graphql-engine/server/src-lib/Hasura/GraphQL/Schema/Relay.hs
Auke Booij 8ccf7724ce server: Metadata origin for definitions (type parameter version v2)
The code that builds the GraphQL schema, and `buildGQLContext` in particular, is partial: not every value of `(ServerConfigCtx, GraphQLQueryType, SourceCache, HashMap RemoteSchemaName (RemoteSchemaCtx, MetadataObject), ActionCache, AnnotatedCustomTypes)` results in a valid GraphQL schema. When it fails, we want to be able to return better error messages than we currently do.

The key thing that is missing is a way to trace back GraphQL type information to their origin from the Hasura metadata. Currently, we have a number of correctness checks of our GraphQL schema. But these correctness checks only have access to pure GraphQL type information, and hence can only report errors in terms of that. Possibly the worst is the "conflicting definitions" error, which, in practice, can only be debugged by Hasura engineers. This is terrible DX for customers.

This PR allows us to print better error messages, by adding a field to the `Definition` type that traces the GraphQL type to its origin in the metadata. So the idea is simple: just add `MetadataObjId`, or `Maybe` that, or some other sum type of that, to `Definition`.

However, we want to avoid having to import a `Hasura.RQL` module from `Hasura.GraphQL.Parser`. So we instead define this additional field of `Definition` through a new type parameter, which is threaded through in `Hasura.GraphQL.Parser`. We then define type synonyms in `Hasura.GraphQL.Schema.Parser` that fill in this type parameter, so that it is not visible for the majority of the codebase.

The idea of associating metadata information to `Definition`s really comes to fruition when combined with hasura/graphql-engine-mono#4517. Their combination would allow us to use the API of fatal errors (just like the current `MonadError QErr`) to report _inconsistencies_ in the metadata. Such inconsistencies are then _automatically_ ignored. So no ad-hoc decisions need to be made on how to cut out inconsistent metadata from the GraphQL schema. This will allow us to report much better errors, as well as improve the likelihood of a successful HGE startup.

PR-URL: https://github.com/hasura/graphql-engine-mono/pull/4770
Co-authored-by: Samir Talwar <47582+SamirTalwar@users.noreply.github.com>
GitOrigin-RevId: 728402b0cae83ae8e83463a826ceeb609001acae
2022-06-28 15:53:44 +00:00

