graphql-engine/server/src-lib/Hasura/SQL/Backend.hs

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

244 lines
9.4 KiB
Haskell
Raw Normal View History

{-# LANGUAGE DeriveAnyClass #-}
{-# LANGUAGE QuasiQuotes #-}
module Hasura.SQL.Backend
( PostgresKind (..),
BackendType (..),
BackendSourceKind (..),
backendShortName,
supportedBackends,
backendTextNames,
backendTypeFromText,
parseBackendTypeFromText,
backendTypeFromBackendSourceKind,
)
where
import Autodocodec (Codec (StringCodec), HasCodec (codec), JSONCodec, bimapCodec, literalTextCodec, parseAlternatives, (<?>))
import Data.Aeson hiding ((<?>))
import Data.Aeson.Types (Parser)
import Data.Text (unpack)
import Data.Text.Extended
import Data.Text.NonEmpty (NonEmptyText, nonEmptyTextQQ)
import Hasura.Backends.DataConnector.Adapter.Types (DataConnectorName (..), mkDataConnectorName)
import Hasura.Prelude
import Language.GraphQL.Draft.Syntax qualified as GQL
import Witch qualified
-- | Argument to Postgres; we represent backends which are variations on Postgres as sub-types of
-- Postgres. This value indicates which "flavour" of Postgres a backend is.
data PostgresKind
= Vanilla
| Citus
| Cockroach
deriving stock (Show, Eq, Ord, Generic)
server: delete the `Cacheable` type class in favor of `Eq` What is the `Cacheable` type class about? ```haskell class Eq a => Cacheable a where unchanged :: Accesses -> a -> a -> Bool default unchanged :: (Generic a, GCacheable (Rep a)) => Accesses -> a -> a -> Bool unchanged accesses a b = gunchanged (from a) (from b) accesses ``` Its only method is an alternative to `(==)`. The added value of `unchanged` (and the additional `Accesses` argument) arises _only_ for one type, namely `Dependency`. Indeed, the `Cacheable (Dependency a)` instance is non-trivial, whereas every other `Cacheable` instance is completely boilerplate (and indeed either generated from `Generic`, or simply `unchanged _ = (==)`). The `Cacheable (Dependency a)` instance is the only one where the `Accesses` argument is not just passed onwards. The only callsite of the `unchanged` method is in the `ArrowCache (Rule m)` method. That is to say that the `Cacheable` type class is used to decide when we can re-use parts of the schema cache between Metadata operations. So what is the `Cacheable (Dependency a)` instance about? Normally, the output of a `Rule m a b` is re-used when the new input (of type `a`) is equal to the old one. But sometimes, that's too coarse: it might be that a certain `Rule m a b` only depends on a small part of its input of type `a`. A `Dependency` allows us to spell out what parts of `a` are being depended on, and these parts are recorded as values of types `Access a` in the state `Accesses`. If the input `a` changes, but not in a way that touches the recorded `Accesses`, then the output `b` of that rule can be re-used without recomputing. So now you understand _why_ we're passing `Accesses` to the `unchanged` method: `unchanged` is an equality check in disguise that just needs some additional context. But we don't need to pass `Accesses` as a function argument. We can use the `reflection` package to pass it as type-level context. So the core of this PR is that we change the instance declaration from ```haskell instance (Cacheable a) => Cacheable (Dependency a) where ``` to ```haskell instance (Given Accesses, Eq a) => Eq (Dependency a) where ``` and use `(==)` instead of `unchanged`. If you haven't seen `reflection` before: it's like a `MonadReader`, but it doesn't require a `Monad`. In order to pass the current `Accesses` value, instead of simply passing the `Accesses` as a function argument, we need to instantiate the `Given Accesses` context. We use the `give` method from the `reflection` package for that. ```haskell give :: forall r. Accesses -> (Given Accesses => r) -> r unchanged :: (Given Accesses => Eq a) => Accesses -> a -> a -> Bool unchanged accesses a b = give accesses (a == b) ``` With these three components in place, we can delete the `Cacheable` type class entirely. The remainder of this PR is just to remove the `Cacheable` type class and its instances. PR-URL: https://github.com/hasura/graphql-engine-mono/pull/6877 GitOrigin-RevId: 7125f5e11d856e7672ab810a23d5bf5ad176e77f
2022-11-21 19:33:56 +03:00
deriving anyclass (Hashable)
-- | An enum that represents each backend we support.
