null? The convergent evolution of @required and CCN’s ! makes me wonder.

Worst case, the approach I outline here would only be viable for clients with a finite support window.

Alternatives to CCN

Our use of CCN to enable this new model, is more opportunistic than designed. CCN offers primitives that smart clients can leverage behind the scenes as a compiler implementation detail. The core behavior we really want is:

  1. A schema that exposes the true nullability of fields, at the same time as…
  2. An execution model that performs no null bubbling

This works because we can expect the smart client to intercept error fields before they reach product code, shielding it from nulls in non-nullable locations.

If we think this model is broadly valuable, it’s possible we would want to explore a more explicit mechanism to enable this execution model rather than simply allowing smart clients to fake this execution model via compiler-inserted CCN annotations.

[2023-09-26] Update Interrobang RFC... to remove the interrobang 😉

Changes:

  1. In the SDL, we add a question mark, this has the behavior of the field being expressly nullable
  2. Anything that does not have a question mark is not supposed to be nullable ie if there is no value, then there MUST be an error in the errors array.
  3. Introspection part: A flag that would indicate that a client is aware of this behavior and gets introspection that’s aware of this nullability

Note: Action Item issues are reviewed and closed during Working Group
meetings.

CCN’s ? can save normalized caches from null bubbling corruption

TL;DR: I believe smart clients like Relay, which maintain a normalized cache, could leverage CCN’s ? to prevent cache corruption due to null bubbling

The Problem

Null bubbling (when a error or null value encountered in a non-nullable field bubbles up to a parent nullable field) is destructive. It causes true and correct data that could be included in the response to be omitted. This can cause problems when trying to write a GraphQL response into a normalized cache.

Imagine two cards A and B side by side, each of which contain information about the current user. A shows “name” and B shows “age”. They are powered by two different queries. The response for query A looks like this:

{ // Response A
  "data": {
    "me": { "id": "10", "name": "Jordan}
  }
}

We can write this into our normalized cache and it will look like:

{ // Normalized cache
  "ROOT": {
    "me": {__id: "10"}
  },
  "10": {
    "id": "10",
    "name": "Jordan"
  }
}

This is fine, and we can nicely render our first card.

Now, however, we get an error in our second query’s response:

{ // Response B
  "data": {
    "me": null // <-- Null due to bubbling
  }
  "errors": [{ "message": "Age errored", "path": ["me", "age"] }]
}

The age field, which was non-nullable, has errored, and we won’t be able to render our second card. Oh well, such is life. However, when we go to write this into our cache, things get worse:

{ // Normalized cache
  "ROOT": {
    "me": null // <-- We had to write null here!
  },
  "10": {
    "id": "10",
    "name": "Jordan"
  }
}

Now card A, which also reads from this normalized store, is broken despite the fact that none of the fields it reads are in an error state.

A possible solution

With the adoption of Relay’s proposed error handling feature, product code is shielded from implicit field errors via framework-level explicit error handling. This means Relay is no-longer dependent on the server’s response shape strictly matching the schema. Specifically, it’s fine for a field marked as an error to be missing, even if it’s non-nullable. In other words, Relay can safely opt-out of server-side null bubbling.

And CCN's ? offers Relay just that option. Relay’s compiler can annotate every non-null field in the query it generates with a ? to opt out of any null bubbling.

Note: I’ve documented the CCN behavior I’m expecting here.

A new world

Lets now imagine how response B would look like in this new world:

{ // Response B (without null bubbling)
  "data": {
    "me": {
      "id": "10",
      "age": null // <-- Errored, but did not bubble!
    }
  }
  "errors": [{
    "message": "Age errored",
    "path": ["me", "age"]
  }]
}

When we write this into our normalized cache we get:

{ // Normalized cache
  "ROOT": {
    "me": {__id: "10"}
  },
  "10": {
    "id": "10",
    "name": "Jordan"
    "age": Error("Age errored")
  }
}

Card B still can’t render, but note that we’ve prevented our normalized cache from getting corrupted due to null bubbling. Thanks CCN!

Related

This post is spiritually related to #19 in that they both explore the benefits of a mode of GraphQL execution that avoids null bubbling. It may be worth exploring other mechanisms where-by clients could opt out of null bubbling, but I wanted to point out that CCN’s ? is a powerful enough primitive to allow a smart client like Relay to opt out of null bubbling.


Hat tip to @RyanHoldren who wrote an excellent internal post articulating this normalization issue. My note here is very much inspired by that post.

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google ❤️ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.