graphql-engine/server/src-lib/Hasura/Backends/Postgres/Translate/Select.hs

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

1782 lines
70 KiB
Haskell
Raw Normal View History

-- | Postgres Translate Select
--
-- This module is a translation layer between IR and postgres-specific select queries.
--
-- There are three main types of selects (as distinguished from the IR):
--
-- * "simple" selects
--
-- * aggregate selects
--
-- * connection selects (used for relay)
--
-- Most exports from this module showcase this distinction. The "interesting" parts
-- of the call tree of these functions is similar:
--
-- * 'selectQuerySQL' -> 'mkSQLSelect' -> 'processAnnSimpleSelect' -> 'processSelectParams'/'processAnnFields'
--
-- * 'selectAggregateQuerySQL' -> 'mkAggregateSelect' -> 'processAnnAggregateSelect' -> 'processSelectParams'/'processAnnFields'
--
-- * 'connetionSelectQuerySQL' -> 'mkConnectionSelect' -> 'processConnectionSelection' -> 'processSelectParams'
--
--
-- Random thoughts that might help when diving deeper in this module:
--
-- * Extractors are a pair of an SQL expression and an alias; they get
-- translated like "[SELECT ...] <expr> as <alias>"
-- * a 'SelectSource' consists of a prefix, a source, a boolean conditional
-- expression, and info on whether sorting or slicing is done
-- (needed to handle the LIMIT optimisation)
-- * For details on creating the selection tree for relationships via
-- @MonadWriter JoinTree@, see 'withWriteJoinTree'
module Hasura.Backends.Postgres.Translate.Select
( selectQuerySQL,
selectAggregateQuerySQL,
connectionSelectQuerySQL,
asSingleRowJsonResp,
mkSQLSelect,
mkAggregateSelect,
mkConnectionSelect,
PostgresAnnotatedFieldJSON,
)
where
import Control.Lens ((^?))
import Control.Monad.Writer.Strict
import Data.HashMap.Strict qualified as HM
import Data.List.NonEmpty qualified as NE
import Data.Text qualified as T
import Data.Text.Extended
import Database.PG.Query qualified as Q
import Hasura.Backends.Postgres.SQL.DML qualified as S
import Hasura.Backends.Postgres.SQL.IdentifierUniqueness
import Hasura.Backends.Postgres.SQL.Types
import Hasura.Backends.Postgres.Translate.BoolExp
import Hasura.Backends.Postgres.Translate.Column (toJSONableExp)
import Hasura.Backends.Postgres.Translate.Types
import Hasura.Base.Error
import Hasura.EncJSON
import Hasura.GraphQL.Schema.Common (currentNodeIdVersion, nodeIdVersionInt)
import Hasura.Prelude
import Hasura.RQL.DML.Internal
import Hasura.RQL.IR.OrderBy
import Hasura.RQL.IR.Select
import Hasura.RQL.Types
import Hasura.SQL.Types
-- | Translates IR to Postgres queries for simple SELECTs (select queries that
-- are not aggregations, including subscriptions).
--
-- See 'mkSQLSelect' for the Postgres AST.
selectQuerySQL ::
forall pgKind.
(Backend ('Postgres pgKind), PostgresAnnotatedFieldJSON pgKind) =>
JsonAggSelect ->
AnnSimpleSelect ('Postgres pgKind) ->
Q.Query
selectQuerySQL jsonAggSelect sel =
Q.fromBuilder $ toSQL $ mkSQLSelect jsonAggSelect sel
-- | Translates IR to Postgres queries for aggregated SELECTs.
--
-- See 'mkAggregateSelect' for the Postgres AST.
selectAggregateQuerySQL ::
forall pgKind.
(Backend ('Postgres pgKind), PostgresAnnotatedFieldJSON pgKind) =>
AnnAggregateSelect ('Postgres pgKind) ->
Q.Query
selectAggregateQuerySQL =
Q.fromBuilder . toSQL . mkAggregateSelect
-- | Translates IR to Postgres queries for "connection" queries (used for Relay).
--
-- See 'mkConnectionSelect' for the Postgres AST.
connectionSelectQuerySQL ::
forall pgKind.
( Backend ('Postgres pgKind),
PostgresAnnotatedFieldJSON pgKind
) =>
ConnectionSelect ('Postgres pgKind) Void S.SQLExp ->
Q.Query
connectionSelectQuerySQL =
Q.fromBuilder . toSQL . mkConnectionSelect
-- | Helper function with no relation to anything else in the module.
--
-- This function is generally used on the result of 'selectQuerySQL',
-- 'selectAggregateQuerySQL' or 'connectionSelectSQL' to run said query and get
-- back the resulting JSON.
--
-- TODO: Perhaps this helper could find a new home.
asSingleRowJsonResp ::
Q.Query ->
[Q.PrepArg] ->
Q.TxE QErr EncJSON
asSingleRowJsonResp query args =
encJFromBS . runIdentity . Q.getRow
<$> Q.rawQE dmlTxErrorHandler query args True
-- | Converts a function name to an 'Identifier'.
--
-- If the schema name is public, it will just use its name, otherwise it will
-- prefix it by the schema name.
functionToIdentifier :: QualifiedFunction -> Identifier
functionToIdentifier = Identifier . qualifiedObjectToText
selectFromToFromItem :: Identifier -> SelectFrom ('Postgres pgKind) -> S.FromItem
selectFromToFromItem pfx = \case
FromTable tn -> S.FISimple tn Nothing
FromIdentifier i -> S.FIIdentifier $ toIdentifier i
allow custom mutations through actions (#3042) * basic doc for actions * custom_types, sync and async actions * switch to graphql-parser-hs on github * update docs * metadata import/export * webhook calls are now supported * relationships in sync actions * initialise.sql is now in sync with the migration file * fix metadata tests * allow specifying arguments of actions * fix blacklist check on check_build_worthiness job * track custom_types and actions related tables * handlers are now triggered on async actions * default to pgjson unless a field is involved in relationships, for generating definition list * use 'true' for action filter for non admin role * fix create_action_permission sql query * drop permissions when dropping an action * add a hdb_role view (and relationships) to fetch all roles in the system * rename 'webhook' key in action definition to 'handler' * allow templating actions wehook URLs with env vars * add 'update_action' /v1/query type * allow forwarding client headers by setting `forward_client_headers` in action definition * add 'headers' configuration in action definition * handle webhook error response based on status codes * support array relationships for custom types * implement single row mutation, see https://github.com/hasura/graphql-engine/issues/3731 * single row mutation: rename 'pk_columns' -> 'columns' and no-op refactor * use top level primary key inputs for delete_by_pk & account select permissions for single row mutations * use only REST semantics to resolve the webhook response * use 'pk_columns' instead of 'columns' for update_by_pk input * add python basic tests for single row mutations * add action context (name) in webhook payload * Async action response is accessible for non admin roles only if the request session vars equals to action's * clean nulls, empty arrays for actions, custom types in export metadata * async action mutation returns only the UUID of the action * unit tests for URL template parser * Basic sync actions python tests * fix output in async query & add async tests * add admin secret header in async actions python test * document async action architecture in Resolve/Action.hs file * support actions returning array of objects * tests for list type response actions * update docs with actions and custom types metadata API reference * update actions python tests as per #f8e1330 Co-authored-by: Tirumarai Selvan <tirumarai.selvan@gmail.com> Co-authored-by: Aravind Shankar <face11301@gmail.com> Co-authored-by: Rakesh Emmadi <12475069+rakeshkky@users.noreply.github.com>
2020-02-13 20:38:23 +03:00
FromFunction qf args defListM ->