data BackendType
= Postgres PostgresKind
| MSSQL
| BigQuery
| MySQL
| DataConnector
deriving stock (Show, Eq, Ord, Generic)
server: delete the `Cacheable` type class in favor of `Eq` What is the `Cacheable` type class about? ```haskell class Eq a => Cacheable a where unchanged :: Accesses -> a -> a -> Bool default unchanged :: (Generic a, GCacheable (Rep a)) => Accesses -> a -> a -> Bool unchanged accesses a b = gunchanged (from a) (from b) accesses ``` Its only method is an alternative to `(==)`. The added value of `unchanged` (and the additional `Accesses` argument) arises _only_ for one type, namely `Dependency`. Indeed, the `Cacheable (Dependency a)` instance is non-trivial, whereas every other `Cacheable` instance is completely boilerplate (and indeed either generated from `Generic`, or simply `unchanged _ = (==)`). The `Cacheable (Dependency a)` instance is the only one where the `Accesses` argument is not just passed onwards. The only callsite of the `unchanged` method is in the `ArrowCache (Rule m)` method. That is to say that the `Cacheable` type class is used to decide when we can re-use parts of the schema cache between Metadata operations. So what is the `Cacheable (Dependency a)` instance about? Normally, the output of a `Rule m a b` is re-used when the new input (of type `a`) is equal to the old one. But sometimes, that's too coarse: it might be that a certain `Rule m a b` only depends on a small part of its input of type `a`. A `Dependency` allows us to spell out what parts of `a` are being depended on, and these parts are recorded as values of types `Access a` in the state `Accesses`. If the input `a` changes, but not in a way that touches the recorded `Accesses`, then the output `b` of that rule can be re-used without recomputing. So now you understand _why_ we're passing `Accesses` to the `unchanged` method: `unchanged` is an equality check in disguise that just needs some additional context. But we don't need to pass `Accesses` as a function argument. We can use the `reflection` package to pass it as type-level context. So the core of this PR is that we change the instance declaration from ```haskell instance (Cacheable a) => Cacheable (Dependency a) where ``` to ```haskell instance (Given Accesses, Eq a) => Eq (Dependency a) where ``` and use `(==)` instead of `unchanged`. If you haven't seen `reflection` before: it's like a `MonadReader`, but it doesn't require a `Monad`. In order to pass the current `Accesses` value, instead of simply passing the `Accesses` as a function argument, we need to instantiate the `Given Accesses` context. We use the `give` method from the `reflection` package for that. ```haskell give :: forall r. Accesses -> (Given Accesses => r) -> r unchanged :: (Given Accesses => Eq a) => Accesses -> a -> a -> Bool unchanged accesses a b = give accesses (a == b) ``` With these three components in place, we can delete the `Cacheable` type class entirely. The remainder of this PR is just to remove the `Cacheable` type class and its instances. PR-URL: https://github.com/hasura/graphql-engine-mono/pull/6877 GitOrigin-RevId: 7125f5e11d856e7672ab810a23d5bf5ad176e77f
2022-11-21 19:33:56 +03:00
deriving anyclass (Hashable)
-- | The name of the backend, as we expect it to appear in our metadata and API.
