graphql-engine/server/src-lib/Hasura/Backends/Postgres/Instances/Schema.hs

861 lines
38 KiB
Haskell
Raw Normal View History

{-# LANGUAGE ApplicativeDo #-}
{-# LANGUAGE UndecidableInstances #-}
{-# OPTIONS_GHC -fno-warn-orphans #-}
-- | Postgres Instances Schema
--
-- Defines a 'Hasura.GraphQL.Schema.Backend.BackendSchema' type class instance for Postgres.
module Hasura.Backends.Postgres.Instances.Schema
(
)
where
import Data.Aeson qualified as J
import Data.Has
import Data.HashMap.Strict qualified as Map
import Data.HashMap.Strict.Extended qualified as M
import Data.List.NonEmpty qualified as NE
import Data.Parser.JSONPath
import Data.Text qualified as T
import Data.Text.Extended
import Hasura.Backends.Postgres.SQL.DML as PG hiding (CountType, incOp)
import Hasura.Backends.Postgres.SQL.Types as PG hiding (FunctionName, TableName)
import Hasura.Backends.Postgres.SQL.Value as PG
import Hasura.Backends.Postgres.Schema.OnConflict
import Hasura.Backends.Postgres.Types.BoolExp
import Hasura.Backends.Postgres.Types.Column
import Hasura.Backends.Postgres.Types.Insert as PGIR
import Hasura.Backends.Postgres.Types.Update as PGIR
import Hasura.Base.Error
import Hasura.GraphQL.Parser hiding (EnumValueInfo, field)
import Hasura.GraphQL.Parser qualified as P
import Hasura.GraphQL.Parser.Internal.Parser hiding (field)
import Hasura.GraphQL.Schema.Backend
( BackendSchema,
ComparisonExp,
MonadBuildSchema,
)
import Hasura.GraphQL.Schema.Backend qualified as BS
import Hasura.GraphQL.Schema.BoolExp
import Hasura.GraphQL.Schema.Build qualified as GSB
import Hasura.GraphQL.Schema.Common
import Hasura.GraphQL.Schema.Mutation qualified as GSB
import Hasura.GraphQL.Schema.Select
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
import Hasura.GraphQL.Schema.Table
import Hasura.GraphQL.Schema.Update qualified as SU
import Hasura.Prelude
import Hasura.RQL.IR.BoolExp
import Hasura.RQL.IR.Root (RemoteRelationshipField)
import Hasura.RQL.IR.Select
( QueryDB (QDBConnection),
)
import Hasura.RQL.IR.Select qualified as IR
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
import Hasura.RQL.IR.Update qualified as IR
import Hasura.RQL.Types.Backend (Backend (..))
import Hasura.RQL.Types.Column
import Hasura.RQL.Types.Common
import Hasura.RQL.Types.Function (FunctionInfo)
import Hasura.RQL.Types.SourceCustomization (mkRootFieldName)
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
import Hasura.RQL.Types.Table (RolePermInfo (..), SelPermInfo, TableInfo, UpdPermInfo)
import Hasura.SQL.Backend (BackendType (Postgres), PostgresKind (Citus, Vanilla))
import Hasura.SQL.Types
import Language.GraphQL.Draft.Syntax qualified as G
----------------------------------------------------------------
-- BackendSchema instance
-- | This class is an implementation detail of 'BackendSchema'.
-- Some functions of 'BackendSchema' differ across different Postgres "kinds",
-- or call to functions (such as those related to Relay) that have not been
-- generalized to all kinds of Postgres and still explicitly work on Vanilla
-- Postgres. This class allows each "kind" to specify its own specific
-- implementation. All common code is directly part of `BackendSchema`.
--
-- Note: Users shouldn't ever put this as a constraint. Use `BackendSchema
-- ('Postgres pgKind)` instead.
class PostgresSchema (pgKind :: PostgresKind) where
pgkBuildTableRelayQueryFields ::
BS.MonadBuildSchema ('Postgres pgKind) r m n =>
SourceName ->
TableName ('Postgres pgKind) ->
TableInfo ('Postgres pgKind) ->
G.Name ->
NESeq (ColumnInfo ('Postgres pgKind)) ->
m [FieldParser n (QueryDB ('Postgres pgKind) (RemoteRelationshipField UnpreparedValue) (UnpreparedValue ('Postgres pgKind)))]
pgkBuildFunctionRelayQueryFields ::
BS.MonadBuildSchema ('Postgres pgKind) r m n =>
SourceName ->
FunctionName ('Postgres pgKind) ->
FunctionInfo ('Postgres pgKind) ->
TableName ('Postgres pgKind) ->
NESeq (ColumnInfo ('Postgres pgKind)) ->
m [FieldParser n (QueryDB ('Postgres pgKind) (RemoteRelationshipField UnpreparedValue) (UnpreparedValue ('Postgres pgKind)))]
pgkRelayExtension ::
Maybe (XRelay ('Postgres pgKind))
pgkNode ::
BS.MonadBuildSchema ('Postgres pgKind) r m n =>
m
( Parser
'Output
n
( HashMap
( TableName ('Postgres pgKind)
)
( SourceName,
SourceConfig ('Postgres pgKind),
SelPermInfo ('Postgres pgKind),
PrimaryKeyColumns ('Postgres pgKind),
AnnotatedFields ('Postgres pgKind)
)
)
)
instance PostgresSchema 'Vanilla where
pgkBuildTableRelayQueryFields = buildTableRelayQueryFields
pgkBuildFunctionRelayQueryFields = buildFunctionRelayQueryFields
pgkRelayExtension = Just ()
pgkNode = nodePG
instance PostgresSchema 'Citus where
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
pgkBuildTableRelayQueryFields _ _ _ _ _ = pure []
pgkBuildFunctionRelayQueryFields _ _ _ _ _ = pure []
pgkRelayExtension = Nothing
pgkNode = undefined
-- postgres schema
instance
( Backend ('Postgres pgKind),
PostgresSchema pgKind
) =>
BackendSchema ('Postgres pgKind)
where
-- top level parsers
buildTableQueryFields = GSB.buildTableQueryFields
buildTableRelayQueryFields = pgkBuildTableRelayQueryFields
buildTableInsertMutationFields = GSB.buildTableInsertMutationFields backendInsertParser
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
buildTableUpdateMutationFields = pgkBuildTableUpdateMutationFields
buildTableDeleteMutationFields = GSB.buildTableDeleteMutationFields
buildFunctionQueryFields = GSB.buildFunctionQueryFields
buildFunctionRelayQueryFields = pgkBuildFunctionRelayQueryFields
buildFunctionMutationFields = GSB.buildFunctionMutationFields
-- table components
tableArguments = defaultTableArgs
mkRelationshipParser = GSB.mkDefaultRelationshipParser backendInsertParser ()
-- backend extensions
relayExtension = pgkRelayExtension @pgKind
nodesAggExtension = Just ()
-- indivdual components
columnParser = columnParser
jsonPathArg = jsonPathArg
orderByOperators = orderByOperators
comparisonExps = comparisonExps
countTypeInput = countTypeInput
aggregateOrderByCountType = PG.PGInteger
computedField = computedFieldPG
node = pgkNode
-- SQL literals
columnDefaultValue = const PG.columnDefaultValue
backendInsertParser ::
forall pgKind m r n.
