graphql-engine/server/src-lib/Hasura/GraphQL/Schema/RemoteRelationship.hs

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

212 lines
10 KiB
Haskell
Raw Normal View History

module Hasura.GraphQL.Schema.RemoteRelationship
( remoteRelationshipField,
)
where
import Control.Lens
import Data.Has
import Data.HashMap.Strict.Extended qualified as Map
import Data.List.NonEmpty qualified as NE
import Data.Text.Extended
import Hasura.Base.Error
import Hasura.GraphQL.Parser
import Hasura.GraphQL.Parser qualified as P
import Hasura.GraphQL.Parser.Constants qualified as G
import Hasura.GraphQL.Parser.Internal.Parser qualified as P
import Hasura.GraphQL.Schema.Backend
import Hasura.GraphQL.Schema.Common
import Hasura.GraphQL.Schema.Instances ()
import Hasura.GraphQL.Schema.Remote
import Hasura.GraphQL.Schema.Select
import Hasura.GraphQL.Schema.Table
import Hasura.Prelude
import Hasura.RQL.DDL.RemoteRelationship.Validate
import Hasura.RQL.IR qualified as IR
import Hasura.RQL.Types.Common (FieldName, RelType (..), relNameToTxt)
import Hasura.RQL.Types.Relationships.Remote
import Hasura.RQL.Types.Relationships.ToSchema
import Hasura.RQL.Types.Relationships.ToSchema qualified as Remote
import Hasura.RQL.Types.RemoteSchema
import Hasura.RQL.Types.ResultCustomization
import Hasura.RQL.Types.SchemaCache hiding (askTableInfo)
2022-05-27 20:21:22 +03:00
import Hasura.RQL.Types.Source
import Hasura.RQL.Types.SourceCustomization (NamingCase (..), mkCustomizedTypename)
import Hasura.SQL.AnyBackend
import Hasura.Session
import Language.GraphQL.Draft.Syntax qualified as G
-- | Remote relationship field parsers
remoteRelationshipField ::
2022-05-27 20:21:22 +03:00
SourceCache ->
RemoteSchemaMap ->
RemoteRelationshipParserBuilder
remoteRelationshipField sourceCache remoteSchemaCache = RemoteRelationshipParserBuilder
\RemoteFieldInfo {..} -> runMaybeT do
Clean Relay's code, break schema cycles, introduce Node ID V2 ## Motivation This PR rewrites most of Relay to achieve the following: - ~~fix a bug in which the same node id could refer to two different tables in the schema~~ - remove one of the few remaining uses of the source cache in the schema building code In doing so, it also: - simplifies the `BackendSchema` class by removing `node` from it, - makes it much easier for other backends to support Relay, - documents, re-organizes, and clarifies the code. ## Description This PR introduces a new `NodeId` version ~~, and adapts the Postgres code to always generate this V2 version~~. This new id contains the source name, in addition to the table name, in order to disambiguate similar table names across different sources (which is now possible with source customization). In doing so, it now explicitly handles that case for V1 node ids, and returns an explicit error message instead of running the risk of _silently returning the wrong information_. Furthermore, it adapts `nodeField` to support multiple backends; most of the code was trivial to generalize, and as a result it lowers the cost of entry for other backends, that now only need to support `AFNodeId` in their translation layer. Finally, it removes one more cycle in the schema building code, by using the same trick we used for remote relationships instead of using the memoization trick of #4576. ## Remaining work - ~~[ ]write a Changelog entry~~ - ~~[x] adapt all tests that were asserting on an old node id~~ ## Future work This PR was adapted from its original form to avoid a breaking change: while it introduces a Node ID V2, we keep generating V1 IDs and the parser rejects V2 IDs. It will be easy to make the switch at a later data in a subsequent PR. PR-URL: https://github.com/hasura/graphql-engine-mono/pull/4593 GitOrigin-RevId: 88e5cb91e8b0646900547fa8c7c0e1463de267a1
2022-06-07 16:35:26 +03:00
queryType <- retrieve scSchemaKind
2022-05-27 20:21:22 +03:00
-- https://github.com/hasura/graphql-engine/issues/5144