221 lines
10 KiB
Haskell

{-# LANGUAGE TemplateHaskellQuotes #-}
-- | This module only exposes one function, 'nodeField', which is used at the
-- root level of the schema to create the 'node' field in the Relay API schema.
module Hasura.GraphQL.Schema.Relay
( nodeInterface,
nodeField,
)
where
import Control.Lens hiding (index)
import Data.Aeson qualified as J
import Data.Aeson.Extended qualified as J
import Data.Align (align)
import Data.Has
import Data.HashMap.Strict.Extended qualified as Map
import Data.Parser.JSONPath
import Data.Sequence.NonEmpty qualified as NESeq
import Data.Text qualified as T
import Data.Text.Extended
import Data.These (partitionThese)
import Hasura.Base.Error
import Hasura.GraphQL.Schema.Backend
import Hasura.GraphQL.Schema.Common
import Hasura.GraphQL.Schema.Instances ()
import Hasura.GraphQL.Schema.Node
import Hasura.GraphQL.Schema.Parser (Kind (..), Parser, memoizeOn)
import Hasura.GraphQL.Schema.Parser qualified as P
import Hasura.GraphQL.Schema.Select
import Hasura.GraphQL.Schema.Table
import Hasura.Name qualified as Name
import Hasura.Prelude
import Hasura.RQL.IR qualified as IR
import Hasura.RQL.Types.Backend
import Hasura.RQL.Types.Column
import Hasura.RQL.Types.Common
import Hasura.RQL.Types.SchemaCache hiding (askTableInfo)
import Hasura.RQL.Types.Source
import Hasura.RQL.Types.SourceCustomization
import Hasura.RQL.Types.Table
import Hasura.SQL.AnyBackend qualified as AB
import Hasura.SQL.Backend
import Language.GraphQL.Draft.Syntax qualified as G
-- | Constructs the parser for the node interface.
--
-- As mentioned in Note [Internal Relay HashMap], this function must parse an
-- incoming query for ANY potential matching table. Its resulting parser returns
-- a 'NodeMap': a container that, to a source name and a table name, associates
-- both the parsed fields and all the relevant table information required to
-- craft a request.
nodeInterface :: SourceCache -> NodeInterfaceParserBuilder
nodeInterface sourceCache = NodeInterfaceParserBuilder $ memoizeOn 'nodeInterface () do
let idDescription = G.Description "A globally unique identifier"
idField = P.selection_ Name._id (Just idDescription) P.identifier
nodeInterfaceDescription = G.Description "An object with globally unique ID"
tCase <- asks getter
tables :: [Parser 'Output n (SourceName, AB.AnyBackend TableMap)] <-
catMaybes . concat <$> for (Map.toList sourceCache) \(sourceName, anySourceInfo) ->
AB.dispatchAnyBackend @BackendSchema anySourceInfo \(sourceInfo :: SourceInfo b) ->
for (Map.toList $ takeValidTables $ _siTables sourceInfo) \(tableName, tableInfo) -> runMaybeT do
tablePkeyColumns <- hoistMaybe $ tableInfo ^? tiCoreInfo . tciPrimaryKey . _Just . pkColumns
selectPermissions <- MaybeT $ tableSelectPermissions tableInfo
annotatedFieldsParser <-
MaybeT $
P.withTypenameCustomization
(mkCustomizedTypename (_scTypeNames $ _siCustomization sourceInfo) tCase)
(tableSelectionSet sourceInfo tableInfo)
pure $
annotatedFieldsParser <&> \fields ->
( sourceName,
AB.mkAnyBackend $
TableMap $
Map.singleton tableName $
NodeInfo (_siConfiguration sourceInfo) selectPermissions tablePkeyColumns fields
)
pure $
Map.fromListWith fuseAnyMaps
<$> P.selectionSetInterface
Name._Node
(Just nodeInterfaceDescription)
[idField]
tables
where
-- this can only ever fail if somehow, within the same source, we ran into
-- two tables of a different type b; in other words, it is impossible.
fuseAnyMaps :: AB.AnyBackend TableMap -> AB.AnyBackend TableMap -> AB.AnyBackend TableMap
fuseAnyMaps m1 m2 =
AB.composeAnyBackend @Backend fuseMaps m1 m2 $
error "panic: two tables of a different backend type within the same source"
fuseMaps :: forall b. Backend b => TableMap b -> TableMap b -> AB.AnyBackend TableMap
fuseMaps (TableMap m1) (TableMap m2) = AB.mkAnyBackend @b $ TableMap $ Map.union m1 m2
-- | Creates a field parser for the top-level "node" field in the QueryRoot.
--
-- It exepcts one argument, the node id. It looks for the targeted table in the
-- 'NodeMap' returned by 'nodeInterface', and, if successful, attempts to craft
-- a corresponding 'QueryRootField' that will extract the requested row.
nodeField ::
forall m n r.
SourceCache ->
MonadBuildSchemaBase r m n =>
m (P.FieldParser n (IR.QueryRootField IR.UnpreparedValue))
nodeField sourceCache = do
let idDescription = G.Description "A globally unique id"
idArgument = P.field Name._id (Just idDescription) P.identifier
