async-graphql/src/resolver_utils/container.rs

309 lines
13 KiB
Rust
Raw Normal View History

2020-10-15 06:38:10 +00:00
use std::collections::BTreeMap;
use std::future::Future;
use std::pin::Pin;
2020-10-12 06:49:32 +00:00
use crate::extensions::{ErrorLogger, ExtensionContext, ResolveInfo};
use crate::parser::types::Selection;
2020-09-12 09:29:52 +00:00
use crate::registry::MetaType;
Rework errors This completely overhauls the error system used in async-graphql. - `Error` has been renamed to `ServerError` and `FieldError` has been renamed to just `Error`. This is because `FieldError` is by far the most common error that users will have to use so it makes sense to use the most obvious error name. Also, the current name didn't make sense as it was used for things other than field errors, such as the data callback for websockets. - `ServerError` has been made completely opaque. Before it was an enum of all the possible errors, but now it just contains an error message, the locations, the path and extensions. It is a shame that we lose information, it makes more sense as _conceptually_ GraphQL does not provide that information. It also frees us to change the internals of async-graphql a lot more. - The path of errors is no longer an opaque JSON value but a regular type, `Vec<PathSegment>`. The type duplication of `PathSegment` and `QueryPathSegment` is unfortunate, I plan to work on this in the future. - Now that `ServerError` is opaque, `RuleError` has been removed from the public API, making it simpler. - Additionally `QueryError` has been completely removed. Instead the error messages are constructed ad-hoc; I took care to never repeat an error message. - Instead of constructing field-not-found errors inside the implementations of field resolvers they now return `Option`s, where a `None` value is representative of the field not being found. - As an unfortunate consequence of the last change, self-referential types based on the output of a subscription resolver can no longer be created. This does not mean anything for users, but causes lifetime issues in the implementation of merged objects. I fixed it with a bit of a hack, but this'll have to be looked into further. - `InputValueError` now has a generic parameter - it's kind of weird but it's necessary for ergonomics. It also improves error messages. - The `ErrorExtensions` trait has been removed. I didn't think the `extend` method was necessary since `From` impls exist. But the ergonomics are still there with a new trait `ExtendError`, which is implemented for both errors and results. - `Response` now supports serializing multiple errors. This allows for nice things like having multiple validation errors not be awkwardly shoved into a single error. - When an error occurs in execution, data is sent as `null`. This is slightly more compliant with the spec but the algorithm described in <https://spec.graphql.org/June2018/#sec-Errors-and-Non-Nullability> has yet to be implemented.
2020-09-29 19:06:44 +00:00
use crate::{
Context, ContextSelectionSet, Name, OutputType, PathSegment, ServerError, ServerResult, Value,
Rework errors This completely overhauls the error system used in async-graphql. - `Error` has been renamed to `ServerError` and `FieldError` has been renamed to just `Error`. This is because `FieldError` is by far the most common error that users will have to use so it makes sense to use the most obvious error name. Also, the current name didn't make sense as it was used for things other than field errors, such as the data callback for websockets. - `ServerError` has been made completely opaque. Before it was an enum of all the possible errors, but now it just contains an error message, the locations, the path and extensions. It is a shame that we lose information, it makes more sense as _conceptually_ GraphQL does not provide that information. It also frees us to change the internals of async-graphql a lot more. - The path of errors is no longer an opaque JSON value but a regular type, `Vec<PathSegment>`. The type duplication of `PathSegment` and `QueryPathSegment` is unfortunate, I plan to work on this in the future. - Now that `ServerError` is opaque, `RuleError` has been removed from the public API, making it simpler. - Additionally `QueryError` has been completely removed. Instead the error messages are constructed ad-hoc; I took care to never repeat an error message. - Instead of constructing field-not-found errors inside the implementations of field resolvers they now return `Option`s, where a `None` value is representative of the field not being found. - As an unfortunate consequence of the last change, self-referential types based on the output of a subscription resolver can no longer be created. This does not mean anything for users, but causes lifetime issues in the implementation of merged objects. I fixed it with a bit of a hack, but this'll have to be looked into further. - `InputValueError` now has a generic parameter - it's kind of weird but it's necessary for ergonomics. It also improves error messages. - The `ErrorExtensions` trait has been removed. I didn't think the `extend` method was necessary since `From` impls exist. But the ergonomics are still there with a new trait `ExtendError`, which is implemented for both errors and results. - `Response` now supports serializing multiple errors. This allows for nice things like having multiple validation errors not be awkwardly shoved into a single error. - When an error occurs in execution, data is sent as `null`. This is slightly more compliant with the spec but the algorithm described in <https://spec.graphql.org/June2018/#sec-Errors-and-Non-Nullability> has yet to be implemented.