-- The above issue is easily fixable by removing the following guard
Clean Relay's code, break schema cycles, introduce Node ID V2 ## Motivation This PR rewrites most of Relay to achieve the following: - ~~fix a bug in which the same node id could refer to two different tables in the schema~~ - remove one of the few remaining uses of the source cache in the schema building code In doing so, it also: - simplifies the `BackendSchema` class by removing `node` from it, - makes it much easier for other backends to support Relay, - documents, re-organizes, and clarifies the code. ## Description This PR introduces a new `NodeId` version ~~, and adapts the Postgres code to always generate this V2 version~~. This new id contains the source name, in addition to the table name, in order to disambiguate similar table names across different sources (which is now possible with source customization). In doing so, it now explicitly handles that case for V1 node ids, and returns an explicit error message instead of running the risk of _silently returning the wrong information_. Furthermore, it adapts `nodeField` to support multiple backends; most of the code was trivial to generalize, and as a result it lowers the cost of entry for other backends, that now only need to support `AFNodeId` in their translation layer. Finally, it removes one more cycle in the schema building code, by using the same trick we used for remote relationships instead of using the memoization trick of #4576. ## Remaining work - ~~[ ]write a Changelog entry~~ - ~~[x] adapt all tests that were asserting on an old node id~~ ## Future work This PR was adapted from its original form to avoid a breaking change: while it introduces a Node ID V2, we keep generating V1 IDs and the parser rejects V2 IDs. It will be easy to make the switch at a later data in a subsequent PR. PR-URL: https://github.com/hasura/graphql-engine-mono/pull/4593 GitOrigin-RevId: 88e5cb91e8b0646900547fa8c7c0e1463de267a1
2022-06-07 16:35:26 +03:00
guard $ isHasuraSchema queryType
2022-05-27 20:21:22 +03:00
case _rfiRHS of
RFISource anyRemoteSourceFieldInfo ->
dispatchAnyBackend @BackendSchema anyRemoteSourceFieldInfo \remoteSourceFieldInfo -> do
fields <- lift $ remoteRelationshipToSourceField sourceCache remoteSourceFieldInfo
pure $ fmap (IR.RemoteSourceField . mkAnyBackend) <$> fields
RFISchema remoteSchema -> do
fields <- MaybeT $ remoteRelationshipToSchemaField remoteSchemaCache _rfiLHS remoteSchema
pure $ pure $ IR.RemoteSchemaField <$> fields
-- | Parser(s) for remote relationship fields to a remote schema
remoteRelationshipToSchemaField ::
forall r m n lhsJoinField.
(MonadBuildSchemaBase r m n) =>
2022-05-27 20:21:22 +03:00
RemoteSchemaMap ->
Map.HashMap FieldName lhsJoinField ->
RemoteSchemaFieldInfo ->
m (Maybe (FieldParser n (IR.RemoteSchemaSelect (IR.RemoteRelationshipField IR.UnpreparedValue))))
2022-05-27 20:21:22 +03:00
remoteRelationshipToSchemaField remoteSchemaCache lhsFields RemoteSchemaFieldInfo {..} = runMaybeT do
remoteSchemaPermsCtx <- retrieve soRemoteSchemaPermsCtx
roleName <- asks getter
remoteSchemaContext <-
Map.lookup _rrfiRemoteSchemaName remoteSchemaCache
`onNothing` throw500 ("invalid remote schema name: " <>> _rrfiRemoteSchemaName)
introspection <- hoistMaybe $ getIntrospectionResult remoteSchemaPermsCtx roleName remoteSchemaContext
let remoteSchemaRelationships = _rscRemoteRelationships remoteSchemaContext
roleIntrospection = irDoc introspection
remoteSchemaRoot = irQueryRoot introspection
remoteSchemaCustomizer = rsCustomizer $ _rscInfo remoteSchemaContext
RemoteSchemaIntrospection typeDefns = roleIntrospection
let hasuraFieldNames = Map.keysSet lhsFields
relationshipDef = ToSchemaRelationshipDef _rrfiRemoteSchemaName hasuraFieldNames _rrfiRemoteFields
(newInpValDefns :: [G.TypeDefinition [G.Name] RemoteSchemaInputValueDefinition], remoteFieldParamMap) <-
if roleName == adminRoleName
then do
