graphql-engine/server/src-lib/Hasura/GraphQL/Schema/Backend.hs
Auke Booij 1007ea27ae server: refactor MonadSchema into MonadMemoize
Followup to hasura/graphql-engine-mono#4713.

The `memoizeOn` method, part of `MonadSchema`, originally had the following type:
```haskell
  memoizeOn
    :: (HasCallStack, Ord a, Typeable a, Typeable b, Typeable k)
    => TH.Name
    -> a
    -> m (Parser k n b)
    -> m (Parser k n b)
```
The reason for operating on `Parser`s specifically was that the `MonadSchema` effect would additionally initialize certain `Unique` values, which appear (nested in) the type of `Parser`.

hasura/graphql-engine-mono#518 changed the type of `memoizeOn`, to additionally allow memoizing `FieldParser`s. These also contained a `Unique` value, which was similarly initialized by the `MonadSchema` effect. The new type of `memoizeOn` was as follows:
```haskell
  memoizeOn
    :: forall p d a b
     . (HasCallStack, HasDefinition (p n b) d, Ord a, Typeable p, Typeable a, Typeable b)
    => TH.Name
    -> a
    -> m (p n b)
    -> m (p n b)
```

Note the type `p n b` of the value being memoized: by choosing `p` to be either `Parser k` or `FieldParser`, both can be memoized. Also note the new `HasDefinition (p n b) d` constraint, which provided a `Lens` for accessing the `Unique` value to be initialized.

A quick simplification is that the `HasCallStack` constraint has never been used by any code. This was realized in hasura/graphql-engine-mono#4713, by removing that constraint.

hasura/graphql-engine-mono#2980 removed the `Unique` value from our GraphQL-related types entirely, as their original purpose was never truly realized. One part of removing `Unique` consisted of dropping the `HasDefinition (p n b) d` constraint from `memoizeOn`.

What I didn't realize at the time was that this meant that the type of `memoizeOn` could be generalized and simplified much further. This PR finally implements that generalization. The new type is as follows:
```haskell
  memoizeOn ::
    forall a p.
    (Ord a, Typeable a, Typeable p) =>
    TH.Name ->
    a ->
    m p ->
    m p
```

This change has a couple of consequences.

1. While constructing the schema, we often output `Maybe (Parser ...)`, to model that the existence of certain pieces of GraphQL schema sometimes depends on the permissions that a certain role has. The previous versions of `memoizeOn` were not able to handle this, as the only thing they could memoize was fully-defined (if not yet fully-evaluated) `(Field)Parser`s. This much more general API _would_ allow memoizing `Maybe (Parser ...)`s. However, we probably have to be continue being cautious with this: if we blindly memoize all `Maybe (Parser ...)`s, the resulting code may never be able to decide whether the value is `Just` or `Nothing` - i.e. it never commits to the existence-or-not of a GraphQL schema fragment. This would manifest as a non-well-founded knot tying, and this would get reported as an error by the implementation of `memoizeOn`.

   tl;dr: This generalization _technically_ allows for memoizing `Maybe` values, but we probably still want to avoid doing so.

   For this reason, the PR adds a specialized version of `memoizeOn` to `Hasura.GraphQL.Schema.Parser`.
2. There is no longer any need to connect the `MonadSchema` knot-tying effect with the `MonadParse` effect. In fact, after this PR, the `memoizeOn` method is completely GraphQL-agnostic, and so we implement hasura/graphql-engine-mono#4726, separating `memoizeOn` from `MonadParse` entirely - `memoizeOn` can be defined and implemented as a general Haskell typeclass method.

   Since `MonadSchema` has been made into a single-type-parameter type class, it has been renamed to something more general, namely `MonadMemoize`. Its only task is to memoize arbitrary `Typeable p` objects under a combined key consisting of a `TH.Name` and a `Typeable a`.

   Also for this reason, the new `MonadMemoize` has been moved to the more general `Control.Monad.Memoize`.
3. After this change, it's somewhat clearer what `memoizeOn` does: it memoizes an arbitrary value of a `Typeable` type. The only thing that needs to be understood in its implementation is how the manual blackholing works. There is no more semantic interaction with _any_ GraphQL code.