MonadBuildSchema ('Postgres pgKind) r m n =>
SourceName ->
TableInfo ('Postgres pgKind) ->
m (InputFieldsParser n (PGIR.BackendInsert pgKind (UnpreparedValue ('Postgres pgKind))))
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
backendInsertParser sourceName tableInfo =
fmap BackendInsert <$> onConflictFieldParser sourceName tableInfo
----------------------------------------------------------------
-- Top level parsers
buildTableRelayQueryFields ::
forall pgKind m n r.
MonadBuildSchema ('Postgres pgKind) r m n =>
SourceName ->
TableName ('Postgres pgKind) ->
TableInfo ('Postgres pgKind) ->
G.Name ->
NESeq (ColumnInfo ('Postgres pgKind)) ->
m [FieldParser n (QueryDB ('Postgres pgKind) (RemoteRelationshipField UnpreparedValue) (UnpreparedValue ('Postgres pgKind)))]
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
buildTableRelayQueryFields sourceName tableName tableInfo gqlName pkeyColumns = do
let fieldDesc = Just $ G.Description $ "fetch data from the table: " <>> tableName
fieldName <- mkRootFieldName $ gqlName <> $$(G.litName "_connection")
fmap afold $
optionalFieldParser QDBConnection $
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
selectTableConnection sourceName tableInfo fieldName fieldDesc pkeyColumns
pgkBuildTableUpdateMutationFields ::
MonadBuildSchema ('Postgres pgKind) r m n =>
-- | The source that the table lives in
SourceName ->
-- | The name of the table being acted on
TableName ('Postgres pgKind) ->
-- | table info
TableInfo ('Postgres pgKind) ->
-- | field display name
G.Name ->
m [FieldParser n (IR.AnnotatedUpdateG ('Postgres pgKind) (RemoteRelationshipField UnpreparedValue) (UnpreparedValue ('Postgres pgKind)))]
pgkBuildTableUpdateMutationFields sourceName tableName tableInfo gqlName =
concat . maybeToList <$> runMaybeT do
updatePerms <- MaybeT $ (_permUpd =<<) <$> tablePermissions tableInfo
lift $
GSB.buildTableUpdateMutationFields
-- TODO: https://github.com/hasura/graphql-engine-mono/issues/2955
(\ti -> fmap BackendUpdate <$> updateOperators ti updatePerms)
sourceName
tableName
tableInfo
gqlName
buildFunctionRelayQueryFields ::
forall pgKind m n r.
MonadBuildSchema ('Postgres pgKind) r m n =>
SourceName ->
FunctionName ('Postgres pgKind) ->
FunctionInfo ('Postgres pgKind) ->
TableName ('Postgres pgKind) ->
NESeq (ColumnInfo ('Postgres pgKind)) ->
m [FieldParser n (QueryDB ('Postgres pgKind) (RemoteRelationshipField UnpreparedValue) (UnpreparedValue ('Postgres pgKind)))]
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
buildFunctionRelayQueryFields sourceName functionName functionInfo tableName pkeyColumns = do
let fieldDesc = Just $ G.Description $ "execute function " <> functionName <<> " which returns " <>> tableName
fmap afold $
optionalFieldParser QDBConnection $
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
selectFunctionConnection sourceName functionInfo fieldDesc pkeyColumns
----------------------------------------------------------------
-- Individual components
columnParser ::
(MonadSchema n m, MonadError QErr m, MonadReader r m, Has P.MkTypename r) =>
ColumnType ('Postgres pgKind) ->
G.Nullability ->
m (Parser 'Both n (ValueWithOrigin (ColumnValue ('Postgres pgKind))))
columnParser columnType (G.Nullability isNullable) =
-- TODO(PDV): It might be worth memoizing this function even though it isnt
-- recursive simply for performance reasons, since its likely to be hammered
-- during schema generation. Need to profile to see whether or not its a win.
server: remove remnants of query plan caching (fix #1795) Query plan caching was introduced by - I believe - hasura/graphql-engine#1934 in order to reduce the query response latency. During the development of PDV in hasura/graphql-engine#4111, it was found out that the new architecture (for which query plan caching wasn't implemented) performed comparably to the pre-PDV architecture with caching. Hence, it was decided to leave query plan caching until some day in the future when it was deemed necessary. Well, we're in the future now, and there still isn't a convincing argument for query plan caching. So the time has come to remove some references to query plan caching from the codebase. For the most part, any code being removed would probably not be very well suited to the post-PDV architecture of query execution, so arguably not much is lost. Apart from simplifying the code, this PR will contribute towards making the GraphQL schema generation more modular, testable, and easier to profile. I'd like to eventually work towards a situation in which it's easy to generate a GraphQL schema parser *in isolation*, without being connected to a database, and then parse a GraphQL query *in isolation*, without even listening any HTTP port. It is important that both of these operations can be examined in detail, and in isolation, since they are two major performance bottlenecks, as well as phases where many important upcoming features hook into. Implementation The following have been removed: - The entirety of `server/src-lib/Hasura/GraphQL/Execute/Plan.hs` - The core phases of query parsing and execution no longer have any references to query plan caching. Note that this is not to be confused with query *response* caching, which is not affected by this PR. This includes removal of the types: - - `Opaque`, which is replaced by a tuple. Note that the old implementation was broken and did not adequately hide the constructors. - - `QueryReusability` (and the `markNotReusable` method). Notably, the implementation of the `ParseT` monad now consists of two, rather than three, monad transformers. - Cache-related tests (in `server/src-test/Hasura/CacheBoundedSpec.hs`) have been removed . - References to query plan caching in the documentation. - The `planCacheOptions` in the `TenantConfig` type class was removed. However, during parsing, unrecognized fields in the YAML config get ignored, so this does not cause a breaking change. (Confirmed manually, as well as in consultation with @sordina.) - The metrics no longer send cache hit/miss messages. There are a few places in which one can still find references to query plan caching: - We still accept the `--query-plan-cache-size` command-line option for backwards compatibility. The `HASURA_QUERY_PLAN_CACHE_SIZE` environment variable is not read. https://github.com/hasura/graphql-engine-mono/pull/1815 GitOrigin-RevId: 17d92b254ec093c62a7dfeec478658ede0813eb7
2021-07-27 14:51:52 +03:00
peelWithOrigin . fmap (ColumnValue columnType) <$> case columnType of
ColumnScalar scalarType ->
possiblyNullable scalarType <$> do
-- We convert the value to JSON and use the FromJSON instance. This avoids
-- having two separate ways of parsing a value in the codebase, which
-- could lead to inconsistencies.
--
-- The mapping from postgres type to GraphQL scalar name is done by
-- 'mkScalarTypeName'. This is confusing, and we might want to fix it
-- later, as we will parse values differently here than how they'd be
-- parsed in other places using the same scalar name; for instance, we
-- will accept strings for postgres columns of type "Integer", despite the
-- fact that they will be represented as GraphQL ints, which otherwise do
-- not accept strings.
--
-- TODO: introduce new dedicated scalars for Postgres column types.