-- we don't validate the remote relationship when the role is admin
-- because it's already been validated, when the remote relationship
-- was created
pure (_rrfiInputValueDefinitions, _rrfiParamMap)
else do
(_, roleRemoteField) <-
afold @(Either _) $
-- TODO: this really needs to go way, we shouldn't be doing
-- validation when building parsers
validateToSchemaRelationship relationshipDef _rrfiLHSIdentifier _rrfiName (_rrfiRemoteSchema, introspection) lhsFields
pure (Remote._rrfiInputValueDefinitions roleRemoteField, Remote._rrfiParamMap roleRemoteField)
let -- add the new input value definitions created by the remote relationship
-- to the existing schema introspection of the role
remoteRelationshipIntrospection = RemoteSchemaIntrospection $ typeDefns <> Map.fromListOn getTypeName newInpValDefns
fieldName <- textToName $ relNameToTxt _rrfiName
-- This selection set parser, should be of the remote node's selection set parser, which comes
-- from the fieldCall
let fieldCalls = unRemoteFields _rrfiRemoteFields
nestedFieldType <- lift $ lookupNestedFieldType remoteSchemaRoot roleIntrospection fieldCalls
let typeName = G.getBaseType nestedFieldType
fieldTypeDefinition <-
onNothing (lookupType roleIntrospection typeName)
-- the below case will never happen because we get the type name
-- from the schema document itself i.e. if a field exists for the
-- given role, then it's return type also must exist
$
throw500 $ "unexpected: " <> typeName <<> " not found "
-- These are the arguments that are given by the user while executing a query
let remoteFieldUserArguments = map snd $ Map.toList remoteFieldParamMap
remoteFld <-
withRemoteSchemaCustomization remoteSchemaCustomizer $
lift $
P.wrapFieldParser nestedFieldType
<$> remoteField remoteRelationshipIntrospection remoteSchemaRelationships remoteSchemaRoot fieldName Nothing remoteFieldUserArguments fieldTypeDefinition
pure $
remoteFld
`P.bindField` \fld@IR.GraphQLField {IR._fArguments = args, IR._fSelectionSet = selSet, IR._fName = fname} -> do
let remoteArgs =
Map.toList args <&> \(argName, argVal) -> IR.RemoteFieldArgument argName $ P.GraphQLValue argVal
let resultCustomizer =
applyFieldCalls fieldCalls $
applyAliasMapping (singletonAliasMapping fname (fcName $ NE.last fieldCalls)) $
makeResultCustomizer remoteSchemaCustomizer fld
pure $
IR.RemoteSchemaSelect
{ IR._rselArgs = remoteArgs,
IR._rselResultCustomizer = resultCustomizer,
IR._rselSelection = selSet,
IR._rselFieldCall = fieldCalls,
IR._rselRemoteSchema = _rrfiRemoteSchema
}
where
-- Apply parent field calls so that the result customizer modifies the nested field
applyFieldCalls :: NonEmpty FieldCall -> ResultCustomizer -> ResultCustomizer
applyFieldCalls fieldCalls resultCustomizer =
foldr (modifyFieldByName . fcName) resultCustomizer $ NE.init fieldCalls
lookupNestedFieldType' ::
(MonadSchema n m, MonadError QErr m) =>
G.Name ->
RemoteSchemaIntrospection ->
FieldCall ->
m G.GType
lookupNestedFieldType' parentTypeName remoteSchemaIntrospection (FieldCall fcName _) =
case lookupObject remoteSchemaIntrospection parentTypeName of
Nothing -> throw400 RemoteSchemaError $ "object with name " <> parentTypeName <<> " not found"
Just G.ObjectTypeDefinition {..} ->
case find ((== fcName) . G._fldName) _otdFieldsDefinition of
Nothing -> throw400 RemoteSchemaError $ "field with name " <> fcName <<> " not found"
Just G.FieldDefinition {..} -> pure _fldType
lookupNestedFieldType ::
(MonadSchema n m, MonadError QErr m) =>
G.Name ->
RemoteSchemaIntrospection ->
NonEmpty FieldCall ->
m G.GType
lookupNestedFieldType parentTypeName remoteSchemaIntrospection (fieldCall :| rest) = do
fieldType <- lookupNestedFieldType' parentTypeName remoteSchemaIntrospection fieldCall
case NE.nonEmpty rest of
Nothing -> pure fieldType
Just rest' -> do
lookupNestedFieldType (G.getBaseType fieldType) remoteSchemaIntrospection rest'
-- | Parser(s) for remote relationship fields to a database table.