S.FIFunc $
S.FunctionExp qf (fromTableRowArgs pfx args) $
Just $ S.mkFunctionAlias (functionToIdentifier qf) defListM
allow custom mutations through actions (#3042) * basic doc for actions * custom_types, sync and async actions * switch to graphql-parser-hs on github * update docs * metadata import/export * webhook calls are now supported * relationships in sync actions * initialise.sql is now in sync with the migration file * fix metadata tests * allow specifying arguments of actions * fix blacklist check on check_build_worthiness job * track custom_types and actions related tables * handlers are now triggered on async actions * default to pgjson unless a field is involved in relationships, for generating definition list * use 'true' for action filter for non admin role * fix create_action_permission sql query * drop permissions when dropping an action * add a hdb_role view (and relationships) to fetch all roles in the system * rename 'webhook' key in action definition to 'handler' * allow templating actions wehook URLs with env vars * add 'update_action' /v1/query type * allow forwarding client headers by setting `forward_client_headers` in action definition * add 'headers' configuration in action definition * handle webhook error response based on status codes * support array relationships for custom types * implement single row mutation, see https://github.com/hasura/graphql-engine/issues/3731 * single row mutation: rename 'pk_columns' -> 'columns' and no-op refactor * use top level primary key inputs for delete_by_pk & account select permissions for single row mutations * use only REST semantics to resolve the webhook response * use 'pk_columns' instead of 'columns' for update_by_pk input * add python basic tests for single row mutations * add action context (name) in webhook payload * Async action response is accessible for non admin roles only if the request session vars equals to action's * clean nulls, empty arrays for actions, custom types in export metadata * async action mutation returns only the UUID of the action * unit tests for URL template parser * Basic sync actions python tests * fix output in async query & add async tests * add admin secret header in async actions python test * document async action architecture in Resolve/Action.hs file * support actions returning array of objects * tests for list type response actions * update docs with actions and custom types metadata API reference * update actions python tests as per #f8e1330 Co-authored-by: Tirumarai Selvan <tirumarai.selvan@gmail.com> Co-authored-by: Aravind Shankar <face11301@gmail.com> Co-authored-by: Rakesh Emmadi <12475069+rakeshkky@users.noreply.github.com>
2020-02-13 20:38:23 +03:00
-- You should be able to retrieve this information
-- from the FromItem generated with selectFromToFromItem
allow custom mutations through actions (#3042) * basic doc for actions * custom_types, sync and async actions * switch to graphql-parser-hs on github * update docs * metadata import/export * webhook calls are now supported * relationships in sync actions * initialise.sql is now in sync with the migration file * fix metadata tests * allow specifying arguments of actions * fix blacklist check on check_build_worthiness job * track custom_types and actions related tables * handlers are now triggered on async actions * default to pgjson unless a field is involved in relationships, for generating definition list * use 'true' for action filter for non admin role * fix create_action_permission sql query * drop permissions when dropping an action * add a hdb_role view (and relationships) to fetch all roles in the system * rename 'webhook' key in action definition to 'handler' * allow templating actions wehook URLs with env vars * add 'update_action' /v1/query type * allow forwarding client headers by setting `forward_client_headers` in action definition * add 'headers' configuration in action definition * handle webhook error response based on status codes * support array relationships for custom types * implement single row mutation, see https://github.com/hasura/graphql-engine/issues/3731 * single row mutation: rename 'pk_columns' -> 'columns' and no-op refactor * use top level primary key inputs for delete_by_pk & account select permissions for single row mutations * use only REST semantics to resolve the webhook response * use 'pk_columns' instead of 'columns' for update_by_pk input * add python basic tests for single row mutations * add action context (name) in webhook payload * Async action response is accessible for non admin roles only if the request session vars equals to action's * clean nulls, empty arrays for actions, custom types in export metadata * async action mutation returns only the UUID of the action * unit tests for URL template parser * Basic sync actions python tests * fix output in async query & add async tests * add admin secret header in async actions python test * document async action architecture in Resolve/Action.hs file * support actions returning array of objects * tests for list type response actions * update docs with actions and custom types metadata API reference * update actions python tests as per #f8e1330 Co-authored-by: Tirumarai Selvan <tirumarai.selvan@gmail.com> Co-authored-by: Aravind Shankar <face11301@gmail.com> Co-authored-by: Rakesh Emmadi <12475069+rakeshkky@users.noreply.github.com>
2020-02-13 20:38:23 +03:00
-- however given from S.FromItem is modelled, it is not
-- possible currently.
--
-- More precisely, 'selectFromToFromItem' is injective but not surjective, so
-- any S.FromItem -> S.Qual function would have to be partial.
selectFromToQual :: SelectFrom ('Postgres pgKind) -> S.Qual
selectFromToQual = \case
FromTable table -> S.QualTable table
FromIdentifier i -> S.QualifiedIdentifier (toIdentifier i) Nothing
FromFunction qf _ _ -> S.QualifiedIdentifier (functionToIdentifier qf) Nothing
aggregateFieldToExp :: AggregateFields ('Postgres pgKind) -> StringifyNumbers -> S.SQLExp
aggregateFieldToExp aggFlds strfyNum = jsonRow
where
jsonRow = S.applyJsonBuildObj (concatMap aggToFlds aggFlds)
withAls fldName sqlExp = [S.SELit fldName, sqlExp]
aggToFlds (FieldName t, fld) = withAls t $ case fld of
AFCount cty -> S.SECount cty
AFOp aggOp -> aggOpToObj aggOp
AFExp e -> S.SELit e
aggOpToObj (AggregateOp opText flds) =
S.applyJsonBuildObj $ concatMap (colFldsToExtr opText) flds
colFldsToExtr opText (FieldName t, CFCol col ty) =
[ S.SELit t,
toJSONableExp strfyNum ty False $
S.SEFnApp opText [S.SEIdentifier $ toIdentifier col] Nothing
]
colFldsToExtr _ (FieldName t, CFExp e) =
[S.SELit t, S.SELit e]
asSingleRowExtr :: S.Alias -> S.SQLExp
asSingleRowExtr col =
S.SEFnApp "coalesce" [jsonAgg, S.SELit "null"] Nothing
where
jsonAgg =
S.SEOpApp
(S.SQLOp "->")
[ S.SEFnApp "json_agg" [S.SEIdentifier $ toIdentifier col] Nothing,
S.SEUnsafe "0"
]
withJsonAggExtr ::
PermissionLimitSubQuery -> Maybe S.OrderByExp -> S.Alias -> S.SQLExp
withJsonAggExtr permLimitSubQuery ordBy alias =
-- if select has aggregations then use subquery to apply permission limit
case permLimitSubQuery of
PLSQRequired permLimit -> withPermLimit permLimit
PLSQNotRequired -> simpleJsonAgg
where
simpleJsonAgg = mkSimpleJsonAgg rowIdenExp ordBy
rowIdenExp = S.SEIdentifier $ S.getAlias alias