2020-09-29 19:06:44 +00:00
};
2020-09-12 09:29:52 +00:00
2020-10-13 02:19:30 +00:00
/// Represents a GraphQL container object.
2020-09-12 09:29:52 +00:00
///
2020-09-30 03:44:18 +00:00
/// This helper trait allows the type to call `resolve_container` on itself in its
/// `OutputType::resolve` implementation.
2020-09-12 09:29:52 +00:00
#[async_trait::async_trait]
pub trait ContainerType: OutputType {
/// This function returns true of type `EmptyMutation` only.
2020-09-12 09:29:52 +00:00
#[doc(hidden)]
fn is_empty() -> bool {
false
}
2020-10-10 02:32:43 +00:00
/// Resolves a field value and outputs it as a json value `async_graphql::Value`.
Rework errors This completely overhauls the error system used in async-graphql. - `Error` has been renamed to `ServerError` and `FieldError` has been renamed to just `Error`. This is because `FieldError` is by far the most common error that users will have to use so it makes sense to use the most obvious error name. Also, the current name didn't make sense as it was used for things other than field errors, such as the data callback for websockets. - `ServerError` has been made completely opaque. Before it was an enum of all the possible errors, but now it just contains an error message, the locations, the path and extensions. It is a shame that we lose information, it makes more sense as _conceptually_ GraphQL does not provide that information. It also frees us to change the internals of async-graphql a lot more. - The path of errors is no longer an opaque JSON value but a regular type, `Vec<PathSegment>`. The type duplication of `PathSegment` and `QueryPathSegment` is unfortunate, I plan to work on this in the future. - Now that `ServerError` is opaque, `RuleError` has been removed from the public API, making it simpler. - Additionally `QueryError` has been completely removed. Instead the error messages are constructed ad-hoc; I took care to never repeat an error message. - Instead of constructing field-not-found errors inside the implementations of field resolvers they now return `Option`s, where a `None` value is representative of the field not being found. - As an unfortunate consequence of the last change, self-referential types based on the output of a subscription resolver can no longer be created. This does not mean anything for users, but causes lifetime issues in the implementation of merged objects. I fixed it with a bit of a hack, but this'll have to be looked into further. - `InputValueError` now has a generic parameter - it's kind of weird but it's necessary for ergonomics. It also improves error messages. - The `ErrorExtensions` trait has been removed. I didn't think the `extend` method was necessary since `From` impls exist. But the ergonomics are still there with a new trait `ExtendError`, which is implemented for both errors and results. - `Response` now supports serializing multiple errors. This allows for nice things like having multiple validation errors not be awkwardly shoved into a single error. - When an error occurs in execution, data is sent as `null`. This is slightly more compliant with the spec but the algorithm described in <https://spec.graphql.org/June2018/#sec-Errors-and-Non-Nullability> has yet to be implemented.
2020-09-29 19:06:44 +00:00
///
/// If the field was not found returns None.
2020-10-10 02:32:43 +00:00
async fn resolve_field(&self, ctx: &Context<'_>) -> ServerResult<Option<Value>>;
2020-09-12 09:29:52 +00:00
2020-09-29 23:45:48 +00:00
/// Collect all the fields of the container that are queried in the selection set.
2020-09-12 09:29:52 +00:00
///
/// Objects do not have to override this, but interfaces and unions must call it on their
/// internal type.