Refactor type name customization Source typename customization (hasura/graphql-engine@aac64f2c81faa6a3aef4d0cf5fae97289ac4383e) introduced a mechanism to change certain names in the GraphQL schema that is exposed. In particular it allows last-minute modification of: 1. the names of some types, and 2. the names of some root fields. The above two items are assigned distinct customization algorithms, and at times both algorithms are in scope. So a need to distinguish them is needed. In the original design, this was addressed by introducing a newtype wrapper `Typename` around GraphQL `Name`s, dedicated to the names of types. However, in the majority of the codebase, type names are also represented by `Name`. For this reason, it was unavoidable to allow for easy conversion. This was supported by a `HasName Typename` instance, as well as by publishing the constructors of `Typename`. This means that the type safety that newtypes can add is lost. In particular, it is now very easy to confuse type name customization with root field name customization. This refactors the above design by instead introducing newtypes around the customization operations: ```haskell newtype MkTypename = MkTypename {runMkTypename :: Name -> Name} deriving (Semigroup, Monoid) via (Endo Name) newtype MkRootFieldName = MkRootFieldName {runMkRootFieldName :: Name -> Name} deriving (Semigroup, Monoid) via (Endo Name) ``` The `Monoid` instance allows easy composition of customization operations, piggybacking off of the type of `Endo`maps. This design allows safe co-existence of the two customization algorithms, while avoiding the syntactic overhead of packing and unpacking newtypes. PR-URL: https://github.com/hasura/graphql-engine-mono/pull/2989 GitOrigin-RevId: da3a353a9b003ee40c8d0a1e02872e99d2edd3ca
2021-11-30 12:51:46 +03:00
name <- mkScalarTypeName scalarType
let schemaType = P.TNamed P.NonNullable $ P.Definition name Nothing P.TIScalar
pure $
Parser
{ pType = schemaType,
pParser =
valueToJSON (P.toGraphQLType schemaType) >=> \case
J.Null -> parseError $ "unexpected null value for type " <>> name
value ->
runAesonParser (parsePGValue scalarType) value
`onLeft` (parseErrorWith ParseFailed . qeError)
}
ColumnEnumReference (EnumReference tableName enumValues) ->
case nonEmpty (Map.toList enumValues) of
Just enumValuesList -> do
name <- qualifiedObjectToName tableName <&> (<> $$(G.litName "_enum")) >>= P.mkTypename
pure $ possiblyNullable PGText $ P.enum name Nothing (mkEnumValue <$> enumValuesList)
Nothing -> throw400 ValidationFailed "empty enum values"
where
possiblyNullable scalarType
| isNullable = fmap (fromMaybe $ PGNull scalarType) . P.nullable
| otherwise = id
mkEnumValue :: (EnumValue, EnumValueInfo) -> (P.Definition P.EnumValueInfo, PGScalarValue)
mkEnumValue (EnumValue value, EnumValueInfo description) =
Remove `Unique` from `Definition` GraphQL types can refer to each other in a circular way. The PDV framework used to use values of type `Unique` to recognize two fragments of GraphQL schema as being the same instance. Internally, this is based on `Data.Unique` from the `base` package, which simply increases a counter on every creation of a `Unique` object. **NB**: The `Unique` values are _not_ used for knot tying the schema combinators themselves (i.e. `Parser`s). The knot tying for `Parser`s is purely based on keys provided to `memoizeOn`. The `Unique` values are _only_ used to recognize two pieces of GraphQL _schema_ as being identical. Originally, the idea was that this would help us with a perfectly correct identification of GraphQL types. But this fully correct equality checking of GraphQL types was never implemented, and does not seem to be necessary to prevent bugs. Specifically, these `Unique` values are stored as part of `data Definition a`, which specifies a part of our internal abstract syntax tree for the GraphQL types that we expose. The `Unique` values get initialized by the `SchemaT` effect. In #2894 and #2895, we are experimenting with how (parts of) the GraphQL types can be hidden behind certain permission predicates. This would allow a single GraphQL schema in memory to serve all roles, implementing #2711. The permission predicates get evaluated at query parsing time when we know what role is doing a certain request, thus outputting the correct GraphQL types for that role. If the approach of #2895 is followed, then the `Definition` objects, and thus the `Unique` values, would be hidden behind the permission predicates. Since the permission predicates are evaluated only after the schema is already supposed to be built, this means that the permission predicates would prevent us from initializing the `Unique` values, rendering them useless. The simplest remedy to this is to remove our usage of `Unique` altogether from the GraphQL schema and schema combinators. It doesn't serve a functional purpose, doesn't prevent bugs, and requires extra bookkeeping. PR-URL: https://github.com/hasura/graphql-engine-mono/pull/2980 GitOrigin-RevId: 50d3f9e0b9fbf578ac49c8fc773ba64a94b1f43d
2021-12-01 19:20:35 +03:00
( P.Definition value (G.Description <$> description) P.EnumValueInfo,
PGValText $ G.unName value
)
jsonPathArg ::
MonadParse n =>
ColumnType ('Postgres pgKind) ->
InputFieldsParser n (Maybe (IR.ColumnOp ('Postgres pgKind)))
jsonPathArg columnType
| isScalarColumnWhere PG.isJSONType columnType =
P.fieldOptional fieldName description P.string `P.bindFields` fmap join . traverse toColExp
| otherwise = pure Nothing
where
fieldName = $$(G.litName "path")
description = Just "JSON select path"
toColExp textValue = case parseJSONPath textValue of
Left err -> parseError $ T.pack $ "parse json path error: " ++ err
Right [] -> pure Nothing
Right jPaths -> pure $ Just $ IR.ColumnOp PG.jsonbPathOp $ PG.SEArray $ map elToColExp jPaths
elToColExp (Key k) = PG.SELit k
elToColExp (Index i) = PG.SELit $ tshow i
orderByOperators ::
NonEmpty (Definition P.EnumValueInfo, (BasicOrderType ('Postgres pgKind), NullsOrderType ('Postgres pgKind)))
orderByOperators =
NE.fromList
[ ( define $$(G.litName "asc") "in ascending order, nulls last",
(PG.OTAsc, PG.NLast)
),
( define $$(G.litName "asc_nulls_first") "in ascending order, nulls first",
(PG.OTAsc, PG.NFirst)
),
( define $$(G.litName "asc_nulls_last") "in ascending order, nulls last",
(PG.OTAsc, PG.NLast)
),
( define $$(G.litName "desc") "in descending order, nulls first",
(PG.OTDesc, PG.NFirst)
),
( define $$(G.litName "desc_nulls_first") "in descending order, nulls first",
(PG.OTDesc, PG.NFirst)
),
( define $$(G.litName "desc_nulls_last") "in descending order, nulls last",
(PG.OTDesc, PG.NLast)
)
]
where
Remove `Unique` from `Definition` GraphQL types can refer to each other in a circular way. The PDV framework used to use values of type `Unique` to recognize two fragments of GraphQL schema as being the same instance. Internally, this is based on `Data.Unique` from the `base` package, which simply increases a counter on every creation of a `Unique` object. **NB**: The `Unique` values are _not_ used for knot tying the schema combinators themselves (i.e. `Parser`s). The knot tying for `Parser`s is purely based on keys provided to `memoizeOn`. The `Unique` values are _only_ used to recognize two pieces of GraphQL _schema_ as being identical. Originally, the idea was that this would help us with a perfectly correct identification of GraphQL types. But this fully correct equality checking of GraphQL types was never implemented, and does not seem to be necessary to prevent bugs. Specifically, these `Unique` values are stored as part of `data Definition a`, which specifies a part of our internal abstract syntax tree for the GraphQL types that we expose. The `Unique` values get initialized by the `SchemaT` effect. In #2894 and #2895, we are experimenting with how (parts of) the GraphQL types can be hidden behind certain permission predicates. This would allow a single GraphQL schema in memory to serve all roles, implementing #2711. The permission predicates get evaluated at query parsing time when we know what role is doing a certain request, thus outputting the correct GraphQL types for that role. If the approach of #2895 is followed, then the `Definition` objects, and thus the `Unique` values, would be hidden behind the permission predicates. Since the permission predicates are evaluated only after the schema is already supposed to be built, this means that the permission predicates would prevent us from initializing the `Unique` values, rendering them useless. The simplest remedy to this is to remove our usage of `Unique` altogether from the GraphQL schema and schema combinators. It doesn't serve a functional purpose, doesn't prevent bugs, and requires extra bookkeeping. PR-URL: https://github.com/hasura/graphql-engine-mono/pull/2980 GitOrigin-RevId: 50d3f9e0b9fbf578ac49c8fc773ba64a94b1f43d
2021-12-01 19:20:35 +03:00
define name desc = P.Definition name (Just desc) P.EnumValueInfo
comparisonExps ::
forall pgKind m n r.