subSelAls = Identifier "sub_query"
unnestTable = Identifier "unnest_table"
mkSimpleJsonAgg rowExp ob =
let jsonAggExp = S.SEFnApp "json_agg" [rowExp] ob
in S.SEFnApp "coalesce" [jsonAggExp, S.SELit "[]"] Nothing
withPermLimit limit =
let subSelect = mkSubSelect limit
rowIdentifier = S.mkQIdenExp subSelAls alias
extr = S.Extractor (mkSimpleJsonAgg rowIdentifier newOrderBy) Nothing
fromExp =
S.FromExp $
pure $
S.mkSelFromItem subSelect $ S.Alias subSelAls
in S.SESelect $
S.mkSelect
{ S.selExtr = pure extr,
S.selFrom = Just fromExp
}
mkSubSelect limit =
let jsonRowExtr =
flip S.Extractor (Just alias) $
S.mkQIdenExp unnestTable alias
obExtrs = flip map newOBAliases $ \a ->
S.Extractor (S.mkQIdenExp unnestTable a) $ Just $ S.Alias a
in S.mkSelect
{ S.selExtr = jsonRowExtr : obExtrs,
S.selFrom = Just $ S.FromExp $ pure unnestFromItem,
S.selLimit = Just $ S.LimitExp $ S.intToSQLExp limit,
S.selOrderBy = newOrderBy
}
unnestFromItem =
let arrayAggItems = flip map (rowIdenExp : obCols) $
\s -> S.SEFnApp "array_agg" [s] Nothing
in S.FIUnnest arrayAggItems (S.Alias unnestTable) $
rowIdenExp : map S.SEIdentifier newOBAliases
newOrderBy = S.OrderByExp <$> NE.nonEmpty newOBItems
(newOBItems, obCols, newOBAliases) = maybe ([], [], []) transformOrderBy ordBy
transformOrderBy (S.OrderByExp l) = unzip3 $
flip map (zip (toList l) [1 ..]) $ \(obItem, i :: Int) ->
let iden = Identifier $ "ob_col_" <> tshow i
in ( obItem {S.oColumn = S.SEIdentifier iden},
S.oColumn obItem,
iden
)
asJsonAggExtr ::
JsonAggSelect -> S.Alias -> PermissionLimitSubQuery -> Maybe S.OrderByExp -> S.Extractor
asJsonAggExtr jsonAggSelect als permLimitSubQuery ordByExpM =
allow custom mutations through actions (#3042) * basic doc for actions * custom_types, sync and async actions * switch to graphql-parser-hs on github * update docs * metadata import/export * webhook calls are now supported * relationships in sync actions * initialise.sql is now in sync with the migration file * fix metadata tests * allow specifying arguments of actions * fix blacklist check on check_build_worthiness job * track custom_types and actions related tables * handlers are now triggered on async actions * default to pgjson unless a field is involved in relationships, for generating definition list * use 'true' for action filter for non admin role * fix create_action_permission sql query * drop permissions when dropping an action * add a hdb_role view (and relationships) to fetch all roles in the system * rename 'webhook' key in action definition to 'handler' * allow templating actions wehook URLs with env vars * add 'update_action' /v1/query type * allow forwarding client headers by setting `forward_client_headers` in action definition * add 'headers' configuration in action definition * handle webhook error response based on status codes * support array relationships for custom types * implement single row mutation, see https://github.com/hasura/graphql-engine/issues/3731 * single row mutation: rename 'pk_columns' -> 'columns' and no-op refactor * use top level primary key inputs for delete_by_pk & account select permissions for single row mutations * use only REST semantics to resolve the webhook response * use 'pk_columns' instead of 'columns' for update_by_pk input * add python basic tests for single row mutations * add action context (name) in webhook payload * Async action response is accessible for non admin roles only if the request session vars equals to action's * clean nulls, empty arrays for actions, custom types in export metadata * async action mutation returns only the UUID of the action * unit tests for URL template parser * Basic sync actions python tests * fix output in async query & add async tests * add admin secret header in async actions python test * document async action architecture in Resolve/Action.hs file * support actions returning array of objects * tests for list type response actions * update docs with actions and custom types metadata API reference * update actions python tests as per #f8e1330 Co-authored-by: Tirumarai Selvan <tirumarai.selvan@gmail.com> Co-authored-by: Aravind Shankar <face11301@gmail.com> Co-authored-by: Rakesh Emmadi <12475069+rakeshkky@users.noreply.github.com>
2020-02-13 20:38:23 +03:00
flip S.Extractor (Just als) $ case jsonAggSelect of
JASMultipleRows -> withJsonAggExtr permLimitSubQuery ordByExpM als
allow custom mutations through actions (#3042) * basic doc for actions * custom_types, sync and async actions * switch to graphql-parser-hs on github * update docs * metadata import/export * webhook calls are now supported * relationships in sync actions * initialise.sql is now in sync with the migration file * fix metadata tests * allow specifying arguments of actions * fix blacklist check on check_build_worthiness job * track custom_types and actions related tables * handlers are now triggered on async actions * default to pgjson unless a field is involved in relationships, for generating definition list * use 'true' for action filter for non admin role * fix create_action_permission sql query * drop permissions when dropping an action * add a hdb_role view (and relationships) to fetch all roles in the system * rename 'webhook' key in action definition to 'handler' * allow templating actions wehook URLs with env vars * add 'update_action' /v1/query type * allow forwarding client headers by setting `forward_client_headers` in action definition * add 'headers' configuration in action definition * handle webhook error response based on status codes * support array relationships for custom types * implement single row mutation, see https://github.com/hasura/graphql-engine/issues/3731 * single row mutation: rename 'pk_columns' -> 'columns' and no-op refactor * use top level primary key inputs for delete_by_pk & account select permissions for single row mutations * use only REST semantics to resolve the webhook response * use 'pk_columns' instead of 'columns' for update_by_pk input * add python basic tests for single row mutations * add action context (name) in webhook payload * Async action response is accessible for non admin roles only if the request session vars equals to action's * clean nulls, empty arrays for actions, custom types in export metadata * async action mutation returns only the UUID of the action * unit tests for URL template parser * Basic sync actions python tests * fix output in async query & add async tests * add admin secret header in async actions python test * document async action architecture in Resolve/Action.hs file * support actions returning array of objects * tests for list type response actions * update docs with actions and custom types metadata API reference * update actions python tests as per #f8e1330 Co-authored-by: Tirumarai Selvan <tirumarai.selvan@gmail.com> Co-authored-by: Aravind Shankar <face11301@gmail.com> Co-authored-by: Rakesh Emmadi <12475069+rakeshkky@users.noreply.github.com>
2020-02-13 20:38:23 +03:00
JASSingleObject -> asSingleRowExtr als
-- array relationships are not grouped, so have to be prefixed by
-- parent's alias
mkUniqArrayRelationAlias :: FieldName -> [FieldName] -> Identifier
mkUniqArrayRelationAlias parAls flds =
let sortedFields = sort flds
in Identifier $
getFieldNameTxt parAls <> "."