instance Witch.From BackendType NonEmptyText where
from (Postgres Vanilla) = [nonEmptyTextQQ|postgres|]
from (Postgres Citus) = [nonEmptyTextQQ|citus|]
from (Postgres Cockroach) = [nonEmptyTextQQ|cockroach|]
from MSSQL = [nonEmptyTextQQ|mssql|]
from BigQuery = [nonEmptyTextQQ|bigquery|]
from MySQL = [nonEmptyTextQQ|mysql|]
from DataConnector = [nonEmptyTextQQ|dataconnector|]
instance ToTxt BackendType where
toTxt = toTxt . Witch.into @NonEmptyText
instance FromJSON BackendType where
parseJSON = withText "backend type" parseBackendTypeFromText
instance ToJSON BackendType where
toJSON = String . toTxt
-- | Similar to 'BackendType', however, in the case of 'DataConnectorKind' we need to be able
-- capture the name of the data connector that should be used by the DataConnector backend.
-- This type correlates to the kind property of 'SourceMetadata', which is usually just
-- postgres, mssql, etc for static backends, but can be a configurable value for DataConnector
-- hence requiring 'DataConnectorName' for 'DataConnectorKind'
--
-- This type cannot entirely replace 'BackendType' because 'BackendType' has a fixed number of
-- possible values which can be enumerated over at compile time, but 'BackendSourceKind' does
-- not because DataConnector fundamentally is configured at runtime with 'DataConnectorName'.
data BackendSourceKind (b :: BackendType) where
PostgresVanillaKind :: BackendSourceKind ('Postgres 'Vanilla)
PostgresCitusKind :: BackendSourceKind ('Postgres 'Citus)
PostgresCockroachKind :: BackendSourceKind ('Postgres 'Cockroach)
MSSQLKind :: BackendSourceKind 'MSSQL
BigQueryKind :: BackendSourceKind 'BigQuery
MySQLKind :: BackendSourceKind 'MySQL
DataConnectorKind :: DataConnectorName -> BackendSourceKind 'DataConnector
deriving instance Show (BackendSourceKind b)
deriving instance Eq (BackendSourceKind b)
deriving instance Ord (BackendSourceKind b)
instance Witch.From (BackendSourceKind b) NonEmptyText where
-- All cases are specified explicitly here to ensure compiler warnings highlight
-- this area for consideration and update if another BackendType is added
from k@PostgresVanillaKind = Witch.into @NonEmptyText $ backendTypeFromBackendSourceKind k
from k@PostgresCitusKind = Witch.into @NonEmptyText $ backendTypeFromBackendSourceKind k
from k@PostgresCockroachKind = Witch.into @NonEmptyText $ backendTypeFromBackendSourceKind k
from k@MSSQLKind = Witch.into @NonEmptyText $ backendTypeFromBackendSourceKind k
from k@BigQueryKind = Witch.into @NonEmptyText $ backendTypeFromBackendSourceKind k
from k@MySQLKind = Witch.into @NonEmptyText $ backendTypeFromBackendSourceKind k
from (DataConnectorKind dataConnectorName) = Witch.into @NonEmptyText dataConnectorName
instance ToTxt (BackendSourceKind b) where
toTxt = toTxt . Witch.into @NonEmptyText
-- If you need to parse an arbitrary string into a BackendSourceKind, you can't because of the
-- b type parameter. You actually want to parse into 'AnyBackend BackendSourceKind'.
-- See 'backendSourceKindFromText' from the AnyBackend module for that.