-- Note that when the target is a database table, an array relationship
-- declaration would have the '_aggregate' field in addition to the array
-- relationship field, hence [FieldParser ...] instead of 'FieldParser'
remoteRelationshipToSourceField ::
forall r m n tgt.
(MonadBuildSchemaBase r m n, BackendSchema tgt) =>
2022-05-27 20:21:22 +03:00
SourceCache ->
RemoteSourceFieldInfo tgt ->
m [FieldParser n (IR.RemoteSourceSelect (IR.RemoteRelationshipField IR.UnpreparedValue) IR.UnpreparedValue tgt)]
2022-05-27 20:21:22 +03:00
remoteRelationshipToSourceField sourceCache RemoteSourceFieldInfo {..} =
withTypenameCustomization (mkCustomizedTypename (Just _rsfiSourceCustomization) HasuraCase) do
2022-05-27 20:21:22 +03:00
sourceInfo <-
onNothing (unsafeSourceInfo @tgt =<< Map.lookup _rsfiSource sourceCache) $
throw500 $ "source not found " <> dquote _rsfiSource
tableInfo <- askTableInfo sourceInfo _rsfiTable
fieldName <- textToName $ relNameToTxt _rsfiName
maybePerms <- tableSelectPermissions @tgt tableInfo
case maybePerms of
Nothing -> pure []
Just tablePerms -> do
parsers <- case _rsfiType of
ObjRel -> do
2022-05-27 20:21:22 +03:00
selectionSetParserM <- tableSelectionSet sourceInfo tableInfo
Role-invariant schema constructors We build the GraphQL schema by combining building blocks such as `tableSelectionSet` and `columnParser`. These building blocks individually build `{InputFields,Field,}Parser` objects. Those object specify the valid GraphQL schema. Since the GraphQL schema is role-dependent, at some point we need to know what fragment of the GraphQL schema a specific role is allowed to access, and this is stored in `{Sel,Upd,Ins,Del}PermInfo` objects. We have passed around these permission objects as function arguments to the schema building blocks since we first started dealing with permissions during the PDV refactor - see hasura/graphql-engine@5168b99e463199b1934d8645bd6cd37eddb64ae1 in hasura/graphql-engine#4111. This means that, for instance, `tableSelectionSet` has as its type: ```haskell tableSelectionSet :: forall b r m n. MonadBuildSchema b r m n => SourceName -> TableInfo b -> SelPermInfo b -> m (Parser 'Output n (AnnotatedFields b)) ``` There are three reasons to change this. 1. We often pass a `Maybe (xPermInfo b)` instead of a proper `xPermInfo b`, and it's not clear what the intended semantics of this is. Some potential improvements on the data types involved are discussed in issue hasura/graphql-engine-mono#3125. 2. In most cases we also already pass a `TableInfo b`, and together with the `MonadRole` that is usually also in scope, this means that we could look up the required permissions regardless: so passing the permissions explicitly undermines the "single source of truth" principle. Breaking this principle also makes the code more difficult to read. 3. We are working towards role-based parsers (see hasura/graphql-engine-mono#2711), where the `{InputFields,Field,}Parser` objects are constructed in a role-invariant way, so that we have a single object that can be used for all roles. In particular, this means that the schema building blocks _need_ to be constructed in a role-invariant way. While this PR doesn't accomplish that, it does reduce the amount of role-specific arguments being passed, thus fixing hasura/graphql-engine-mono#3068. Concretely, this PR simply drops the `xPermInfo b` argument from almost all schema building blocks. Instead these objects are looked up from the `TableInfo b` as-needed. The resulting code is considerably simpler and shorter. One way to interpret this change is as follows. Before this PR, we figured out permissions at the top-level in `Hasura.GraphQL.Schema`, passing down the obtained `xPermInfo` objects as required. After this PR, we have a bottom-up approach where the schema building blocks themselves decide whether they want to be included for a particular role. So this moves some permission logic out of `Hasura.GraphQL.Schema`, which is very complex. PR-URL: https://github.com/hasura/graphql-engine-mono/pull/3608 GitOrigin-RevId: 51a744f34ec7d57bc8077667ae7f9cb9c4f6c962