<> T.intercalate "." (map getFieldNameTxt sortedFields)
mkArrayRelationTableAlias :: Identifier -> FieldName -> [FieldName] -> Identifier
mkArrayRelationTableAlias pfx parAls flds =
pfx <> Identifier ".ar." <> uniqArrRelAls
where
uniqArrRelAls = mkUniqArrayRelationAlias parAls flds
mkObjectRelationTableAlias :: Identifier -> RelName -> Identifier
mkObjectRelationTableAlias pfx relName =
pfx <> Identifier ".or." <> toIdentifier relName
mkComputedFieldTableAlias :: Identifier -> FieldName -> Identifier
mkComputedFieldTableAlias pfx fldAls =
pfx <> Identifier ".cf." <> toIdentifier fldAls
mkBaseTableAlias :: Identifier -> Identifier
mkBaseTableAlias pfx =
pfx <> Identifier ".base"
mkBaseTableColumnAlias :: Identifier -> PGCol -> Identifier
mkBaseTableColumnAlias pfx pgColumn =
pfx <> Identifier ".pg." <> toIdentifier pgColumn
mkOrderByFieldName :: ToTxt a => a -> FieldName
mkOrderByFieldName name =
FieldName $ toTxt name <> "." <> "order_by"
mkAggregateOrderByAlias :: AnnotatedAggregateOrderBy ('Postgres pgKind) -> S.Alias
mkAggregateOrderByAlias =
(S.Alias . Identifier) . \case
AAOCount -> "count"
AAOOp opText col -> opText <> "." <> getPGColTxt (ciColumn col)
mkArrayRelationSourcePrefix ::
Identifier ->
FieldName ->
HM.HashMap FieldName [FieldName] ->
FieldName ->
Identifier
mkArrayRelationSourcePrefix parentSourcePrefix parentFieldName similarFieldsMap fieldName =
mkArrayRelationTableAlias parentSourcePrefix parentFieldName $
HM.lookupDefault [fieldName] fieldName similarFieldsMap
mkArrayRelationAlias ::
FieldName ->
HM.HashMap FieldName [FieldName] ->
FieldName ->
S.Alias
mkArrayRelationAlias parentFieldName similarFieldsMap fieldName =
S.Alias $
mkUniqArrayRelationAlias parentFieldName $
HM.lookupDefault [fieldName] fieldName similarFieldsMap
fromTableRowArgs ::
Identifier -> FunctionArgsExpTableRow S.SQLExp -> S.FunctionArgs
fromTableRowArgs prefix = toFunctionArgs . fmap toSQLExp
where
toFunctionArgs (FunctionArgsExp positional named) =
S.FunctionArgs positional named
toSQLExp =
onArgumentExp
(S.SERowIdentifier alias)
(S.mkQIdenExp alias . Identifier)
alias = mkBaseTableAlias prefix
-- uses row_to_json to build a json object
withRowToJSON ::
FieldName -> [S.Extractor] -> (S.Alias, S.SQLExp)
withRowToJSON parAls extrs =
(S.toAlias parAls, jsonRow)
where
jsonRow = S.applyRowToJson extrs
-- uses json_build_object to build a json object
withJsonBuildObj ::
FieldName -> [S.SQLExp] -> (S.Alias, S.SQLExp)
withJsonBuildObj parAls exps =
(S.toAlias parAls, jsonRow)
where
jsonRow = S.applyJsonBuildObj exps
-- | Forces aggregation
withForceAggregation :: S.TypeAnn -> S.SQLExp -> S.SQLExp
withForceAggregation tyAnn e =
-- bool_or to force aggregation
S.SEFnApp "coalesce" [e, S.SETyAnn (S.SEUnsafe "bool_or('true')") tyAnn] Nothing
mkAggregateOrderByExtractorAndFields ::
forall pgKind.
Backend ('Postgres pgKind) =>
AnnotatedAggregateOrderBy ('Postgres pgKind) ->
(S.Extractor, AggregateFields ('Postgres pgKind))
mkAggregateOrderByExtractorAndFields annAggOrderBy =
case annAggOrderBy of
AAOCount ->
( S.Extractor S.countStar alias,
[(FieldName "count", AFCount S.CTStar)]
)
AAOOp opText pgColumnInfo ->
let pgColumn = ciColumn pgColumnInfo
pgType = ciType pgColumnInfo
in ( S.Extractor (S.SEFnApp opText [S.SEIdentifier $ toIdentifier pgColumn] Nothing) alias,
[ ( FieldName opText,
AFOp $
AggregateOp
opText
[ ( fromCol @('Postgres pgKind) pgColumn,
CFCol pgColumn pgType
)
]
)
]
)
where
alias = Just $ mkAggregateOrderByAlias annAggOrderBy
-- | Generate alias for order by extractors
mkAnnOrderByAlias ::
Identifier -> FieldName -> SimilarArrayFields -> AnnotatedOrderByElement ('Postgres pgKind) v -> S.Alias
mkAnnOrderByAlias pfx parAls similarFields = \case
AOCColumn pgColumnInfo ->
let pgColumn = ciColumn pgColumnInfo
obColAls = mkBaseTableColumnAlias pfx pgColumn
in S.Alias obColAls
-- "pfx.or.relname"."pfx.ob.or.relname.rest" AS "pfx.ob.or.relname.rest"
AOCObjectRelation relInfo _ rest ->
let rn = riName relInfo
relPfx = mkObjectRelationTableAlias pfx rn
ordByFldName = mkOrderByFieldName rn
nesAls = mkAnnOrderByAlias relPfx ordByFldName mempty rest
in nesAls
AOCArrayAggregation relInfo _ aggOrderBy ->
let rn = riName relInfo
arrPfx =
mkArrayRelationSourcePrefix pfx parAls similarFields $
mkOrderByFieldName rn
obAls = arrPfx <> Identifier "." <> toIdentifier (mkAggregateOrderByAlias aggOrderBy)
in S.Alias obAls
AOCComputedField cfOrderBy ->
let fieldName = fromComputedField $ _cfobName cfOrderBy
in case _cfobOrderByElement cfOrderBy of
CFOBEScalar _ -> S.Alias $ mkComputedFieldTableAlias pfx fieldName
CFOBETableAggregation _ _ aggOrderBy ->
let cfPfx = mkComputedFieldTableAlias pfx fieldName
obAls = cfPfx <> Identifier "." <> toIdentifier (mkAggregateOrderByAlias aggOrderBy)
in S.Alias obAls
applyDistinctOnAtBase ::
NE.NonEmpty PGCol -> S.DistinctExpr
applyDistinctOnAtBase =
S.DistinctOn . map (S.SEIdentifier . toIdentifier) . toList
applyDistinctOnAtNode ::
Identifier ->
NE.NonEmpty PGCol ->
( S.DistinctExpr,
[(S.Alias, S.SQLExp)] -- additional column extractors
)
applyDistinctOnAtNode pfx neCols = (distOnExp, colExtrs)
where
cols = toList neCols
distOnExp = S.DistinctOn $ map (S.SEIdentifier . toIdentifier . mkQColAls) cols
mkQCol c = S.mkQIdenExp (mkBaseTableAlias pfx) $ toIdentifier c
mkQColAls = S.Alias . mkBaseTableColumnAlias pfx
colExtrs = flip map cols $ mkQColAls &&& mkQCol
type SimilarArrayFields = HM.HashMap FieldName [FieldName]
mkSimilarArrayFields ::
forall pgKind v.
(Backend ('Postgres pgKind), Eq v) =>
AnnFieldsG ('Postgres pgKind) Void v ->
Maybe (NE.NonEmpty (AnnotatedOrderByItemG ('Postgres pgKind) v)) ->
SimilarArrayFields
mkSimilarArrayFields annFields maybeOrderBys =
HM.fromList $
flip map allTuples $
\(relNameAndArgs, fieldName) -> (fieldName, getSimilarFields relNameAndArgs)
where
getSimilarFields relNameAndArgs = map snd $ filter ((== relNameAndArgs) . fst) allTuples
allTuples = arrayRelationTuples <> aggOrderByRelationTuples
arrayRelationTuples =
let arrayFields = mapMaybe getAnnArr annFields
in flip map arrayFields $
\(f, relSel) -> (getArrayRelNameAndSelectArgs relSel, f)
aggOrderByRelationTuples =
let mkItem (relName, fieldName) =
( (relName, noSelectArgs),
fieldName
)
in map mkItem $
maybe
[]
(mapMaybe (fetchAggOrderByRels . obiColumn) . toList)
maybeOrderBys
fetchAggOrderByRels (AOCArrayAggregation ri _ _) =
Just (riName ri, mkOrderByFieldName $ riName ri)
fetchAggOrderByRels _ = Nothing
getArrayRelNameAndSelectArgs ::
ArraySelectG ('Postgres pgKind) r v ->
(RelName, SelectArgsG ('Postgres pgKind) v)
getArrayRelNameAndSelectArgs = \case
ASSimple r -> (_aarRelationshipName r, _asnArgs $ _aarAnnSelect r)
ASAggregate r -> (_aarRelationshipName r, _asnArgs $ _aarAnnSelect r)
ASConnection r -> (_aarRelationshipName r, _asnArgs $ _csSelect $ _aarAnnSelect r)
getAnnArr ::
(a, AnnFieldG ('Postgres pgKind) r v) ->
Maybe (a, ArraySelectG ('Postgres pgKind) r v)
getAnnArr (f, annFld) = case annFld of
AFArrayRelation (ASConnection _) -> Nothing
AFArrayRelation ar -> Just (f, ar)
_ -> Nothing
-- | This is the lowest level function which deals with @MonadWriter JoinTree@, whose
-- purpose is to essentially create the selection tree across relationships.