instance ToJSON (BackendSourceKind b) where
toJSON = String . toTxt
instance FromJSON (BackendSourceKind ('Postgres 'Vanilla)) where
parseJSON = mkParseStaticBackendSourceKind PostgresVanillaKind
instance FromJSON (BackendSourceKind ('Postgres 'Citus)) where
parseJSON = mkParseStaticBackendSourceKind PostgresCitusKind
instance FromJSON (BackendSourceKind ('Postgres 'Cockroach)) where
parseJSON = mkParseStaticBackendSourceKind PostgresCockroachKind
instance FromJSON (BackendSourceKind ('MSSQL)) where
parseJSON = mkParseStaticBackendSourceKind MSSQLKind
instance FromJSON (BackendSourceKind ('BigQuery)) where
parseJSON = mkParseStaticBackendSourceKind BigQueryKind
instance FromJSON (BackendSourceKind ('MySQL)) where
parseJSON = mkParseStaticBackendSourceKind MySQLKind
instance FromJSON (BackendSourceKind ('DataConnector)) where
parseJSON v = DataConnectorKind <$> parseJSON v
mkParseStaticBackendSourceKind :: BackendSourceKind b -> (Value -> Parser (BackendSourceKind b))
mkParseStaticBackendSourceKind backendSourceKind =
withText "BackendSourceKind" $ \text ->
if text `elem` validValues
then pure backendSourceKind
else fail ("got: " <> unpack text <> ", expected one of: " <> unpack (commaSeparated validValues))
where
validValues = backendTextNames $ backendTypeFromBackendSourceKind backendSourceKind
server: polymorphic codec for metadata sources This PR expands the OpenAPI specification generated for metadata to include separate definitions for `SourceMetadata` for each native database type, and for DataConnector. For the most part the changes add `HasCodec` implementations, and don't modify existing code otherwise. The generated OpenAPI spec can be used to generate TypeScript definitions that distinguish different source metadata types based on the value of the `kind` properly. There is a problem: because the specified `kind` value for a data connector source is any string, when TypeScript gets a source with a `kind` value of, say, `"postgres"`, it cannot unambiguously determine whether the source is postgres, or a data connector. For example, ```ts function consumeSourceMetadata(source: SourceMetadata) { if (source.kind === "postgres" || source.kind === "pg") { // At this point TypeScript infers that `source` is either an instance // of `PostgresSourceMetadata`, or `DataconnectorSourceMetadata`. It // can't narrow further. source } if (source.kind === "something else") { // TypeScript infers that this `source` must be an instance of // `DataconnectorSourceMetadata` because `source.kind` does not match // any of the other options. source } } ``` The simplest way I can think of to fix this would be to add a boolean property to the `SourceMetadata` type along the lines of `isNative` or `isDataConnector`. This could be a field that only exists in serialized data, like the metadata version field. The combination of one of the native database names for `kind`, and a true value for `isNative` would be enough for TypeScript to unambiguously distinguish the source kinds. But note that in the current state TypeScript is able to reference the short `"pg"` name correctly! ~~Tests are not passing yet due to some discrepancies in DTO serialization vs existing Metadata serialization. I'm working on that.~~ The placeholders that I used for table and function metadata are not compatible with the ordered JSON serialization in use. I think the best solution is to write compatible codecs for those types in another PR. For now I have disabled some DTO tests for this PR. Here are the generated [OpenAPI spec](https://github.com/hasura/graphql-engine-mono/files/9397333/openapi.tar.gz) based on these changes, and the generated [TypeScript client code](https://github.com/hasura/graphql-engine-mono/files/9397339/client-typescript.tar.gz) based on that spec. Ticket: [MM-66](https://hasurahq.atlassian.net/browse/MM-66) PR-URL: https://github.com/hasura/graphql-engine-mono/pull/5582 GitOrigin-RevId: e1446191c6c832879db04f129daa397a3be03f62
2022-08-25 21:34:44 +03:00
instance HasCodec (BackendSourceKind ('Postgres 'Vanilla)) where
codec = mkCodecStaticBackendSourceKind PostgresVanillaKind
instance HasCodec (BackendSourceKind ('Postgres 'Citus)) where
codec = mkCodecStaticBackendSourceKind PostgresCitusKind