( BackendSchema ('Postgres pgKind),
MonadSchema n m,
MonadError QErr m,
MonadReader r m,
Has QueryContext r,
Has MkTypename r
) =>
ColumnType ('Postgres pgKind) ->
m (Parser 'Input n [ComparisonExp ('Postgres pgKind)])
comparisonExps = P.memoize 'comparisonExps \columnType -> do
-- see Note [Columns in comparison expression are never nullable]
collapseIfNull <- asks $ qcDangerousBooleanCollapse . getter
-- parsers used for comparison arguments
geogInputParser <- geographyWithinDistanceInput
geomInputParser <- geometryWithinDistanceInput
ignInputParser <- intersectsGeomNbandInput
ingInputParser <- intersectsNbandGeomInput
typedParser <- columnParser columnType (G.Nullability False)
nullableTextParser <- columnParser (ColumnScalar PGText) (G.Nullability True)
textParser <- columnParser (ColumnScalar PGText) (G.Nullability False)
-- `lquery` represents a regular-expression-like pattern for matching `ltree` values.
lqueryParser <- columnParser (ColumnScalar PGLquery) (G.Nullability False)
-- `ltxtquery` represents a full-text-search-like pattern for matching `ltree` values.
ltxtqueryParser <- columnParser (ColumnScalar PGLtxtquery) (G.Nullability False)
maybeCastParser <- castExp columnType
Refactor type name customization Source typename customization (hasura/graphql-engine@aac64f2c81faa6a3aef4d0cf5fae97289ac4383e) introduced a mechanism to change certain names in the GraphQL schema that is exposed. In particular it allows last-minute modification of: 1. the names of some types, and 2. the names of some root fields. The above two items are assigned distinct customization algorithms, and at times both algorithms are in scope. So a need to distinguish them is needed. In the original design, this was addressed by introducing a newtype wrapper `Typename` around GraphQL `Name`s, dedicated to the names of types. However, in the majority of the codebase, type names are also represented by `Name`. For this reason, it was unavoidable to allow for easy conversion. This was supported by a `HasName Typename` instance, as well as by publishing the constructors of `Typename`. This means that the type safety that newtypes can add is lost. In particular, it is now very easy to confuse type name customization with root field name customization. This refactors the above design by instead introducing newtypes around the customization operations: ```haskell newtype MkTypename = MkTypename {runMkTypename :: Name -> Name} deriving (Semigroup, Monoid) via (Endo Name) newtype MkRootFieldName = MkRootFieldName {runMkRootFieldName :: Name -> Name} deriving (Semigroup, Monoid) via (Endo Name) ``` The `Monoid` instance allows easy composition of customization operations, piggybacking off of the type of `Endo`maps. This design allows safe co-existence of the two customization algorithms, while avoiding the syntactic overhead of packing and unpacking newtypes. PR-URL: https://github.com/hasura/graphql-engine-mono/pull/2989 GitOrigin-RevId: da3a353a9b003ee40c8d0a1e02872e99d2edd3ca
2021-11-30 12:51:46 +03:00
let name = P.getName typedParser <> $$(G.litName "_comparison_exp")
desc =
G.Description $
"Boolean expression to compare columns of type "
<> P.getName typedParser
<<> ". All fields are combined with logical 'AND'."
textListParser = fmap openValueOrigin <$> P.list textParser
server: remove remnants of query plan caching (fix #1795) Query plan caching was introduced by - I believe - hasura/graphql-engine#1934 in order to reduce the query response latency. During the development of PDV in hasura/graphql-engine#4111, it was found out that the new architecture (for which query plan caching wasn't implemented) performed comparably to the pre-PDV architecture with caching. Hence, it was decided to leave query plan caching until some day in the future when it was deemed necessary. Well, we're in the future now, and there still isn't a convincing argument for query plan caching. So the time has come to remove some references to query plan caching from the codebase. For the most part, any code being removed would probably not be very well suited to the post-PDV architecture of query execution, so arguably not much is lost. Apart from simplifying the code, this PR will contribute towards making the GraphQL schema generation more modular, testable, and easier to profile. I'd like to eventually work towards a situation in which it's easy to generate a GraphQL schema parser *in isolation*, without being connected to a database, and then parse a GraphQL query *in isolation*, without even listening any HTTP port. It is important that both of these operations can be examined in detail, and in isolation, since they are two major performance bottlenecks, as well as phases where many important upcoming features hook into. Implementation The following have been removed: - The entirety of `server/src-lib/Hasura/GraphQL/Execute/Plan.hs` - The core phases of query parsing and execution no longer have any references to query plan caching. Note that this is not to be confused with query *response* caching, which is not affected by this PR. This includes removal of the types: - - `Opaque`, which is replaced by a tuple. Note that the old implementation was broken and did not adequately hide the constructors. - - `QueryReusability` (and the `markNotReusable` method). Notably, the implementation of the `ParseT` monad now consists of two, rather than three, monad transformers. - Cache-related tests (in `server/src-test/Hasura/CacheBoundedSpec.hs`) have been removed . - References to query plan caching in the documentation. - The `planCacheOptions` in the `TenantConfig` type class was removed. However, during parsing, unrecognized fields in the YAML config get ignored, so this does not cause a breaking change. (Confirmed manually, as well as in consultation with @sordina.) - The metrics no longer send cache hit/miss messages. There are a few places in which one can still find references to query plan caching: - We still accept the `--query-plan-cache-size` command-line option for backwards compatibility. The `HASURA_QUERY_PLAN_CACHE_SIZE` environment variable is not read. https://github.com/hasura/graphql-engine-mono/pull/1815 GitOrigin-RevId: 17d92b254ec093c62a7dfeec478658ede0813eb7
2021-07-27 14:51:52 +03:00
columnListParser = fmap openValueOrigin <$> P.list typedParser
pure $
P.object name (Just desc) $
fmap catMaybes $
sequenceA $
concat
[ flip (maybe []) maybeCastParser $ \castParser ->
[ P.fieldOptional $$(G.litName "_cast") Nothing (ACast <$> castParser)
],