--
-- Each type of relationship uses a different kind of update function; see
-- 'withWriteObjectRelation', 'withWriteArrayRelation', 'withWriteArrayConnection',
-- and 'withWriteComputedFieldTableSet'.
--
-- See the definition of 'JoinTree' for details before diving further
-- (particularly its components and Monoid instance).
withWriteJoinTree ::
(MonadWriter JoinTree m) =>
(JoinTree -> b -> JoinTree) ->
m (a, b) ->
m a
withWriteJoinTree joinTreeUpdater action =
pass $ do
(out, result) <- action
let fromJoinTree joinTree =
joinTreeUpdater joinTree result
pure (out, fromJoinTree)
withWriteObjectRelation ::
(MonadWriter JoinTree m) =>
m
( ObjectRelationSource,
HM.HashMap S.Alias S.SQLExp,
a
) ->
m a
withWriteObjectRelation action =
withWriteJoinTree updateJoinTree $ do
(source, nodeExtractors, out) <- action
pure (out, (source, nodeExtractors))
where
updateJoinTree joinTree (source, nodeExtractors) =
let selectNode = SelectNode nodeExtractors joinTree
in mempty {_jtObjectRelations = HM.singleton source selectNode}
withWriteArrayRelation ::
(MonadWriter JoinTree m) =>
m
( ArrayRelationSource,
S.Extractor,
HM.HashMap S.Alias S.SQLExp,
a
) ->
m a
withWriteArrayRelation action =
withWriteJoinTree updateJoinTree $ do
(source, topExtractor, nodeExtractors, out) <- action
pure (out, (source, topExtractor, nodeExtractors))
where
updateJoinTree joinTree (source, topExtractor, nodeExtractors) =
let arraySelectNode =
MultiRowSelectNode [topExtractor] $
SelectNode nodeExtractors joinTree
in mempty {_jtArrayRelations = HM.singleton source arraySelectNode}
withWriteArrayConnection ::
(MonadWriter JoinTree m) =>
m
( ArrayConnectionSource,
S.Extractor,
HM.HashMap S.Alias S.SQLExp,
a
) ->
m a
withWriteArrayConnection action =
withWriteJoinTree updateJoinTree $ do
(source, topExtractor, nodeExtractors, out) <- action
pure (out, (source, topExtractor, nodeExtractors))
where
updateJoinTree joinTree (source, topExtractor, nodeExtractors) =
let arraySelectNode =
MultiRowSelectNode [topExtractor] $
SelectNode nodeExtractors joinTree
in mempty {_jtArrayConnections = HM.singleton source arraySelectNode}
withWriteComputedFieldTableSet ::
(MonadWriter JoinTree m) =>
m
( ComputedFieldTableSetSource,
S.Extractor,
HM.HashMap S.Alias S.SQLExp,
a
) ->
m a
withWriteComputedFieldTableSet action =
withWriteJoinTree updateJoinTree $ do
(source, topExtractor, nodeExtractors, out) <- action
pure (out, (source, topExtractor, nodeExtractors))
where
updateJoinTree joinTree (source, topExtractor, nodeExtractors) =
let selectNode = MultiRowSelectNode [topExtractor] $ SelectNode nodeExtractors joinTree
in mempty {_jtComputedFieldTableSets = HM.singleton source selectNode}
processAnnSimpleSelect ::
forall pgKind m.
( MonadReader StringifyNumbers m,
MonadWriter JoinTree m,
Backend ('Postgres pgKind),
PostgresAnnotatedFieldJSON pgKind
) =>
SourcePrefixes ->
FieldName ->
PermissionLimitSubQuery ->
AnnSimpleSelect ('Postgres pgKind) ->
m
( SelectSource,
HM.HashMap S.Alias S.SQLExp
)
processAnnSimpleSelect sourcePrefixes fieldAlias permLimitSubQuery annSimpleSel = do
(selectSource, orderByAndDistinctExtrs, _) <-
processSelectParams
sourcePrefixes
fieldAlias
similarArrayFields
tableFrom
permLimitSubQuery
tablePermissions
tableArgs
annFieldsExtr <- processAnnFields (_pfThis sourcePrefixes) fieldAlias similarArrayFields annSelFields
let allExtractors = HM.fromList $ annFieldsExtr : orderByAndDistinctExtrs
pure (selectSource, allExtractors)
where
AnnSelectG annSelFields tableFrom tablePermissions tableArgs _ = annSimpleSel
similarArrayFields =
mkSimilarArrayFields annSelFields $ _saOrderBy tableArgs
processAnnAggregateSelect ::
forall pgKind m.
( MonadReader StringifyNumbers m,
MonadWriter JoinTree m,
Backend ('Postgres pgKind),
PostgresAnnotatedFieldJSON pgKind
) =>
SourcePrefixes ->
FieldName ->
AnnAggregateSelect ('Postgres pgKind) ->
m
( SelectSource,
HM.HashMap S.Alias S.SQLExp,
S.Extractor
)
processAnnAggregateSelect sourcePrefixes fieldAlias annAggSel = do
(selectSource, orderByAndDistinctExtrs, _) <-
processSelectParams
sourcePrefixes
fieldAlias
similarArrayFields
tableFrom
permLimitSubQuery
tablePermissions
tableArgs
let thisSourcePrefix = _pfThis sourcePrefixes
processedFields <- forM aggSelFields $ \(fieldName, field) ->
(fieldName,)
<$> case field of
TAFAgg aggFields ->
pure
( aggregateFieldsToExtractorExps thisSourcePrefix aggFields,
aggregateFieldToExp aggFields strfyNum
)
TAFNodes _ annFields -> do
annFieldExtr <- processAnnFields thisSourcePrefix fieldName similarArrayFields annFields
pure
( [annFieldExtr],
withJsonAggExtr permLimitSubQuery (orderByForJsonAgg selectSource) $
S.Alias $ toIdentifier fieldName
)
TAFExp e ->
pure
( [],
withForceAggregation S.textTypeAnn $ S.SELit e
)
let topLevelExtractor =
flip S.Extractor (Just $ S.Alias $ toIdentifier fieldAlias) $
S.applyJsonBuildObj $
flip concatMap (map (second snd) processedFields) $
\(FieldName fieldText, fieldExp) -> [S.SELit fieldText, fieldExp]
nodeExtractors =
HM.fromList $
concatMap (fst . snd) processedFields <> orderByAndDistinctExtrs
pure (selectSource, nodeExtractors, topLevelExtractor)
where
AnnSelectG aggSelFields tableFrom tablePermissions tableArgs strfyNum = annAggSel
permLimit = _tpLimit tablePermissions
orderBy = _saOrderBy tableArgs
permLimitSubQuery = mkPermissionLimitSubQuery permLimit aggSelFields orderBy
similarArrayFields = HM.unions $
flip map (map snd aggSelFields) $ \case
TAFAgg _ -> mempty
TAFNodes _ annFlds ->
mkSimilarArrayFields annFlds orderBy
TAFExp _ -> mempty
mkPermissionLimitSubQuery ::
Maybe Int ->
TableAggregateFields ('Postgres pgKind) ->
Maybe (NE.NonEmpty (AnnotatedOrderByItem ('Postgres pgKind))) ->
PermissionLimitSubQuery
mkPermissionLimitSubQuery permLimit aggFields orderBys =
case permLimit of
Nothing -> PLSQNotRequired
Just limit ->
if hasAggregateField || hasAggOrderBy
then PLSQRequired limit
else PLSQNotRequired
where
hasAggregateField = flip any (map snd aggFields) $
\case
TAFAgg _ -> True
_ -> False
hasAggOrderBy = case orderBys of
Nothing -> False
Just l -> flip any (concatMap toList $ toList l) $
\case
AOCArrayAggregation {} -> True
_ -> False
processArrayRelation ::
forall pgKind m.