instance HasCodec (BackendSourceKind ('Postgres 'Cockroach)) where
codec = mkCodecStaticBackendSourceKind PostgresCockroachKind
instance HasCodec (BackendSourceKind ('MSSQL)) where
codec = mkCodecStaticBackendSourceKind MSSQLKind
instance HasCodec (BackendSourceKind ('BigQuery)) where
codec = mkCodecStaticBackendSourceKind BigQueryKind
instance HasCodec (BackendSourceKind ('MySQL)) where
codec = mkCodecStaticBackendSourceKind MySQLKind
instance HasCodec (BackendSourceKind ('DataConnector)) where
codec = bimapCodec dec enc gqlNameCodec
server: polymorphic codec for metadata sources This PR expands the OpenAPI specification generated for metadata to include separate definitions for `SourceMetadata` for each native database type, and for DataConnector. For the most part the changes add `HasCodec` implementations, and don't modify existing code otherwise. The generated OpenAPI spec can be used to generate TypeScript definitions that distinguish different source metadata types based on the value of the `kind` properly. There is a problem: because the specified `kind` value for a data connector source is any string, when TypeScript gets a source with a `kind` value of, say, `"postgres"`, it cannot unambiguously determine whether the source is postgres, or a data connector. For example, ```ts function consumeSourceMetadata(source: SourceMetadata) { if (source.kind === "postgres" || source.kind === "pg") { // At this point TypeScript infers that `source` is either an instance // of `PostgresSourceMetadata`, or `DataconnectorSourceMetadata`. It // can't narrow further. source } if (source.kind === "something else") { // TypeScript infers that this `source` must be an instance of // `DataconnectorSourceMetadata` because `source.kind` does not match // any of the other options. source } } ``` The simplest way I can think of to fix this would be to add a boolean property to the `SourceMetadata` type along the lines of `isNative` or `isDataConnector`. This could be a field that only exists in serialized data, like the metadata version field. The combination of one of the native database names for `kind`, and a true value for `isNative` would be enough for TypeScript to unambiguously distinguish the source kinds. But note that in the current state TypeScript is able to reference the short `"pg"` name correctly! ~~Tests are not passing yet due to some discrepancies in DTO serialization vs existing Metadata serialization. I'm working on that.~~ The placeholders that I used for table and function metadata are not compatible with the ordered JSON serialization in use. I think the best solution is to write compatible codecs for those types in another PR. For now I have disabled some DTO tests for this PR. Here are the generated [OpenAPI spec](https://github.com/hasura/graphql-engine-mono/files/9397333/openapi.tar.gz) based on these changes, and the generated [TypeScript client code](https://github.com/hasura/graphql-engine-mono/files/9397339/client-typescript.tar.gz) based on that spec. Ticket: [MM-66](https://hasurahq.atlassian.net/browse/MM-66) PR-URL: https://github.com/hasura/graphql-engine-mono/pull/5582 GitOrigin-RevId: e1446191c6c832879db04f129daa397a3be03f62
2022-08-25 21:34:44 +03:00
where
dec :: GQL.Name -> Either String (BackendSourceKind 'DataConnector)
dec n = DataConnectorKind <$> mkDataConnectorName n
enc :: BackendSourceKind ('DataConnector) -> GQL.Name
enc (DataConnectorKind dcName) = unDataConnectorName dcName
gqlNameCodec :: JSONCodec GQL.Name
gqlNameCodec =
bimapCodec
parseName
GQL.unName
(StringCodec (Just "GraphQLName"))
<?> "A valid GraphQL name"
parseName text = GQL.mkName text `onNothing` Left (unpack text <> " is not a valid GraphQL name")