-- Common ops for all types
equalityOperators
collapseIfNull
(mkParameter <$> typedParser)
(mkListLiteral columnType <$> columnListParser),
-- Comparison ops for non Raster types
guard (isScalarColumnWhere (/= PGRaster) columnType)
*> comparisonOperators
collapseIfNull
(mkParameter <$> typedParser),
-- Ops for Raster types
guard (isScalarColumnWhere (== PGRaster) columnType)
*> [ mkBoolOperator
collapseIfNull
$$(G.litName "_st_intersects_rast")
Nothing
(ABackendSpecific . ASTIntersectsRast . mkParameter <$> typedParser),
mkBoolOperator
collapseIfNull
$$(G.litName "_st_intersects_nband_geom")
Nothing
(ABackendSpecific . ASTIntersectsNbandGeom <$> ingInputParser),
mkBoolOperator
collapseIfNull
$$(G.litName "_st_intersects_geom_nband")
Nothing
(ABackendSpecific . ASTIntersectsGeomNband <$> ignInputParser)
],
-- Ops for String like types
guard (isScalarColumnWhere isStringType columnType)
*> [ mkBoolOperator
collapseIfNull
$$(G.litName "_like")
(Just "does the column match the given pattern")
(ALIKE . mkParameter <$> typedParser),
mkBoolOperator
collapseIfNull
$$(G.litName "_nlike")
(Just "does the column NOT match the given pattern")
(ANLIKE . mkParameter <$> typedParser),
mkBoolOperator
collapseIfNull
$$(G.litName "_ilike")
(Just "does the column match the given case-insensitive pattern")
(ABackendSpecific . AILIKE . mkParameter <$> typedParser),
mkBoolOperator
collapseIfNull
$$(G.litName "_nilike")
(Just "does the column NOT match the given case-insensitive pattern")
(ABackendSpecific . ANILIKE . mkParameter <$> typedParser),
mkBoolOperator
collapseIfNull
$$(G.litName "_similar")
(Just "does the column match the given SQL regular expression")
(ABackendSpecific . ASIMILAR . mkParameter <$> typedParser),
mkBoolOperator
collapseIfNull
$$(G.litName "_nsimilar")
(Just "does the column NOT match the given SQL regular expression")
(ABackendSpecific . ANSIMILAR . mkParameter <$> typedParser),
mkBoolOperator
collapseIfNull
$$(G.litName "_regex")
(Just "does the column match the given POSIX regular expression, case sensitive")
(ABackendSpecific . AREGEX . mkParameter <$> typedParser),
mkBoolOperator
collapseIfNull
$$(G.litName "_iregex")
(Just "does the column match the given POSIX regular expression, case insensitive")
(ABackendSpecific . AIREGEX . mkParameter <$> typedParser),
mkBoolOperator
collapseIfNull
$$(G.litName "_nregex")
(Just "does the column NOT match the given POSIX regular expression, case sensitive")
(ABackendSpecific . ANREGEX . mkParameter <$> typedParser),
mkBoolOperator
collapseIfNull
$$(G.litName "_niregex")
(Just "does the column NOT match the given POSIX regular expression, case insensitive")
(ABackendSpecific . ANIREGEX . mkParameter <$> typedParser)
],
-- Ops for JSONB type
guard (isScalarColumnWhere (== PGJSONB) columnType)
*> [ mkBoolOperator
collapseIfNull
$$(G.litName "_contains")
(Just "does the column contain the given json value at the top level")
(ABackendSpecific . AContains . mkParameter <$> typedParser),
mkBoolOperator
collapseIfNull
$$(G.litName "_contained_in")
(Just "is the column contained in the given json value")
(ABackendSpecific . AContainedIn . mkParameter <$> typedParser),
mkBoolOperator
collapseIfNull
$$(G.litName "_has_key")
(Just "does the string exist as a top-level key in the column")
(ABackendSpecific . AHasKey . mkParameter <$> nullableTextParser),
mkBoolOperator
collapseIfNull
$$(G.litName "_has_keys_any")
(Just "do any of these strings exist as top-level keys in the column")
(ABackendSpecific . AHasKeysAny . mkListLiteral (ColumnScalar PGText) <$> textListParser),
mkBoolOperator
collapseIfNull
$$(G.litName "_has_keys_all")
(Just "do all of these strings exist as top-level keys in the column")
(ABackendSpecific . AHasKeysAll . mkListLiteral (ColumnScalar PGText) <$> textListParser)
],
-- Ops for Geography type
guard (isScalarColumnWhere (== PGGeography) columnType)
*> [ mkBoolOperator
collapseIfNull
$$(G.litName "_st_intersects")
(Just "does the column spatially intersect the given geography value")
(ABackendSpecific . ASTIntersects . mkParameter <$> typedParser),
mkBoolOperator
collapseIfNull
$$(G.litName "_st_d_within")
(Just "is the column within a given distance from the given geography value")
(ABackendSpecific . ASTDWithinGeog <$> geogInputParser)
],
-- Ops for Geometry type
guard (isScalarColumnWhere (== PGGeometry) columnType)
*> [ mkBoolOperator
collapseIfNull
$$(G.litName "_st_contains")
(Just "does the column contain the given geometry value")
(ABackendSpecific . ASTContains . mkParameter <$> typedParser),
mkBoolOperator
collapseIfNull
$$(G.litName "_st_crosses")
(Just "does the column cross the given geometry value")
(ABackendSpecific . ASTCrosses . mkParameter <$> typedParser),
mkBoolOperator
collapseIfNull
$$(G.litName "_st_equals")
(Just "is the column equal to given geometry value (directionality is ignored)")
(ABackendSpecific . ASTEquals . mkParameter <$> typedParser),
mkBoolOperator
collapseIfNull
$$(G.litName "_st_overlaps")
(Just "does the column 'spatially overlap' (intersect but not completely contain) the given geometry value")
(ABackendSpecific . ASTOverlaps . mkParameter <$> typedParser),
mkBoolOperator
collapseIfNull
$$(G.litName "_st_touches")
(Just "does the column have atleast one point in common with the given geometry value")
(ABackendSpecific . ASTTouches . mkParameter <$> typedParser),
mkBoolOperator
collapseIfNull
$$(G.litName "_st_within")
(Just "is the column contained in the given geometry value")
(ABackendSpecific . ASTWithin . mkParameter <$> typedParser),
mkBoolOperator
collapseIfNull
$$(G.litName "_st_intersects")
(Just "does the column spatially intersect the given geometry value")
(ABackendSpecific . ASTIntersects . mkParameter <$> typedParser),
mkBoolOperator
collapseIfNull
$$(G.litName "_st_3d_intersects")
(Just "does the column spatially intersect the given geometry value in 3D")
(ABackendSpecific . AST3DIntersects . mkParameter <$> typedParser),
mkBoolOperator
collapseIfNull
$$(G.litName "_st_d_within")
(Just "is the column within a given distance from the given geometry value")
(ABackendSpecific . ASTDWithinGeom <$> geomInputParser),
mkBoolOperator
collapseIfNull
$$(G.litName "_st_3d_d_within")
(Just "is the column within a given 3D distance from the given geometry value")
(ABackendSpecific . AST3DDWithinGeom <$> geomInputParser)
],
-- Ops for Ltree type
guard (isScalarColumnWhere (== PGLtree) columnType)
*> [ mkBoolOperator
collapseIfNull
$$(G.litName "_ancestor")
(Just "is the left argument an ancestor of right (or equal)?")