fn collect_all_fields<'a>(
&'a self,
ctx: &ContextSelectionSet<'a>,
fields: &mut Fields<'a>,
Rework errors This completely overhauls the error system used in async-graphql. - `Error` has been renamed to `ServerError` and `FieldError` has been renamed to just `Error`. This is because `FieldError` is by far the most common error that users will have to use so it makes sense to use the most obvious error name. Also, the current name didn't make sense as it was used for things other than field errors, such as the data callback for websockets. - `ServerError` has been made completely opaque. Before it was an enum of all the possible errors, but now it just contains an error message, the locations, the path and extensions. It is a shame that we lose information, it makes more sense as _conceptually_ GraphQL does not provide that information. It also frees us to change the internals of async-graphql a lot more. - The path of errors is no longer an opaque JSON value but a regular type, `Vec<PathSegment>`. The type duplication of `PathSegment` and `QueryPathSegment` is unfortunate, I plan to work on this in the future. - Now that `ServerError` is opaque, `RuleError` has been removed from the public API, making it simpler. - Additionally `QueryError` has been completely removed. Instead the error messages are constructed ad-hoc; I took care to never repeat an error message. - Instead of constructing field-not-found errors inside the implementations of field resolvers they now return `Option`s, where a `None` value is representative of the field not being found. - As an unfortunate consequence of the last change, self-referential types based on the output of a subscription resolver can no longer be created. This does not mean anything for users, but causes lifetime issues in the implementation of merged objects. I fixed it with a bit of a hack, but this'll have to be looked into further. - `InputValueError` now has a generic parameter - it's kind of weird but it's necessary for ergonomics. It also improves error messages. - The `ErrorExtensions` trait has been removed. I didn't think the `extend` method was necessary since `From` impls exist. But the ergonomics are still there with a new trait `ExtendError`, which is implemented for both errors and results. - `Response` now supports serializing multiple errors. This allows for nice things like having multiple validation errors not be awkwardly shoved into a single error. - When an error occurs in execution, data is sent as `null`. This is slightly more compliant with the spec but the algorithm described in <https://spec.graphql.org/June2018/#sec-Errors-and-Non-Nullability> has yet to be implemented.
2020-09-29 19:06:44 +00:00
) -> ServerResult<()>
2020-09-12 09:29:52 +00:00
where
Self: Sized + Send + Sync,
{
fields.add_set(ctx, self)
}
/// Find the GraphQL entity with the given name from the parameter.
///
/// Objects should override this in case they are the query root.
2020-10-10 02:32:43 +00:00
async fn find_entity(&self, _: &Context<'_>, _params: &Value) -> ServerResult<Option<Value>> {
Rework errors This completely overhauls the error system used in async-graphql. - `Error` has been renamed to `ServerError` and `FieldError` has been renamed to just `Error`. This is because `FieldError` is by far the most common error that users will have to use so it makes sense to use the most obvious error name. Also, the current name didn't make sense as it was used for things other than field errors, such as the data callback for websockets. - `ServerError` has been made completely opaque. Before it was an enum of all the possible errors, but now it just contains an error message, the locations, the path and extensions. It is a shame that we lose information, it makes more sense as _conceptually_ GraphQL does not provide that information. It also frees us to change the internals of async-graphql a lot more. - The path of errors is no longer an opaque JSON value but a regular type, `Vec<PathSegment>`. The type duplication of `PathSegment` and `QueryPathSegment` is unfortunate, I plan to work on this in the future. - Now that `ServerError` is opaque, `RuleError` has been removed from the public API, making it simpler. - Additionally `QueryError` has been completely removed. Instead the error messages are constructed ad-hoc; I took care to never repeat an error message. - Instead of constructing field-not-found errors inside the implementations of field resolvers they now return `Option`s, where a `None` value is representative of the field not being found. - As an unfortunate consequence of the last change, self-referential types based on the output of a subscription resolver can no longer be created. This does not mean anything for users, but causes lifetime issues in the implementation of merged objects. I fixed it with a bit of a hack, but this'll have to be looked into further. - `InputValueError` now has a generic parameter - it's kind of weird but it's necessary for ergonomics. It also improves error messages. - The `ErrorExtensions` trait has been removed. I didn't think the `extend` method was necessary since `From` impls exist. But the ergonomics are still there with a new trait `ExtendError`, which is implemented for both errors and results. - `Response` now supports serializing multiple errors. This allows for nice things like having multiple validation errors not be awkwardly shoved into a single error. - When an error occurs in execution, data is sent as `null`. This is slightly more compliant with the spec but the algorithm described in <https://spec.graphql.org/June2018/#sec-Errors-and-Non-Nullability> has yet to be implemented.