( MonadReader StringifyNumbers m,
MonadWriter JoinTree m,
Backend ('Postgres pgKind),
PostgresAnnotatedFieldJSON pgKind
) =>
SourcePrefixes ->
FieldName ->
S.Alias ->
ArraySelect ('Postgres pgKind) ->
m ()
processArrayRelation sourcePrefixes fieldAlias relAlias arrSel =
case arrSel of
ASSimple annArrRel -> withWriteArrayRelation $ do
let AnnRelationSelectG _ colMapping sel = annArrRel
permLimitSubQuery =
maybe PLSQNotRequired PLSQRequired $ _tpLimit $ _asnPerm sel
(source, nodeExtractors) <-
processAnnSimpleSelect sourcePrefixes fieldAlias permLimitSubQuery sel
let topExtr =
asJsonAggExtr
JASMultipleRows
(S.toAlias fieldAlias)
permLimitSubQuery
$ orderByForJsonAgg source
pure
( ArrayRelationSource relAlias colMapping source,
topExtr,
nodeExtractors,
()
)
ASAggregate aggSel -> withWriteArrayRelation $ do
let AnnRelationSelectG _ colMapping sel = aggSel
(source, nodeExtractors, topExtr) <-
processAnnAggregateSelect sourcePrefixes fieldAlias sel
pure
( ArrayRelationSource relAlias colMapping source,
topExtr,
nodeExtractors,
()
)
ASConnection connSel -> withWriteArrayConnection $ do
let AnnRelationSelectG _ colMapping sel = connSel
(source, topExtractor, nodeExtractors) <-
processConnectionSelect sourcePrefixes fieldAlias relAlias colMapping sel
pure
( source,
topExtractor,
nodeExtractors,
()
)
{- Note [Optimizing queries using limit/offset]
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Refer to the issue https://github.com/hasura/graphql-engine/issues/5745
Before this change, limit/offset/distinct_on is applied at outer selection
node along with order by clause. This greatly reduces query performance if
our base selection table contains many rows and relationships are selected
which joins remote tables. We need to optimize application of limit wrt to
order by input.
If "Order by" is not present:
Apply limit/offset/distinct on at the base table selection
Else if "Order by" contains only columns:
Apply limit/offset/distinct_on at the base table selection along with order by
Otherwise:
Apply limit/offset/distinct_on at the node selection along with order by
-}
processSelectParams ::
forall pgKind m.
( MonadReader StringifyNumbers m,
MonadWriter JoinTree m,
Backend ('Postgres pgKind)
) =>
SourcePrefixes ->
FieldName ->
SimilarArrayFields ->
SelectFrom ('Postgres pgKind) ->
PermissionLimitSubQuery ->
TablePerm ('Postgres pgKind) ->
SelectArgs ('Postgres pgKind) ->
m
( SelectSource,
[(S.Alias, S.SQLExp)],
Maybe S.SQLExp -- Order by cursor
)
processSelectParams
sourcePrefixes
fieldAlias
similarArrFields
selectFrom
permLimitSubQ
tablePermissions
tableArgs = do
(additionalExtrs, selectSorting, cursorExp) <-
processOrderByItems thisSourcePrefix fieldAlias similarArrFields distM orderByM
let fromItem = selectFromToFromItem (_pfBase sourcePrefixes) selectFrom
finalWhere =
toSQLBoolExp (selectFromToQual selectFrom) $
maybe permFilter (andAnnBoolExps permFilter) whereM
sortingAndSlicing = SortingAndSlicing selectSorting selectSlicing
selectSource =
SelectSource
thisSourcePrefix
fromItem
finalWhere
sortingAndSlicing
pure
( selectSource,
additionalExtrs,
cursorExp
)
where
thisSourcePrefix = _pfThis sourcePrefixes
SelectArgs whereM orderByM inpLimitM offsetM distM = tableArgs
TablePerm permFilter permLimit = tablePermissions
selectSlicing = SelectSlicing finalLimit offsetM
finalLimit =
-- if sub query is required, then only use input limit
-- because permission limit is being applied in subquery
-- else compare input and permission limits
case permLimitSubQ of
PLSQRequired _ -> inpLimitM
PLSQNotRequired -> compareLimits
compareLimits =
case (inpLimitM, permLimit) of
(inpLim, Nothing) -> inpLim
(Nothing, permLim) -> permLim
(Just inp, Just perm) -> Just if inp < perm then inp else perm
processOrderByItems ::
forall pgKind m.
( MonadReader StringifyNumbers m,
MonadWriter JoinTree m,
Backend ('Postgres pgKind)
) =>
Identifier ->
FieldName ->
SimilarArrayFields ->
Maybe (NE.NonEmpty PGCol) ->
Maybe (NE.NonEmpty (AnnotatedOrderByItem ('Postgres pgKind))) ->
m
( [(S.Alias, S.SQLExp)], -- Order by Extractors
SelectSorting,
Maybe S.SQLExp -- The cursor expression
)
processOrderByItems sourcePrefix' fieldAlias' similarArrayFields distOnCols = \case
Nothing -> pure ([], NoSorting $ applyDistinctOnAtBase <$> distOnCols, Nothing)
Just orderByItems -> do
orderByItemExps <- forM orderByItems processAnnOrderByItem
let (sorting, distinctOnExtractors) = generateSorting orderByItemExps
orderByExtractors = concat $ toList $ map snd . toList <$> orderByItemExps
cursor = mkCursorExp $ toList orderByItemExps
pure (orderByExtractors <> distinctOnExtractors, sorting, Just cursor)
where
processAnnOrderByItem ::
AnnotatedOrderByItem ('Postgres pgKind) ->
m (OrderByItemG ('Postgres pgKind) (AnnotatedOrderByElement ('Postgres pgKind) (SQLExpression ('Postgres pgKind)), (S.Alias, SQLExpression ('Postgres pgKind))))
processAnnOrderByItem orderByItem =
forM orderByItem $ \ordByCol ->
(ordByCol,)
<$> processAnnotatedOrderByElement sourcePrefix' fieldAlias' ordByCol
processAnnotatedOrderByElement ::
Identifier -> FieldName -> AnnotatedOrderByElement ('Postgres pgKind) S.SQLExp -> m (S.Alias, (SQLExpression ('Postgres pgKind)))
processAnnotatedOrderByElement sourcePrefix fieldAlias annObCol = do
let ordByAlias = mkAnnOrderByAlias sourcePrefix fieldAlias similarArrayFields annObCol
(ordByAlias,) <$> case annObCol of
AOCColumn pgColInfo ->
pure $
S.mkQIdenExp (mkBaseTableAlias sourcePrefix) $ toIdentifier $ ciColumn pgColInfo
AOCObjectRelation relInfo relFilter rest -> withWriteObjectRelation $ do