(ABackendSpecific . AAncestor . mkParameter <$> typedParser),
mkBoolOperator
collapseIfNull
$$(G.litName "_ancestor_any")
(Just "does array contain an ancestor of `ltree`?")
(ABackendSpecific . AAncestorAny . mkListLiteral columnType <$> columnListParser),
mkBoolOperator
collapseIfNull
$$(G.litName "_descendant")
(Just "is the left argument a descendant of right (or equal)?")
(ABackendSpecific . ADescendant . mkParameter <$> typedParser),
mkBoolOperator
collapseIfNull
$$(G.litName "_descendant_any")
(Just "does array contain a descendant of `ltree`?")
(ABackendSpecific . ADescendantAny . mkListLiteral columnType <$> columnListParser),
mkBoolOperator
collapseIfNull
$$(G.litName "_matches")
(Just "does `ltree` match `lquery`?")
(ABackendSpecific . AMatches . mkParameter <$> lqueryParser),
mkBoolOperator
collapseIfNull
$$(G.litName "_matches_any")
(Just "does `ltree` match any `lquery` in array?")
(ABackendSpecific . AMatchesAny . mkListLiteral (ColumnScalar PGLquery) <$> textListParser),
mkBoolOperator
collapseIfNull
$$(G.litName "_matches_fulltext")
(Just "does `ltree` match `ltxtquery`?")
(ABackendSpecific . AMatchesFulltext . mkParameter <$> ltxtqueryParser)
]
]
where
mkListLiteral :: ColumnType ('Postgres pgKind) -> [ColumnValue ('Postgres pgKind)] -> UnpreparedValue ('Postgres pgKind)
mkListLiteral columnType columnValues =
P.UVLiteral $
SETyAnn
(SEArray $ txtEncoder . cvValue <$> columnValues)
(mkTypeAnn $ CollectableTypeArray $ unsafePGColumnToBackend columnType)
castExp :: ColumnType ('Postgres pgKind) -> m (Maybe (Parser 'Input n (CastExp ('Postgres pgKind) (UnpreparedValue ('Postgres pgKind)))))
castExp sourceType = do
let maybeScalars = case sourceType of
ColumnScalar PGGeography -> Just (PGGeography, PGGeometry)
ColumnScalar PGGeometry -> Just (PGGeometry, PGGeography)
_ -> Nothing
forM maybeScalars $ \(sourceScalar, targetScalar) -> do
sourceName <- mkScalarTypeName sourceScalar <&> (<> $$(G.litName "_cast_exp"))
targetName <- mkScalarTypeName targetScalar
targetOpExps <- comparisonExps $ ColumnScalar targetScalar
let field = P.fieldOptional targetName Nothing $ (targetScalar,) <$> targetOpExps
Refactor type name customization Source typename customization (hasura/graphql-engine@aac64f2c81faa6a3aef4d0cf5fae97289ac4383e) introduced a mechanism to change certain names in the GraphQL schema that is exposed. In particular it allows last-minute modification of: 1. the names of some types, and 2. the names of some root fields. The above two items are assigned distinct customization algorithms, and at times both algorithms are in scope. So a need to distinguish them is needed. In the original design, this was addressed by introducing a newtype wrapper `Typename` around GraphQL `Name`s, dedicated to the names of types. However, in the majority of the codebase, type names are also represented by `Name`. For this reason, it was unavoidable to allow for easy conversion. This was supported by a `HasName Typename` instance, as well as by publishing the constructors of `Typename`. This means that the type safety that newtypes can add is lost. In particular, it is now very easy to confuse type name customization with root field name customization. This refactors the above design by instead introducing newtypes around the customization operations: ```haskell newtype MkTypename = MkTypename {runMkTypename :: Name -> Name} deriving (Semigroup, Monoid) via (Endo Name) newtype MkRootFieldName = MkRootFieldName {runMkRootFieldName :: Name -> Name} deriving (Semigroup, Monoid) via (Endo Name) ``` The `Monoid` instance allows easy composition of customization operations, piggybacking off of the type of `Endo`maps. This design allows safe co-existence of the two customization algorithms, while avoiding the syntactic overhead of packing and unpacking newtypes. PR-URL: https://github.com/hasura/graphql-engine-mono/pull/2989 GitOrigin-RevId: da3a353a9b003ee40c8d0a1e02872e99d2edd3ca
2021-11-30 12:51:46 +03:00
pure $ P.object sourceName Nothing $ M.fromList . maybeToList <$> field
geographyWithinDistanceInput ::
forall pgKind m n r.
(MonadSchema n m, MonadError QErr m, MonadReader r m, Has MkTypename r) =>
m (Parser 'Input n (DWithinGeogOp (UnpreparedValue ('Postgres pgKind))))
geographyWithinDistanceInput = do
geographyParser <- columnParser (ColumnScalar PGGeography) (G.Nullability False)
-- FIXME
-- It doesn't make sense for this value to be nullable; it only is for
-- backwards compatibility; if an explicit Null value is given, it will be
-- forwarded to the underlying SQL function, that in turns treat a null value
-- as an error. We can fix this by rejecting explicit null values, by marking
-- this field non-nullable in a future release.
booleanParser <- columnParser (ColumnScalar PGBoolean) (G.Nullability True)
floatParser <- columnParser (ColumnScalar PGFloat) (G.Nullability False)
pure $
Refactor type name customization Source typename customization (hasura/graphql-engine@aac64f2c81faa6a3aef4d0cf5fae97289ac4383e) introduced a mechanism to change certain names in the GraphQL schema that is exposed. In particular it allows last-minute modification of: 1. the names of some types, and 2. the names of some root fields. The above two items are assigned distinct customization algorithms, and at times both algorithms are in scope. So a need to distinguish them is needed. In the original design, this was addressed by introducing a newtype wrapper `Typename` around GraphQL `Name`s, dedicated to the names of types. However, in the majority of the codebase, type names are also represented by `Name`. For this reason, it was unavoidable to allow for easy conversion. This was supported by a `HasName Typename` instance, as well as by publishing the constructors of `Typename`. This means that the type safety that newtypes can add is lost. In particular, it is now very easy to confuse type name customization with root field name customization. This refactors the above design by instead introducing newtypes around the customization operations: ```haskell newtype MkTypename = MkTypename {runMkTypename :: Name -> Name} deriving (Semigroup, Monoid) via (Endo Name) newtype MkRootFieldName = MkRootFieldName {runMkRootFieldName :: Name -> Name} deriving (Semigroup, Monoid) via (Endo Name) ``` The `Monoid` instance allows easy composition of customization operations, piggybacking off of the type of `Endo`maps. This design allows safe co-existence of the two customization algorithms, while avoiding the syntactic overhead of packing and unpacking newtypes. PR-URL: https://github.com/hasura/graphql-engine-mono/pull/2989 GitOrigin-RevId: da3a353a9b003ee40c8d0a1e02872e99d2edd3ca
2021-11-30 12:51:46 +03:00
P.object $$(G.litName "st_d_within_geography_input") Nothing $
DWithinGeogOp <$> (mkParameter <$> P.field $$(G.litName "distance") Nothing floatParser)
<*> (mkParameter <$> P.field $$(G.litName "from") Nothing geographyParser)
<*> (mkParameter <$> P.fieldWithDefault $$(G.litName "use_spheroid") Nothing (G.VBoolean True) booleanParser)
geometryWithinDistanceInput ::
forall pgKind m n r.