2020-09-29 19:06:44 +00:00
Ok(None)
2020-09-12 09:29:52 +00:00
}
}
#[async_trait::async_trait]
2020-09-29 23:45:48 +00:00
impl<T: ContainerType + Send + Sync> ContainerType for &T {
2020-10-10 02:32:43 +00:00
async fn resolve_field(&self, ctx: &Context<'_>) -> ServerResult<Option<Value>> {
2020-09-12 09:29:52 +00:00
T::resolve_field(*self, ctx).await
}
async fn find_entity(&self, ctx: &Context<'_>, params: &Value) -> ServerResult<Option<Value>> {
T::find_entity(*self, ctx, params).await
}
2020-09-12 09:29:52 +00:00
}
2020-09-29 23:45:48 +00:00
/// Resolve an container by executing each of the fields concurrently.
pub async fn resolve_container<'a, T: ContainerType + Send + Sync>(
2020-09-12 09:29:52 +00:00
ctx: &ContextSelectionSet<'a>,
root: &'a T,
2020-10-10 02:32:43 +00:00
) -> ServerResult<Value> {
resolve_container_inner(ctx, root, true).await
2020-09-12 09:29:52 +00:00
}
2020-09-29 23:45:48 +00:00
/// Resolve an container by executing each of the fields serially.
pub async fn resolve_container_serial<'a, T: ContainerType + Send + Sync>(
2020-09-12 09:29:52 +00:00
ctx: &ContextSelectionSet<'a>,
root: &'a T,
) -> ServerResult<Value> {
resolve_container_inner(ctx, root, false).await
}
fn insert_value(target: &mut BTreeMap<Name, Value>, name: Name, value: Value) {
if let Some(prev_value) = target.get_mut(&name) {
if let Value::Object(target_map) = prev_value {
if let Value::Object(obj) = value {
for (key, value) in obj.into_iter() {
insert_value(target_map, key, value);
}
}
} else if let Value::List(target_list) = prev_value {
if let Value::List(list) = value {
for (idx, value) in list.into_iter().enumerate() {
if let Some(Value::Object(target_map)) = target_list.get_mut(idx) {
if let Value::Object(obj) = value {
for (key, value) in obj.into_iter() {
insert_value(target_map, key, value);
}
}
}
}
}
}
} else {
target.insert(name, value);
}
}
async fn resolve_container_inner<'a, T: ContainerType + Send + Sync>(
ctx: &ContextSelectionSet<'a>,
root: &'a T,
parallel: bool,
2020-10-10 02:32:43 +00:00
) -> ServerResult<Value> {
2020-09-12 09:29:52 +00:00
let mut fields = Fields(Vec::new());
fields.add_set(ctx, root)?;
let res = if parallel {
2020-10-16 06:49:22 +00:00
futures_util::future::try_join_all(fields.0).await?
} else {
let mut results = Vec::with_capacity(fields.0.len());
for field in fields.0 {
results.push(field.await?);
}
results
};
2020-09-12 09:29:52 +00:00
let mut map = BTreeMap::new();
for (name, value) in res {
insert_value(&mut map, name, value);
2020-09-12 09:29:52 +00:00
}
2020-10-10 02:32:43 +00:00
Ok(Value::Object(map))
2020-09-12 09:29:52 +00:00
}
2020-10-10 02:32:43 +00:00
type BoxFieldFuture<'a> = Pin<Box<dyn Future<Output = ServerResult<(Name, Value)>> + 'a + Send>>;
2020-09-12 09:29:52 +00:00
2020-09-30 03:44:18 +00:00
/// A set of fields on an container that are being selected.
2020-09-12 09:29:52 +00:00
pub struct Fields<'a>(Vec<BoxFieldFuture<'a>>);
impl<'a> Fields<'a> {
2020-09-30 03:44:18 +00:00
/// Add another set of fields to this set of fields using the given container.