stringifyNum <- retrieve soStringifyNum
nodeObject <-
retrieve scSchemaKind >>= \case
HasuraSchema -> throw500 "internal error: the node field should only be built for the Relay schema"
RelaySchema nodeBuilder -> runNodeBuilder nodeBuilder
pure $
P.subselection Name._node Nothing idArgument nodeObject `P.bindField` \(ident, parseds) -> do
nodeId <- parseNodeId ident
case nodeId of
NodeIdV1 (V1NodeId tableName pKeys) -> do
-- Node id V1.
--
-- We don't have the source name in a V1 node; we attempt all of them
-- and pick the first one we find; there is a risk we might pick the
-- wrong one if two tables with the same name exist in different
-- sources! It is, however, unlikely; the engine emits V2 IDs, meaning
-- if ever encounter a V1 ID it means it has been manually entered bya
-- user, saved from an older version of the engine?
let matchingTables = flip mapMaybe (Map.keys sourceCache) \sourceName ->
(sourceName,) <$> findNode @('Postgres 'Vanilla) sourceName tableName parseds
case matchingTables of
[(sourceName, nodeValue)] -> createRootField stringifyNum sourceName tableName nodeValue pKeys
[] -> throwInvalidNodeId $ "no such table found: " <>> tableName
l ->
throwInvalidNodeId $
"this V1 node id matches more than one table across different sources: " <> tableName
<<> " exists in sources "
<> commaSeparated (fst <$> l)
NodeIdV2 nodev2 ->
-- Node id V2.
--
-- We have the source name and table name, we can extract the relevant
-- info directly.
AB.dispatchAnyBackend @Backend nodev2 \(V2NodeId sourceName tableName pKeys :: V2NodeId b) -> do
nodeValue <-
findNode @b sourceName tableName parseds
`onNothing` throwInvalidNodeId ("no table " <> tableName <<> " found in source " <>> sourceName)
createRootField stringifyNum sourceName tableName nodeValue pKeys
where
throwInvalidNodeId :: Text -> n a
throwInvalidNodeId t = P.withKey (Key "args") $ P.withKey (Key "id") $ P.parseError $ "invalid node id: " <> t
parseNodeId :: Text -> n NodeId
parseNodeId = either (throwInvalidNodeId . T.pack) pure . J.eitherDecode . base64Decode
-- Given all the node id information about a table, and the extracted
-- 'NodeInfo', craft the top-level query. This relies on the assumption
-- that all backends that support relay use the same IR for single row
-- selection.
createRootField ::
Backend b =>
StringifyNumbers ->
SourceName ->
TableName b ->
NodeInfo b ->
NESeq.NESeq J.Value ->
n (IR.QueryRootField IR.UnpreparedValue)
createRootField stringifyNum sourceName tableName (NodeInfo sourceConfig perms pKeys fields) columnValues = do
whereExp <- buildNodeIdBoolExp columnValues pKeys
pure $
IR.RFDB sourceName $
AB.mkAnyBackend $
IR.SourceConfigWith sourceConfig Nothing $
IR.QDBR $
IR.QDBSingleRow $
IR.AnnSelectG
{ IR._asnFields = fields,
IR._asnFrom = IR.FromTable tableName,
IR._asnPerm = tablePermissionsInfo perms,
IR._asnArgs =
IR.SelectArgs
{ IR._saWhere = Just whereExp,
IR._saOrderBy = Nothing,
IR._saLimit = Nothing,
IR._saOffset = Nothing,
IR._saDistinct = Nothing
},
IR._asnStrfyNum = stringifyNum
}
-- Craft the 'where' condition of the query by making an `AEQ` entry for
-- each primary key. This might fail if the given node id doesn't exactly
-- have a valid entry for each primary key.
buildNodeIdBoolExp ::
Backend b =>
NESeq.NESeq J.Value ->
NESeq.NESeq (ColumnInfo b) ->
n (IR.AnnBoolExp b (IR.UnpreparedValue b))
buildNodeIdBoolExp columnValues pkeyColumns = do
let firstPkColumn NESeq.:<|| remainingPkColumns = pkeyColumns
firstColumnValue NESeq.:<|| remainingColumns = columnValues
(nonAlignedPkColumns, nonAlignedColumnValues, alignedTuples) =
partitionThese $ toList $ align remainingPkColumns remainingColumns
unless (null nonAlignedPkColumns) $
throwInvalidNodeId $
"primary key columns " <> dquoteList (map ciColumn nonAlignedPkColumns) <> " are missing"
unless (null nonAlignedColumnValues) $
throwInvalidNodeId $
"unexpected column values " <> J.encodeToStrictText nonAlignedColumnValues
let allTuples = (firstPkColumn, firstColumnValue) : alignedTuples
IR.BoolAnd <$> for allTuples \(columnInfo, columnValue) -> do
let columnType = ciType columnInfo
parsedValue <-
parseScalarValueColumnType columnType columnValue `onLeft` \e ->
P.parseErrorWith ParseFailed $ "value of column " <> ciColumn columnInfo <<> " in node id: " <> qeError e
pure $
IR.BoolFld $
IR.AVColumn
columnInfo
[IR.AEQ True $ IR.UVParameter Nothing $ ColumnValue columnType parsedValue]