2022-02-17 11:16:20 +03:00
pure $ case selectionSetParserM of
Nothing -> []
Just selectionSetParser ->
pure $
subselection_ fieldName Nothing selectionSetParser <&> \fields ->
IR.SourceRelationshipObject $
IR.AnnObjectSelectG fields _rsfiTable $ IR._tpFilter $ tablePermissionsInfo tablePerms
ArrRel -> do
let aggFieldName = fieldName <> G.__aggregate
2022-05-27 20:21:22 +03:00
selectionSetParser <- selectTable sourceInfo tableInfo fieldName Nothing
aggSelectionSetParser <- selectTableAggregate sourceInfo tableInfo aggFieldName Nothing
pure $
catMaybes
Role-invariant schema constructors We build the GraphQL schema by combining building blocks such as `tableSelectionSet` and `columnParser`. These building blocks individually build `{InputFields,Field,}Parser` objects. Those object specify the valid GraphQL schema. Since the GraphQL schema is role-dependent, at some point we need to know what fragment of the GraphQL schema a specific role is allowed to access, and this is stored in `{Sel,Upd,Ins,Del}PermInfo` objects. We have passed around these permission objects as function arguments to the schema building blocks since we first started dealing with permissions during the PDV refactor - see hasura/graphql-engine@5168b99e463199b1934d8645bd6cd37eddb64ae1 in hasura/graphql-engine#4111. This means that, for instance, `tableSelectionSet` has as its type: ```haskell tableSelectionSet :: forall b r m n. MonadBuildSchema b r m n => SourceName -> TableInfo b -> SelPermInfo b -> m (Parser 'Output n (AnnotatedFields b)) ``` There are three reasons to change this. 1. We often pass a `Maybe (xPermInfo b)` instead of a proper `xPermInfo b`, and it's not clear what the intended semantics of this is. Some potential improvements on the data types involved are discussed in issue hasura/graphql-engine-mono#3125. 2. In most cases we also already pass a `TableInfo b`, and together with the `MonadRole` that is usually also in scope, this means that we could look up the required permissions regardless: so passing the permissions explicitly undermines the "single source of truth" principle. Breaking this principle also makes the code more difficult to read. 3. We are working towards role-based parsers (see hasura/graphql-engine-mono#2711), where the `{InputFields,Field,}Parser` objects are constructed in a role-invariant way, so that we have a single object that can be used for all roles. In particular, this means that the schema building blocks _need_ to be constructed in a role-invariant way. While this PR doesn't accomplish that, it does reduce the amount of role-specific arguments being passed, thus fixing hasura/graphql-engine-mono#3068. Concretely, this PR simply drops the `xPermInfo b` argument from almost all schema building blocks. Instead these objects are looked up from the `TableInfo b` as-needed. The resulting code is considerably simpler and shorter. One way to interpret this change is as follows. Before this PR, we figured out permissions at the top-level in `Hasura.GraphQL.Schema`, passing down the obtained `xPermInfo` objects as required. After this PR, we have a bottom-up approach where the schema building blocks themselves decide whether they want to be included for a particular role. So this moves some permission logic out of `Hasura.GraphQL.Schema`, which is very complex. PR-URL: https://github.com/hasura/graphql-engine-mono/pull/3608 GitOrigin-RevId: 51a744f34ec7d57bc8077667ae7f9cb9c4f6c962
2022-02-17 11:16:20 +03:00
[ selectionSetParser <&> fmap IR.SourceRelationshipArray,
aggSelectionSetParser <&> fmap IR.SourceRelationshipArrayAggregate
]
pure $
parsers <&> fmap \select ->
IR.RemoteSourceSelect _rsfiSource _rsfiSourceConfig select _rsfiMapping