PR-URL: https://github.com/hasura/graphql-engine-mono/pull/4725
Co-authored-by: Daniel Harvey <4729125+danieljharvey@users.noreply.github.com>
GitOrigin-RevId: 089fa2e82c2ce29da76850e994eabb1e261f9c92
2022-08-04 13:45:53 +00:00

372 lines
13 KiB
Haskell

-- | This module defines the type class 'BackendSchema' and auxiliary types.
--
-- 'BackendSchema' represents the part of the interface that a backend driver
-- presents to the GraphQL Engine core that is responsible for generating
-- the backend's Schema Parsers.
--
-- The Schema Parsers recognise (and reflect) the schema that a backend exposes.
--
-- The 'BackendSchema' methods are used by
-- 'Hasura.GraphQL.Schema.buildGQLContext', which is the core's entrypoint to
-- schema generation.
--
-- Many of the 'BackendSchema' methods will have default implementations that a
-- backend driver may use. These may be found (chiefly) in the modules:
--
-- * The module "Hasura.GraphQL.Schema.Build", commonly qualified @GSB@
-- * "Hasura.GraphQL.Schema.Select", commonly qualified @GSS@
-- * "Hasura.GraphQL.Schema.BoolExp"
--
-- For more information see:
--
-- * <https://github.com/hasura/graphql-engine/blob/master/server/documentation/schema.md Technical overview of Schema Generation >
-- * The type 'Hasura.GraphQL.Parser.Parser', and associated source code notes
-- in the same folder (not exposed with Haddock unfortunately)
module Hasura.GraphQL.Schema.Backend
( -- * Main Types
BackendSchema (..),
BackendTableSelectSchema (..),
MonadBuildSchema,
-- * Auxiliary Types
ComparisonExp,
-- * Note: @BackendSchema@ modelling principles
-- $modelling
)
where
import Data.Has
import Data.Text.Casing (GQLNameIdentifier)
import Hasura.Base.Error
import Hasura.GraphQL.ApolloFederation (ApolloFederationParserFunction)
import Hasura.GraphQL.Schema.Common
import Hasura.GraphQL.Schema.NamingCase
import Hasura.GraphQL.Schema.Parser hiding (Type)
import Hasura.GraphQL.Schema.Typename
import Hasura.Prelude
import Hasura.RQL.IR
import Hasura.RQL.IR.Insert qualified as IR
import Hasura.RQL.IR.Select qualified as IR
import Hasura.RQL.Types.Backend
import Hasura.RQL.Types.Column hiding (EnumValueInfo)
import Hasura.RQL.Types.ComputedField
import Hasura.RQL.Types.Function
import Hasura.RQL.Types.Relationships.Local
import Hasura.RQL.Types.SchemaCache
import Hasura.RQL.Types.Source
import Hasura.RQL.Types.SourceCustomization (MkRootFieldName)
import Hasura.SQL.Backend
import Hasura.Server.Types (StreamingSubscriptionsCtx)
import Language.GraphQL.Draft.Syntax qualified as G
-- | Bag of constraints available to the methods of @BackendSchema@.
--
-- Note that @BackendSchema b@ is itself part of this, so a methods may also
-- call other methods. This might seem trivial, but it can be easy to miss when
-- the functions used to implement a class instance are defined in multiple
-- modules.
type MonadBuildSchema b r m n =
( BackendSchema b,
MonadBuildSchemaBase r m n
)
-- | This type class is responsible for generating the schema of a backend.
-- Its methods are called by the common core that orchestrates the various
-- backend drivers.
--
-- Its purpose in life is to make it convenient to express the GraphQL schema we
-- want to expose for the backends that we support. This means balancing the
-- desire to have consistency with the desire to differentiate the schema of a
-- backend.
--
-- This means that it is expected to evolve over time as we add new backends,
-- and that you have the license to change it: Whatever form it currently takes
-- only reflects status quo current implementation.
--
-- The module "Hasura.GraphQL.Schema.Build" (commonly qualified as @GSB@)
-- provides standard building blocks for implementing many methods of this
-- class. And as such, these two modules are very much expected to evolve in
-- tandem.
--
-- See <#modelling Note BackendSchema modelling principles>.
class
Backend b =>
BackendSchema (b :: BackendType)
where
-- top level parsers
buildTableQueryAndSubscriptionFields ::
MonadBuildSchema b r m n =>
MkRootFieldName ->
SourceInfo b ->
TableName b ->
TableInfo b ->
StreamingSubscriptionsCtx ->
GQLNameIdentifier ->
m
( [FieldParser n (QueryDB b (RemoteRelationshipField UnpreparedValue) (UnpreparedValue b))],
[FieldParser n (QueryDB b (RemoteRelationshipField UnpreparedValue) (UnpreparedValue b))],
Maybe (G.Name, Parser 'Output n (ApolloFederationParserFunction n))