server: fix the nullability of object relationships (fix hasura/graphql-engine#7201) When adding object relationships, we set the nullability of the generated GraphQL field based on whether the database backend enforces that the referenced data always exists. For manual relationships (corresponding to `manual_configuration`), the database backend is unaware of any relationship between data, and hence such fields are always set to be nullable. For relationships generated from foreign key constraints (corresponding to `foreign_key_constraint_on`), we distinguish between two cases: 1. The "forward" object relationship from a referencing table (i.e. which has the foreign key constraint) to a referenced table. This should be set to be non-nullable when all referencing columns are non-nullable. But in fact, it used to set it to be non-nullable if *any* referencing column is non-nullable, which is only correct in Postgres when `MATCH FULL` is set (a flag we don't consider). This fixes that by changing a boolean conjunction to a disjunction. 2. The "reverse" object relationship from a referenced table to a referencing table which has the foreign key constraint. This should always be set to be nullable. But in fact, it used to always be set to non-nullable, as was reported in hasura/graphql-engine#7201. This fixes that. Moreover, we have moved the computation of the nullability from `Hasura.RQL.DDL.Relationship` to `Hasura.GraphQL.Schema.Select`: this nullability used to be passed through the `riIsNullable` field of `RelInfo`, but for array relationships this information is not actually used, and moreover the remaining fields of `RelInfo` are already enough to deduce the nullability. This also adds regression tests for both (1) and (2) above. https://github.com/hasura/graphql-engine-mono/pull/2159 GitOrigin-RevId: 617f12765614f49746d18d3368f41dfae2f3e6ca
2021-08-26 18:26:43 +03:00
let RelInfo relName _ colMapping relTable _ _ = relInfo
relSourcePrefix = mkObjectRelationTableAlias sourcePrefix relName
fieldName = mkOrderByFieldName relName
(relOrderByAlias, relOrdByExp) <-
processAnnotatedOrderByElement relSourcePrefix fieldName rest
let selectSource =
ObjectSelectSource
relSourcePrefix
(S.FISimple relTable Nothing)
(toSQLBoolExp (S.QualTable relTable) relFilter)
relSource = ObjectRelationSource relName colMapping selectSource
pure
( relSource,
HM.singleton relOrderByAlias relOrdByExp,
S.mkQIdenExp relSourcePrefix relOrderByAlias
)
AOCArrayAggregation relInfo relFilter aggOrderBy -> withWriteArrayRelation $ do
server: fix the nullability of object relationships (fix hasura/graphql-engine#7201) When adding object relationships, we set the nullability of the generated GraphQL field based on whether the database backend enforces that the referenced data always exists. For manual relationships (corresponding to `manual_configuration`), the database backend is unaware of any relationship between data, and hence such fields are always set to be nullable. For relationships generated from foreign key constraints (corresponding to `foreign_key_constraint_on`), we distinguish between two cases: 1. The "forward" object relationship from a referencing table (i.e. which has the foreign key constraint) to a referenced table. This should be set to be non-nullable when all referencing columns are non-nullable. But in fact, it used to set it to be non-nullable if *any* referencing column is non-nullable, which is only correct in Postgres when `MATCH FULL` is set (a flag we don't consider). This fixes that by changing a boolean conjunction to a disjunction. 2. The "reverse" object relationship from a referenced table to a referencing table which has the foreign key constraint. This should always be set to be nullable. But in fact, it used to always be set to non-nullable, as was reported in hasura/graphql-engine#7201. This fixes that. Moreover, we have moved the computation of the nullability from `Hasura.RQL.DDL.Relationship` to `Hasura.GraphQL.Schema.Select`: this nullability used to be passed through the `riIsNullable` field of `RelInfo`, but for array relationships this information is not actually used, and moreover the remaining fields of `RelInfo` are already enough to deduce the nullability. This also adds regression tests for both (1) and (2) above. https://github.com/hasura/graphql-engine-mono/pull/2159 GitOrigin-RevId: 617f12765614f49746d18d3368f41dfae2f3e6ca
2021-08-26 18:26:43 +03:00
let RelInfo relName _ colMapping relTable _ _ = relInfo
fieldName = mkOrderByFieldName relName
relSourcePrefix =
mkArrayRelationSourcePrefix
sourcePrefix
fieldAlias
similarArrayFields
fieldName
relAlias = mkArrayRelationAlias fieldAlias similarArrayFields fieldName
(topExtractor, fields) = mkAggregateOrderByExtractorAndFields aggOrderBy
selectSource =
SelectSource
relSourcePrefix
(S.FISimple relTable Nothing)
(toSQLBoolExp (S.QualTable relTable) relFilter)
noSortingAndSlicing
relSource = ArrayRelationSource relAlias colMapping selectSource
pure
( relSource,
topExtractor,
HM.fromList $ aggregateFieldsToExtractorExps relSourcePrefix fields,
S.mkQIdenExp relSourcePrefix (mkAggregateOrderByAlias aggOrderBy)
)
AOCComputedField ComputedFieldOrderBy {..} ->
case _cfobOrderByElement of
CFOBEScalar _ -> do
let functionArgs = fromTableRowArgs sourcePrefix _cfobFunctionArgsExp
functionExp = S.FunctionExp _cfobFunction functionArgs Nothing
pure $ S.SEFunction functionExp
CFOBETableAggregation _ tableFilter aggOrderBy -> withWriteComputedFieldTableSet $ do
let fieldName = mkOrderByFieldName _cfobName
computedFieldSourcePrefix = mkComputedFieldTableAlias sourcePrefix fieldName
(topExtractor, fields) = mkAggregateOrderByExtractorAndFields aggOrderBy
fromItem =
selectFromToFromItem sourcePrefix $
FromFunction _cfobFunction _cfobFunctionArgsExp Nothing
functionQual = S.QualifiedIdentifier (functionToIdentifier _cfobFunction) Nothing
selectSource =
SelectSource
computedFieldSourcePrefix
fromItem
(toSQLBoolExp functionQual tableFilter)
noSortingAndSlicing
source = ComputedFieldTableSetSource fieldName selectSource
pure
( source,
topExtractor,
HM.fromList $ aggregateFieldsToExtractorExps computedFieldSourcePrefix fields,
S.mkQIdenExp computedFieldSourcePrefix (mkAggregateOrderByAlias aggOrderBy)
)
generateSorting ::
NE.NonEmpty (OrderByItemG ('Postgres pgKind) (AnnotatedOrderByElement ('Postgres pgKind) (SQLExpression ('Postgres pgKind)), (S.Alias, SQLExpression ('Postgres pgKind)))) ->
( SelectSorting,
[(S.Alias, SQLExpression ('Postgres pgKind))] -- 'distinct on' column extractors
)
generateSorting orderByExps@(firstOrderBy NE.:| restOrderBys) =
case fst $ obiColumn firstOrderBy of
AOCColumn columnInfo ->
-- If rest order by expressions are all columns then apply order by clause at base selection.
if all (isJust . getColumnOrderBy . obiColumn) restOrderBys
then -- Collect column order by expressions from the rest.
let restColumnOrderBys = mapMaybe (sequenceA . (getColumnOrderBy <$>)) restOrderBys
firstColumnOrderBy = firstOrderBy {obiColumn = columnInfo}
in sortAtNodeAndBase $ firstColumnOrderBy NE.:| restColumnOrderBys
else -- Else rest order by expressions contain atleast one non-column order by.
-- So, apply order by clause at node selection.
sortOnlyAtNode
_ ->
-- First order by expression is non-column. So, apply order by clause at node selection.