(MonadSchema n m, MonadError QErr m, MonadReader r m, Has MkTypename r) =>
m (Parser 'Input n (DWithinGeomOp (UnpreparedValue ('Postgres pgKind))))
geometryWithinDistanceInput = do
geometryParser <- columnParser (ColumnScalar PGGeometry) (G.Nullability False)
floatParser <- columnParser (ColumnScalar PGFloat) (G.Nullability False)
pure $
Refactor type name customization Source typename customization (hasura/graphql-engine@aac64f2c81faa6a3aef4d0cf5fae97289ac4383e) introduced a mechanism to change certain names in the GraphQL schema that is exposed. In particular it allows last-minute modification of: 1. the names of some types, and 2. the names of some root fields. The above two items are assigned distinct customization algorithms, and at times both algorithms are in scope. So a need to distinguish them is needed. In the original design, this was addressed by introducing a newtype wrapper `Typename` around GraphQL `Name`s, dedicated to the names of types. However, in the majority of the codebase, type names are also represented by `Name`. For this reason, it was unavoidable to allow for easy conversion. This was supported by a `HasName Typename` instance, as well as by publishing the constructors of `Typename`. This means that the type safety that newtypes can add is lost. In particular, it is now very easy to confuse type name customization with root field name customization. This refactors the above design by instead introducing newtypes around the customization operations: ```haskell newtype MkTypename = MkTypename {runMkTypename :: Name -> Name} deriving (Semigroup, Monoid) via (Endo Name) newtype MkRootFieldName = MkRootFieldName {runMkRootFieldName :: Name -> Name} deriving (Semigroup, Monoid) via (Endo Name) ``` The `Monoid` instance allows easy composition of customization operations, piggybacking off of the type of `Endo`maps. This design allows safe co-existence of the two customization algorithms, while avoiding the syntactic overhead of packing and unpacking newtypes. PR-URL: https://github.com/hasura/graphql-engine-mono/pull/2989 GitOrigin-RevId: da3a353a9b003ee40c8d0a1e02872e99d2edd3ca
2021-11-30 12:51:46 +03:00
P.object $$(G.litName "st_d_within_input") Nothing $
DWithinGeomOp <$> (mkParameter <$> P.field $$(G.litName "distance") Nothing floatParser)
<*> (mkParameter <$> P.field $$(G.litName "from") Nothing geometryParser)
intersectsNbandGeomInput ::
forall pgKind m n r.
(MonadSchema n m, MonadError QErr m, MonadReader r m, Has MkTypename r) =>
m (Parser 'Input n (STIntersectsNbandGeommin (UnpreparedValue ('Postgres pgKind))))
intersectsNbandGeomInput = do
geometryParser <- columnParser (ColumnScalar PGGeometry) (G.Nullability False)
integerParser <- columnParser (ColumnScalar PGInteger) (G.Nullability False)
pure $
Refactor type name customization Source typename customization (hasura/graphql-engine@aac64f2c81faa6a3aef4d0cf5fae97289ac4383e) introduced a mechanism to change certain names in the GraphQL schema that is exposed. In particular it allows last-minute modification of: 1. the names of some types, and 2. the names of some root fields. The above two items are assigned distinct customization algorithms, and at times both algorithms are in scope. So a need to distinguish them is needed. In the original design, this was addressed by introducing a newtype wrapper `Typename` around GraphQL `Name`s, dedicated to the names of types. However, in the majority of the codebase, type names are also represented by `Name`. For this reason, it was unavoidable to allow for easy conversion. This was supported by a `HasName Typename` instance, as well as by publishing the constructors of `Typename`. This means that the type safety that newtypes can add is lost. In particular, it is now very easy to confuse type name customization with root field name customization. This refactors the above design by instead introducing newtypes around the customization operations: ```haskell newtype MkTypename = MkTypename {runMkTypename :: Name -> Name} deriving (Semigroup, Monoid) via (Endo Name) newtype MkRootFieldName = MkRootFieldName {runMkRootFieldName :: Name -> Name} deriving (Semigroup, Monoid) via (Endo Name) ``` The `Monoid` instance allows easy composition of customization operations, piggybacking off of the type of `Endo`maps. This design allows safe co-existence of the two customization algorithms, while avoiding the syntactic overhead of packing and unpacking newtypes. PR-URL: https://github.com/hasura/graphql-engine-mono/pull/2989 GitOrigin-RevId: da3a353a9b003ee40c8d0a1e02872e99d2edd3ca
2021-11-30 12:51:46 +03:00
P.object $$(G.litName "st_intersects_nband_geom_input") Nothing $
STIntersectsNbandGeommin <$> (mkParameter <$> P.field $$(G.litName "nband") Nothing integerParser)
<*> (mkParameter <$> P.field $$(G.litName "geommin") Nothing geometryParser)
intersectsGeomNbandInput ::
forall pgKind m n r.
(MonadSchema n m, MonadError QErr m, MonadReader r m, Has MkTypename r) =>
m (Parser 'Input n (STIntersectsGeomminNband (UnpreparedValue ('Postgres pgKind))))
intersectsGeomNbandInput = do
geometryParser <- columnParser (ColumnScalar PGGeometry) (G.Nullability False)
integerParser <- columnParser (ColumnScalar PGInteger) (G.Nullability False)
pure $
Refactor type name customization Source typename customization (hasura/graphql-engine@aac64f2c81faa6a3aef4d0cf5fae97289ac4383e) introduced a mechanism to change certain names in the GraphQL schema that is exposed. In particular it allows last-minute modification of: 1. the names of some types, and 2. the names of some root fields. The above two items are assigned distinct customization algorithms, and at times both algorithms are in scope. So a need to distinguish them is needed. In the original design, this was addressed by introducing a newtype wrapper `Typename` around GraphQL `Name`s, dedicated to the names of types. However, in the majority of the codebase, type names are also represented by `Name`. For this reason, it was unavoidable to allow for easy conversion. This was supported by a `HasName Typename` instance, as well as by publishing the constructors of `Typename`. This means that the type safety that newtypes can add is lost. In particular, it is now very easy to confuse type name customization with root field name customization. This refactors the above design by instead introducing newtypes around the customization operations: ```haskell newtype MkTypename = MkTypename {runMkTypename :: Name -> Name} deriving (Semigroup, Monoid) via (Endo Name) newtype MkRootFieldName = MkRootFieldName {runMkRootFieldName :: Name -> Name} deriving (Semigroup, Monoid) via (Endo Name) ``` The `Monoid` instance allows easy composition of customization operations, piggybacking off of the type of `Endo`maps. This design allows safe co-existence of the two customization algorithms, while avoiding the syntactic overhead of packing and unpacking newtypes. PR-URL: https://github.com/hasura/graphql-engine-mono/pull/2989 GitOrigin-RevId: da3a353a9b003ee40c8d0a1e02872e99d2edd3ca
2021-11-30 12:51:46 +03:00
P.object $$(G.litName "st_intersects_geom_nband_input") Nothing $
STIntersectsGeomminNband
<$> (mkParameter <$> P.field $$(G.litName "geommin") Nothing geometryParser)
<*> (fmap mkParameter <$> P.fieldOptional $$(G.litName "nband") Nothing integerParser)
countTypeInput ::
MonadParse n =>
Maybe (Parser 'Both n (Column ('Postgres pgKind))) ->
InputFieldsParser n (IR.CountDistinct -> CountType ('Postgres pgKind))
countTypeInput = \case
Just columnEnum -> do
columns <- P.fieldOptional $$(G.litName "columns") Nothing (P.list columnEnum)
pure $ flip mkCountType columns
Nothing -> pure $ flip mkCountType Nothing
where
mkCountType :: IR.CountDistinct -> Maybe [Column ('Postgres pgKind)] -> CountType ('Postgres pgKind)
mkCountType _ Nothing = PG.CTStar
mkCountType IR.SelectCountDistinct (Just cols) = PG.CTDistinct cols
mkCountType IR.SelectCountNonDistinct (Just cols) = PG.CTSimple cols
-- | Update operator that prepends a value to a column containing jsonb arrays.