)
buildTableStreamingSubscriptionFields ::
MonadBuildSchema b r m n =>
MkRootFieldName ->
SourceInfo b ->
TableName b ->
TableInfo b ->
GQLNameIdentifier ->
m [FieldParser n (QueryDB b (RemoteRelationshipField UnpreparedValue) (UnpreparedValue b))]
buildTableRelayQueryFields ::
MonadBuildSchema b r m n =>
MkRootFieldName ->
SourceInfo b ->
TableName b ->
TableInfo b ->
GQLNameIdentifier ->
NESeq (ColumnInfo b) ->
m [FieldParser n (QueryDB b (RemoteRelationshipField UnpreparedValue) (UnpreparedValue b))]
buildTableInsertMutationFields ::
MonadBuildSchema b r m n =>
MkRootFieldName ->
Scenario ->
SourceInfo b ->
TableName b ->
TableInfo b ->
GQLNameIdentifier ->
m [FieldParser n (AnnotatedInsert b (RemoteRelationshipField UnpreparedValue) (UnpreparedValue b))]
-- | This method is responsible for building the GraphQL Schema for mutations
-- backed by @UPDATE@ statements on some table, as described in
-- @https://hasura.io/docs/latest/graphql/core/databases/postgres/mutations/update.html@.
--
-- The suggested way to implement this is using building blocks in GSB, c.f.
-- its namesake @GSB.@'Hasura.GraphQL.Schema.Build.buildTableUpdateMutationFields'.
buildTableUpdateMutationFields ::
MonadBuildSchema b r m n =>
MkRootFieldName ->
Scenario ->
-- | The source that the table lives in
SourceInfo b ->
-- | The name of the table being acted on
TableName b ->
-- | table info
TableInfo b ->
-- | field display name
GQLNameIdentifier ->
m [FieldParser n (AnnotatedUpdateG b (RemoteRelationshipField UnpreparedValue) (UnpreparedValue b))]
buildTableDeleteMutationFields ::
MonadBuildSchema b r m n =>
MkRootFieldName ->
Scenario ->
SourceInfo b ->
TableName b ->
TableInfo b ->
GQLNameIdentifier ->
m [FieldParser n (AnnDelG b (RemoteRelationshipField UnpreparedValue) (UnpreparedValue b))]
buildFunctionQueryFields ::
MonadBuildSchema b r m n =>
MkRootFieldName ->
SourceInfo b ->
FunctionName b ->
FunctionInfo b ->
TableName b ->
m [FieldParser n (QueryDB b (RemoteRelationshipField UnpreparedValue) (UnpreparedValue b))]
buildFunctionRelayQueryFields ::
MonadBuildSchema b r m n =>
MkRootFieldName ->
SourceInfo b ->
FunctionName b ->
FunctionInfo b ->
TableName b ->
NESeq (ColumnInfo b) ->
m [FieldParser n (QueryDB b (RemoteRelationshipField UnpreparedValue) (UnpreparedValue b))]
buildFunctionMutationFields ::
MonadBuildSchema b r m n =>
MkRootFieldName ->
SourceInfo b ->
FunctionName b ->
FunctionInfo b ->
TableName b ->
m [FieldParser n (MutationDB b (RemoteRelationshipField UnpreparedValue) (UnpreparedValue b))]
-- | Make a parser for relationships. Default implementaton elides
-- relationships altogether.
mkRelationshipParser ::
MonadBuildSchema b r m n =>
SourceInfo b ->
RelInfo b ->
m (Maybe (InputFieldsParser n (Maybe (IR.AnnotatedInsertField b (UnpreparedValue b)))))
mkRelationshipParser _ _ = pure Nothing
-- backend extensions
relayExtension :: Maybe (XRelay b)
nodesAggExtension :: Maybe (XNodesAgg b)
streamSubscriptionExtension :: Maybe (XStreamingSubscription b)
-- individual components
columnParser ::
(MonadParse n, MonadError QErr m, MonadReader r m, Has MkTypename r, Has NamingCase r) =>
ColumnType b ->
G.Nullability -> -- TODO maybe use Hasura.GraphQL.Parser.Schema.Nullability instead?
m (Parser 'Both n (ValueWithOrigin (ColumnValue b)))
-- | Parser for arguments on scalar fields in a selection set
scalarSelectionArgumentsParser ::
MonadParse n =>
ColumnType b ->
InputFieldsParser n (Maybe (ScalarSelectionArguments b))
orderByOperators ::
SourceInfo b ->
NamingCase ->
(G.Name, NonEmpty (Definition EnumValueInfo, (BasicOrderType b, NullsOrderType b)))
comparisonExps ::
MonadBuildSchema b r m n =>
ColumnType b ->
m (Parser 'Input n [ComparisonExp b])
-- | The input fields parser, for "count" aggregate field, yielding a function
-- which generates @'CountType b' from optional "distinct" field value
countTypeInput ::
MonadParse n =>
Maybe (Parser 'Both n (Column b)) ->
InputFieldsParser n (CountDistinct -> CountType b)
aggregateOrderByCountType :: ScalarType b
-- | Computed field parser
computedField ::
MonadBuildSchema b r m n =>
SourceInfo b ->
ComputedFieldInfo b ->
TableName b ->
TableInfo b ->
m (Maybe (FieldParser n (AnnotatedField b)))
-- | The public interface for the schema of table queries exposed by a backend.
--
-- Remote Schemas and the Relay schema are the chief backend-agnostic clients of