2020-09-29 23:45:48 +00:00
pub fn add_set<T: ContainerType + Send + Sync>(
2020-09-12 09:29:52 +00:00
&mut self,
ctx: &ContextSelectionSet<'a>,
root: &'a T,
Rework errors This completely overhauls the error system used in async-graphql. - `Error` has been renamed to `ServerError` and `FieldError` has been renamed to just `Error`. This is because `FieldError` is by far the most common error that users will have to use so it makes sense to use the most obvious error name. Also, the current name didn't make sense as it was used for things other than field errors, such as the data callback for websockets. - `ServerError` has been made completely opaque. Before it was an enum of all the possible errors, but now it just contains an error message, the locations, the path and extensions. It is a shame that we lose information, it makes more sense as _conceptually_ GraphQL does not provide that information. It also frees us to change the internals of async-graphql a lot more. - The path of errors is no longer an opaque JSON value but a regular type, `Vec<PathSegment>`. The type duplication of `PathSegment` and `QueryPathSegment` is unfortunate, I plan to work on this in the future. - Now that `ServerError` is opaque, `RuleError` has been removed from the public API, making it simpler. - Additionally `QueryError` has been completely removed. Instead the error messages are constructed ad-hoc; I took care to never repeat an error message. - Instead of constructing field-not-found errors inside the implementations of field resolvers they now return `Option`s, where a `None` value is representative of the field not being found. - As an unfortunate consequence of the last change, self-referential types based on the output of a subscription resolver can no longer be created. This does not mean anything for users, but causes lifetime issues in the implementation of merged objects. I fixed it with a bit of a hack, but this'll have to be looked into further. - `InputValueError` now has a generic parameter - it's kind of weird but it's necessary for ergonomics. It also improves error messages. - The `ErrorExtensions` trait has been removed. I didn't think the `extend` method was necessary since `From` impls exist. But the ergonomics are still there with a new trait `ExtendError`, which is implemented for both errors and results. - `Response` now supports serializing multiple errors. This allows for nice things like having multiple validation errors not be awkwardly shoved into a single error. - When an error occurs in execution, data is sent as `null`. This is slightly more compliant with the spec but the algorithm described in <https://spec.graphql.org/June2018/#sec-Errors-and-Non-Nullability> has yet to be implemented.
2020-09-29 19:06:44 +00:00
) -> ServerResult<()> {
2020-09-12 09:29:52 +00:00
for selection in &ctx.item.node.items {
if ctx.is_skip(&selection.node.directives())? {
continue;
}
match &selection.node {
Selection::Field(field) => {
if field.node.name.node == "__typename" {
// Get the typename
let ctx_field = ctx.with_field(field);
2020-10-10 02:32:43 +00:00
let field_name = ctx_field.item.node.response_key().node.clone();
2020-09-12 09:29:52 +00:00
let typename = root.introspection_type_name().into_owned();
self.0.push(Box::pin(async move {
2020-10-10 02:32:43 +00:00
Ok((field_name, Value::String(typename)))
2020-09-12 09:29:52 +00:00
}));
continue;
}
if ctx.is_ifdef(&field.node.directives) {
if let Some(MetaType::Object { fields, .. }) =
ctx.schema_env.registry.types.get(T::type_name().as_ref())
{
if !fields.contains_key(field.node.name.node.as_str()) {
continue;
}
}
}
self.0.push(Box::pin({
// TODO: investigate removing this
2020-09-12 09:29:52 +00:00
let ctx = ctx.clone();
async move {
let ctx_field = ctx.with_field(field);
2020-10-10 02:32:43 +00:00
let field_name = ctx_field.item.node.response_key().node.clone();
2020-09-12 09:29:52 +00:00
2020-10-24 00:59:35 +00:00
let res = if ctx_field.query_env.extensions.is_empty() {
match root.resolve_field(&ctx_field).await {
Ok(value) => Ok((field_name, value.unwrap_or_default())),
Err(e) => {
Err(e.path(PathSegment::Field(field_name.to_string())))
}
}?
} else {
2020-10-24 00:59:35 +00:00
let ctx_extension = ExtensionContext {
schema_data: &ctx.schema_env.data,
query_data: &ctx.query_env.ctx_data,
};
let type_name = T::type_name();
let resolve_info = ResolveInfo {
resolve_id: ctx_field.resolve_id,
path_node: ctx_field.path_node.as_ref().unwrap(),
parent_type: &type_name,
return_type: match ctx_field
.schema_env
.registry
.types
.get(type_name.as_ref())
.and_then(|ty| {
ty.field_by_name(field.node.name.node.as_str())
})
.map(|field| &field.ty)
{
Some(ty) => &ty,
None => {
return Err(ServerError::new(format!(
r#"Cannot query field "{}" on type "{}"."#,
field_name, type_name
))
.at(ctx_field.item.pos)
.path(PathSegment::Field(field_name.to_string())));
}
},
};
ctx_field
.query_env
.extensions
.resolve_start(&ctx_extension, &resolve_info);
let res = match root.resolve_field(&ctx_field).await {