--
-- Note: Currently this is Postgres specific because json columns have not been ported
-- to other backends yet.
prependOp ::
forall pgKind m n r.
( BackendSchema ('Postgres pgKind),
MonadReader r m,
MonadError QErr m,
MonadSchema n m,
Has MkTypename r
) =>
SU.UpdateOperator ('Postgres pgKind) m n (UnpreparedValue ('Postgres pgKind))
prependOp = SU.UpdateOperator {..}
where
updateOperatorApplicableColumn = isScalarColumnWhere (== PGJSONB) . ciType
updateOperatorParser tableGQLName tableName columns = do
let typedParser columnInfo =
fmap P.mkParameter
<$> BS.columnParser
(ciType columnInfo)
(G.Nullability $ ciIsNullable columnInfo)
desc = "prepend existing jsonb value of filtered columns with new jsonb value"
SU.updateOperator
tableGQLName
$$(G.litName "_prepend")
typedParser
columns
desc
desc
-- | Update operator that appends a value to a column containing jsonb arrays.
--
-- Note: Currently this is Postgres specific because json columns have not been ported
-- to other backends yet.
appendOp ::
forall pgKind m n r.
( BackendSchema ('Postgres pgKind),
MonadReader r m,
MonadError QErr m,
MonadSchema n m,
Has MkTypename r
) =>
SU.UpdateOperator ('Postgres pgKind) m n (UnpreparedValue ('Postgres pgKind))
appendOp = SU.UpdateOperator {..}
where
updateOperatorApplicableColumn = isScalarColumnWhere (== PGJSONB) . ciType
updateOperatorParser tableGQLName tableName columns = do
let typedParser columnInfo =
fmap P.mkParameter
<$> BS.columnParser
(ciType columnInfo)
(G.Nullability $ ciIsNullable columnInfo)
desc = "append existing jsonb value of filtered columns with new jsonb value"
SU.updateOperator
tableGQLName
$$(G.litName "_append")
typedParser
columns
desc
desc
-- | Update operator that deletes a value at a specified key from a column
-- containing jsonb objects.
--
-- Note: Currently this is Postgres specific because json columns have not been ported
-- to other backends yet.
deleteKeyOp ::
forall pgKind m n r.
( BackendSchema ('Postgres pgKind),
MonadReader r m,
MonadError QErr m,
MonadSchema n m,
Has MkTypename r
) =>
SU.UpdateOperator ('Postgres pgKind) m n (UnpreparedValue ('Postgres pgKind))
deleteKeyOp = SU.UpdateOperator {..}
where
updateOperatorApplicableColumn = isScalarColumnWhere (== PGJSONB) . ciType
updateOperatorParser tableGQLName tableName columns = do
let nullableTextParser _ = fmap P.mkParameter <$> columnParser (ColumnScalar PGText) (G.Nullability True)
desc = "delete key/value pair or string element. key/value pairs are matched based on their key value"
SU.updateOperator
tableGQLName
$$(G.litName "_delete_key")
nullableTextParser
columns
desc
desc
-- | Update operator that deletes a value at a specific index from a column
-- containing jsonb arrays.
--
-- Note: Currently this is Postgres specific because json columns have not been ported
-- to other backends yet.
deleteElemOp ::
forall pgKind m n r.
( BackendSchema ('Postgres pgKind),
MonadReader r m,
MonadError QErr m,
MonadSchema n m,
Has MkTypename r
) =>
SU.UpdateOperator ('Postgres pgKind) m n (UnpreparedValue ('Postgres pgKind))
deleteElemOp = SU.UpdateOperator {..}
where
updateOperatorApplicableColumn = isScalarColumnWhere (== PGJSONB) . ciType
updateOperatorParser tableGQLName tableName columns = do
let nonNullableIntParser _ = fmap P.mkParameter <$> columnParser (ColumnScalar PGInteger) (G.Nullability False)
desc =
"delete the array element with specified index (negative integers count from the end). "
<> "throws an error if top level container is not an array"
SU.updateOperator
tableGQLName
$$(G.litName "_delete_elem")
nonNullableIntParser
columns
desc
desc
-- | Update operator that deletes a field at a certan path from a column
-- containing jsonb objects.
--
-- Note: Currently this is Postgres specific because json columns have not been ported
-- to other backends yet.
deleteAtPathOp ::
forall pgKind m n r.
( BackendSchema ('Postgres pgKind),
MonadReader r m,
MonadError QErr m,
MonadSchema n m,
Has MkTypename r
) =>
SU.UpdateOperator ('Postgres pgKind) m n [UnpreparedValue ('Postgres pgKind)]
deleteAtPathOp = SU.UpdateOperator {..}
where
updateOperatorApplicableColumn = isScalarColumnWhere (== PGJSONB) . ciType
updateOperatorParser tableGQLName tableName columns = do
let nonNullableTextListParser _ = P.list . fmap P.mkParameter <$> columnParser (ColumnScalar PGText) (G.Nullability False)
desc = "delete the field or element with specified path (for JSON arrays, negative integers count from the end)"
SU.updateOperator
tableGQLName
$$(G.litName "_delete_at_path")
nonNullableTextListParser
columns
desc
desc
-- | The update operators that we support on Postgres.
updateOperators ::
forall pgKind m n r.
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
MonadBuildSchema ('Postgres pgKind) r m n =>
TableInfo ('Postgres pgKind) ->
UpdPermInfo ('Postgres pgKind) ->
m (InputFieldsParser n (HashMap (Column ('Postgres pgKind)) (UpdateOpExpression (UnpreparedValue ('Postgres pgKind)))))
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
updateOperators tableInfo updatePermissions = do
SU.buildUpdateOperators
(PGIR.UpdateSet <$> SU.presetColumns updatePermissions)
[ PGIR.UpdateSet <$> SU.setOp,
PGIR.UpdateInc <$> SU.incOp,
PGIR.UpdatePrepend <$> prependOp,
PGIR.UpdateAppend <$> appendOp,
PGIR.UpdateDeleteKey <$> deleteKeyOp,
PGIR.UpdateDeleteElem <$> deleteElemOp,
PGIR.UpdateDeleteAtPath <$> deleteAtPathOp
]
tableInfo