server: polymorphic codec for metadata sources This PR expands the OpenAPI specification generated for metadata to include separate definitions for `SourceMetadata` for each native database type, and for DataConnector. For the most part the changes add `HasCodec` implementations, and don't modify existing code otherwise. The generated OpenAPI spec can be used to generate TypeScript definitions that distinguish different source metadata types based on the value of the `kind` properly. There is a problem: because the specified `kind` value for a data connector source is any string, when TypeScript gets a source with a `kind` value of, say, `"postgres"`, it cannot unambiguously determine whether the source is postgres, or a data connector. For example, ```ts function consumeSourceMetadata(source: SourceMetadata) { if (source.kind === "postgres" || source.kind === "pg") { // At this point TypeScript infers that `source` is either an instance // of `PostgresSourceMetadata`, or `DataconnectorSourceMetadata`. It // can't narrow further. source } if (source.kind === "something else") { // TypeScript infers that this `source` must be an instance of // `DataconnectorSourceMetadata` because `source.kind` does not match // any of the other options. source } } ``` The simplest way I can think of to fix this would be to add a boolean property to the `SourceMetadata` type along the lines of `isNative` or `isDataConnector`. This could be a field that only exists in serialized data, like the metadata version field. The combination of one of the native database names for `kind`, and a true value for `isNative` would be enough for TypeScript to unambiguously distinguish the source kinds. But note that in the current state TypeScript is able to reference the short `"pg"` name correctly! ~~Tests are not passing yet due to some discrepancies in DTO serialization vs existing Metadata serialization. I'm working on that.~~ The placeholders that I used for table and function metadata are not compatible with the ordered JSON serialization in use. I think the best solution is to write compatible codecs for those types in another PR. For now I have disabled some DTO tests for this PR. Here are the generated [OpenAPI spec](https://github.com/hasura/graphql-engine-mono/files/9397333/openapi.tar.gz) based on these changes, and the generated [TypeScript client code](https://github.com/hasura/graphql-engine-mono/files/9397339/client-typescript.tar.gz) based on that spec. Ticket: [MM-66](https://hasurahq.atlassian.net/browse/MM-66) PR-URL: https://github.com/hasura/graphql-engine-mono/pull/5582 GitOrigin-RevId: e1446191c6c832879db04f129daa397a3be03f62
2022-08-25 21:34:44 +03:00
mkCodecStaticBackendSourceKind :: BackendSourceKind b -> JSONCodec (BackendSourceKind b)
mkCodecStaticBackendSourceKind backendSourceKind =
bimapCodec dec enc $
parseAlternatives (literalTextCodec longName) (literalTextCodec <$> aliases)
where
dec text =
if text `elem` validValues
then Right backendSourceKind
else Left ("got: " <> unpack text <> ", expected one of: " <> unpack (commaSeparated validValues))
enc = toTxt
validValues = backendTextNames $ backendTypeFromBackendSourceKind backendSourceKind
longName = head validValues
aliases = tail validValues
-- | Some generated APIs use a shortened version of the backend's name rather than its full
-- name. This function returns the "short form" of a backend, if any.
backendShortName :: BackendType -> Maybe Text
backendShortName = \case
Postgres Vanilla -> Just "pg"
_ -> Nothing
supportedBackends :: [BackendType]
supportedBackends =
[ Postgres Vanilla,
Postgres Citus,
Postgres Cockroach,
MSSQL,
BigQuery,
MySQL,
DataConnector
]
backendTextNames :: BackendType -> [Text]
backendTextNames b =
catMaybes
[ Just (toTxt b), -- long form
backendShortName b -- short form
]
backendTextNameLookup :: [(Text, BackendType)]
backendTextNameLookup =
supportedBackends >>= (\b -> (,b) <$> backendTextNames b)
-- | This uses this lookup mechanism to avoid having to duplicate and hardcode the
-- backend string. We accept both the short form and the long form of the backend's name.
backendTypeFromText :: Text -> Maybe BackendType
backendTypeFromText txt =
lookup txt backendTextNameLookup
parseBackendTypeFromText :: Text -> Parser BackendType
parseBackendTypeFromText txt =
let uniqueBackends = commaSeparated $ fst <$> backendTextNameLookup
in backendTypeFromText txt
`onNothing` fail ("got: " <> unpack txt <> ", expected one of: " <> unpack uniqueBackends)
backendTypeFromBackendSourceKind :: BackendSourceKind b -> BackendType
backendTypeFromBackendSourceKind = \case
PostgresVanillaKind -> Postgres Vanilla
PostgresCitusKind -> Postgres Citus
PostgresCockroachKind -> Postgres Cockroach
MSSQLKind -> MSSQL
BigQueryKind -> BigQuery
MySQLKind -> MySQL
DataConnectorKind _ -> DataConnector