-- this typeclass.
--
-- Some of schema building components in the "Hasura.GraphQL.Schema" namespace
-- also make use of these methods, ensuring backends expose a consistent schema
-- regardless of the mode it's referenced.
--
-- Default implementations exist for all of these in
-- 'Hasura.GraphQL.Schema.Select'.
class Backend b => BackendTableSelectSchema (b :: BackendType) where
tableArguments ::
MonadBuildSchemaBase r m n =>
SourceInfo b ->
TableInfo b ->
m (InputFieldsParser n (IR.SelectArgsG b (UnpreparedValue b)))
tableSelectionSet ::
MonadBuildSchemaBase r m n =>
SourceInfo b ->
TableInfo b ->
m (Maybe (Parser 'Output n (AnnotatedFields b)))
selectTable ::
MonadBuildSchemaBase r m n =>
SourceInfo b ->
-- | table info
TableInfo b ->
-- | field display name
G.Name ->
-- | field description, if any
Maybe G.Description ->
m (Maybe (FieldParser n (SelectExp b)))
selectTableAggregate ::
MonadBuildSchemaBase r m n =>
SourceInfo b ->
-- | table info
TableInfo b ->
-- | field display name
G.Name ->
-- | field description, if any
Maybe G.Description ->
m (Maybe (FieldParser n (AggSelectExp b)))
type ComparisonExp b = OpExpG b (UnpreparedValue b)
-- $modelling
-- #modelling#
--
-- In its current form, we model every component, from the top level (query,
-- insert mutation, etc.) of the schema down to its leaf values, as a type class
-- method of @BackendSchema@.
--
-- Consider, for example, the following query for a given table "author":
--
-- > query {
-- > author(where: {id: {_eq: 2}}) {
-- > name
-- > }
-- > }
--
-- The chain of functions leading to a parser for this RootField will be along
-- the lines of:
--
-- > > BackendSchema.buildTableQueryAndSubscriptionFields (Suggested default its GSB namesake)
-- > > GSS.selectTable
-- > > BackendSchema.tableArguments (Suggested default implementation being
-- > GSS.defaultTableArgs)
-- > > GSS.tableWhereArg
-- > > GSBE.boolExp
-- > > BackendSchema.comparisonExps
-- > > BackendSchema.columnParser
-- >
-- > > tableSelectionSet (...)
-- > > fieldSelection
--
-- (where the abbreviation @GSB@ refers to "Hasura.GraphQL.Schema.Build" and @GSS@
-- refers to "Hasura.GraphQL.Schema.Select", and @GSBE@ refers to
-- "Hasura.GraphQL.Schema.BoolExp".)
--
-- Several of those steps are part of the class, meaning that a backend can
-- customize part of this tree without having to reimplement all of it. For
-- instance, a backend that supports a different set ot table arguments can
-- choose to reimplement 'tableArguments', but can still use
-- 'Hasura.GraphQL.Schema.Select.tableWhereArg' in its custom implementation.
--
-- Applying the above modelling guidelines has pros and cons:
--
-- * Pro: You can specify both shared and diverging behavior.
-- * Pro: You can specify a lot of behavior implicitly, i.e. it's easy to write.
-- * Con: You can specify a lot of behavior implicitly, i.e. it's hard do
-- understand without tracing through implementations.
-- * Con: You get a proliferation of type class methods and it's difficult to
-- understand how they fit together.
--
-- == Going forward, we want to follow some different modelling guidelines:
--
-- We should break up / refactor the building blocks (in
-- "Hasura.GraphQL.Schema.Build" etc.) which are used to implement the top-level
-- type class methods (e.g. @BackendSchema@.'buildTableQueryAndSubscriptionFields', c.f.
-- @GSB.@'Hasura.GraphQL.Schema.Build.buildTableQueryAndSubscriptionFields', etc.) and have them
-- invoke the backend-specific behaviors they rely on via /function arguments/
-- instead of other type class methods.
--
-- When we do this, the function call sites (which will often be in @instance
-- BackendSchema ...@) becomes the centralised place where we decide which behavior
-- variation to follow.
--
-- When faced with answering the question of "what does this method do, and how does
-- it do it?", at least you will have listed the other components it depends on
-- front and center without having to trace through its implementation.
--
-- That is of course, if we refactor our building blocks mindfully into
-- conceptually meaningful units. Otherwise we'll just end up with an
-- incomprehensible mass of poorly shaped pieces. And we will still have a hard
-- time explaining what they do.
--
-- In other words, It is still the case that if you don't clean your room
-- you'll be living in a mess.