Ok(value) => Ok((field_name, value.unwrap_or_default())),
Err(e) => {
Err(e.path(PathSegment::Field(field_name.to_string())))
}
}
.log_error(&ctx_extension, &ctx_field.query_env.extensions)?;
ctx_field
.query_env
.extensions
.resolve_end(&ctx_extension, &resolve_info);
res
2020-10-23 16:18:37 +00:00
};
2020-09-12 09:29:52 +00:00
2020-10-23 16:18:37 +00:00
Ok(res)
2020-09-12 09:29:52 +00:00
}
}));
}
selection => {
let (type_condition, selection_set) = match selection {
Selection::Field(_) => unreachable!(),
Selection::FragmentSpread(spread) => {
let fragment =
ctx.query_env.fragments.get(&spread.node.fragment_name.node);
2020-09-12 09:29:52 +00:00
let fragment = match fragment {
Some(fragment) => fragment,
None => {
Rework errors This completely overhauls the error system used in async-graphql. - `Error` has been renamed to `ServerError` and `FieldError` has been renamed to just `Error`. This is because `FieldError` is by far the most common error that users will have to use so it makes sense to use the most obvious error name. Also, the current name didn't make sense as it was used for things other than field errors, such as the data callback for websockets. - `ServerError` has been made completely opaque. Before it was an enum of all the possible errors, but now it just contains an error message, the locations, the path and extensions. It is a shame that we lose information, it makes more sense as _conceptually_ GraphQL does not provide that information. It also frees us to change the internals of async-graphql a lot more. - The path of errors is no longer an opaque JSON value but a regular type, `Vec<PathSegment>`. The type duplication of `PathSegment` and `QueryPathSegment` is unfortunate, I plan to work on this in the future. - Now that `ServerError` is opaque, `RuleError` has been removed from the public API, making it simpler. - Additionally `QueryError` has been completely removed. Instead the error messages are constructed ad-hoc; I took care to never repeat an error message. - Instead of constructing field-not-found errors inside the implementations of field resolvers they now return `Option`s, where a `None` value is representative of the field not being found. - As an unfortunate consequence of the last change, self-referential types based on the output of a subscription resolver can no longer be created. This does not mean anything for users, but causes lifetime issues in the implementation of merged objects. I fixed it with a bit of a hack, but this'll have to be looked into further. - `InputValueError` now has a generic parameter - it's kind of weird but it's necessary for ergonomics. It also improves error messages. - The `ErrorExtensions` trait has been removed. I didn't think the `extend` method was necessary since `From` impls exist. But the ergonomics are still there with a new trait `ExtendError`, which is implemented for both errors and results. - `Response` now supports serializing multiple errors. This allows for nice things like having multiple validation errors not be awkwardly shoved into a single error. - When an error occurs in execution, data is sent as `null`. This is slightly more compliant with the spec but the algorithm described in <https://spec.graphql.org/June2018/#sec-Errors-and-Non-Nullability> has yet to be implemented.
2020-09-29 19:06:44 +00:00
return Err(ServerError::new(format!(
r#"Unknown fragment "{}"."#,
spread.node.fragment_name.node
))
.at(spread.pos));
2020-09-12 09:29:52 +00:00
}
};
(
Some(&fragment.node.type_condition),
&fragment.node.selection_set,
)
}
Selection::InlineFragment(fragment) => (
fragment.node.type_condition.as_ref(),
&fragment.node.selection_set,
),
};
let type_condition =
type_condition.map(|condition| condition.node.on.node.as_str());
let introspection_type_name = root.introspection_type_name();
2020-09-12 16:07:46 +00:00
let applies_concrete_object = type_condition.map_or(false, |condition| {
2020-09-12 09:29:52 +00:00
introspection_type_name == condition
|| ctx
.schema_env
.registry
.implements
.get(&*introspection_type_name)
.map_or(false, |interfaces| interfaces.contains(condition))
2020-09-12 16:07:46 +00:00
});
if applies_concrete_object {
2020-09-12 09:29:52 +00:00
// The fragment applies to the concrete object type.
// TODO: This solution isn't ideal. If there are two interfaces InterfaceA
// and InterfaceB and one type MyObj that implements both, then if you have
// a type condition for `InterfaceA` on an `InterfaceB` and when resolving,
// the `InterfaceB` is actually a `MyObj` then the contents of the fragment
// will be treated as a `MyObj` rather than an `InterfaceB`. Example:
//
// myObjAsInterfaceB {
// ... on InterfaceA {
// # here you can query MyObj fields even when you should only be
// # able to query InterfaceA fields.
// }
// }
root.collect_all_fields(&ctx.with_selection_set(selection_set), self)?;
} else if type_condition.map_or(true, |condition| T::type_name() == condition) {
// The fragment applies to an interface type.
self.add_set(&ctx.with_selection_set(selection_set), root)?;
}
}
}
}
Ok(())
}
}