sortOnlyAtNode
where
getColumnOrderBy = (^? _AOCColumn) . fst
(nodeOrderBy, nodeDistinctOn, nodeDistinctOnExtractors) =
let toOrderByExp orderByItemExp =
let OrderByItemG obTyM expAlias obNullsM = fst . snd <$> orderByItemExp
in S.OrderByItem (S.SEIdentifier $ toIdentifier expAlias) obTyM obNullsM
orderByExp = S.OrderByExp $ toOrderByExp <$> orderByExps
(maybeDistOn, distOnExtrs) = NE.unzip $ applyDistinctOnAtNode sourcePrefix' <$> distOnCols
in (orderByExp, maybeDistOn, fromMaybe [] distOnExtrs)
sortOnlyAtNode =
(Sorting $ ASorting (nodeOrderBy, nodeDistinctOn) Nothing, nodeDistinctOnExtractors)
sortAtNodeAndBase baseColumnOrderBys =
let mkBaseOrderByItem (OrderByItemG orderByType columnInfo nullsOrder) =
S.OrderByItem
(S.SEIdentifier $ toIdentifier $ ciColumn columnInfo)
orderByType
nullsOrder
baseOrderByExp = S.OrderByExp $ mkBaseOrderByItem <$> baseColumnOrderBys
baseDistOnExp = applyDistinctOnAtBase <$> distOnCols
sorting = Sorting $ ASorting (nodeOrderBy, nodeDistinctOn) $ Just (baseOrderByExp, baseDistOnExp)
in (sorting, nodeDistinctOnExtractors)
mkCursorExp ::
[OrderByItemG ('Postgres pgKind) (AnnotatedOrderByElement ('Postgres pgKind) (SQLExpression ('Postgres pgKind)), (S.Alias, SQLExpression ('Postgres pgKind)))] ->
S.SQLExp
mkCursorExp orderByItemExps =
S.applyJsonBuildObj $
flip concatMap orderByItemExps $
\orderByItemExp ->
let OrderByItemG _ (annObCol, (_, valExp)) _ = orderByItemExp
in annObColToJSONField valExp annObCol
where
mkAggOrderByValExp valExp = \case
AAOCount -> [S.SELit "count", valExp]
AAOOp opText colInfo ->
[ S.SELit opText,
S.applyJsonBuildObj [S.SELit $ getPGColTxt $ ciColumn colInfo, valExp]
]
annObColToJSONField valExp = \case
AOCColumn pgCol -> [S.SELit $ getPGColTxt $ ciColumn pgCol, valExp]
AOCObjectRelation relInfo _ obCol ->
[ S.SELit $ relNameToTxt $ riName relInfo,
S.applyJsonBuildObj $ annObColToJSONField valExp obCol
]
AOCArrayAggregation relInfo _ aggOrderBy ->
[ S.SELit $ relNameToTxt (riName relInfo) <> "_aggregate",
S.applyJsonBuildObj $ mkAggOrderByValExp valExp aggOrderBy
]
AOCComputedField cfOrderBy ->
let fieldNameText = computedFieldNameToText $ _cfobName cfOrderBy
in case _cfobOrderByElement cfOrderBy of
CFOBEScalar _ -> [S.SELit fieldNameText, valExp]
CFOBETableAggregation _ _ aggOrderBy ->
[ S.SELit $ fieldNameText <> "_aggregate",
S.applyJsonBuildObj $ mkAggOrderByValExp valExp aggOrderBy
]
aggregateFieldsToExtractorExps ::
Identifier -> AggregateFields ('Postgres pgKind) -> [(S.Alias, S.SQLExp)]
aggregateFieldsToExtractorExps sourcePrefix aggregateFields =
flip concatMap aggregateFields $ \(_, field) ->
case field of
AFCount cty -> case cty of
S.CTStar -> []
S.CTSimple cols -> colsToExps cols
S.CTDistinct cols -> colsToExps cols
AFOp aggOp -> aggOpToExps aggOp
AFExp _ -> []
where
colsToExps = fmap mkColExp
aggOpToExps = mapMaybe colToMaybeExp . _aoFields
colToMaybeExp = \case
(_, CFCol col _) -> Just $ mkColExp col
_ -> Nothing
mkColExp c =
let qualCol = S.mkQIdenExp (mkBaseTableAlias sourcePrefix) (toIdentifier c)
colAls = toIdentifier c
in (S.Alias colAls, qualCol)
[Preview] Inherited roles for postgres read queries fixes #3868 docker image - `hasura/graphql-engine:inherited-roles-preview-48b73a2de` Note: To be able to use the inherited roles feature, the graphql-engine should be started with the env variable `HASURA_GRAPHQL_EXPERIMENTAL_FEATURES` set to `inherited_roles`. Introduction ------------ This PR implements the idea of multiple roles as presented in this [paper](https://www.microsoft.com/en-us/research/wp-content/uploads/2016/02/FGALanguageICDE07.pdf). The multiple roles feature in this PR can be used via inherited roles. An inherited role is a role which can be created by combining multiple singular roles. For example, if there are two roles `author` and `editor` configured in the graphql-engine, then we can create a inherited role with the name of `combined_author_editor` role which will combine the select permissions of the `author` and `editor` roles and then make GraphQL queries using the `combined_author_editor`. How are select permissions of different roles are combined? ------------------------------------------------------------ A select permission includes 5 things: 1. Columns accessible to the role 2. Row selection filter 3. Limit 4. Allow aggregation 5. Scalar computed fields accessible to the role Suppose there are two roles, `role1` gives access to the `address` column with row filter `P1` and `role2` gives access to both the `address` and the `phone` column with row filter `P2` and we create a new role `combined_roles` which combines `role1` and `role2`. Let's say the following GraphQL query is queried with the `combined_roles` role. ```graphql query { employees { address phone } } ``` This will translate to the following SQL query: ```sql select (case when (P1 or P2) then address else null end) as address, (case when P2 then phone else null end) as phone from employee where (P1 or P2) ``` The other parameters of the select permission will be combined in the following manner: 1. Limit - Minimum of the limits will be the limit of the inherited role 2. Allow aggregations - If any of the role allows aggregation, then the inherited role will allow aggregation 3. Scalar computed fields - same as table column fields, as in the above example APIs for inherited roles: ---------------------- 1. `add_inherited_role` `add_inherited_role` is the [metadata API](https://hasura.io/docs/1.0/graphql/core/api-reference/index.html#schema-metadata-api) to create a new inherited role. It accepts two arguments `role_name`: the name of the inherited role to be added (String) `role_set`: list of roles that need to be combined (Array of Strings) Example: ```json { "type": "add_inherited_role", "args": { "role_name":"combined_user", "role_set":[ "user", "user1" ] } } ``` After adding the inherited role, the inherited role can be used like single roles like earlier Note: An inherited role can only be created with non-inherited/singular roles. 2. `drop_inherited_role` The `drop_inherited_role` API accepts the name of the inherited role and drops it from the metadata. It accepts a single argument: `role_name`: name of the inherited role to be dropped Example: ```json { "type": "drop_inherited_role", "args": { "role_name":"combined_user" } } ``` Metadata --------- The derived roles metadata will be included under the `experimental_features` key while exporting the metadata. ```json { "experimental_features": { "derived_roles": [ { "role_name": "manager_is_employee_too", "role_set": [ "employee", "manager" ] } ] } } ``` Scope ------ Only postgres queries and subscriptions are supported in this PR. Important points: ----------------- 1. All columns exposed to an inherited role will be marked as `nullable`, this is done so that cell value nullification can be done. TODOs ------- - [ ] Tests - [ ] Test a GraphQL query running with a inherited role without enabling inherited roles in experimental features - [] Tests for aggregate queries, limit, computed fields, functions, subscriptions (?) - [ ] Introspection test with a inherited role (nullability changes in a inherited role) - [ ] Docs - [ ] Changelog Co-authored-by: Vamshi Surabhi <6562944+0x777@users.noreply.github.com> GitOrigin-RevId: 3b8ee1e11f5ceca80fe294f8c074d42fbccfec63
2021-03-08 14:14:13 +03:00
{- Note: [SQL generation for inherited roles]
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
When a query is executed by an inherited role, each column may contain a predicate
(AnnColumnCaseBoolExp ('Postgres pgKind) SQLExp) along with it. The predicate is then
[Preview] Inherited roles for postgres read queries fixes #3868 docker image - `hasura/graphql-engine:inherited-roles-preview-48b73a2de` Note: To be able to use the inherited roles feature, the graphql-engine should be started with the env variable `HASURA_GRAPHQL_EXPERIMENTAL_FEATURES` set to `inherited_roles`. Introduction ------------ This PR implements the idea of multiple roles as presented in this [paper](https://www.microsoft.com/en-us/research/wp-content/uploads/2016/02/FGALanguageICDE07.pdf). The multiple roles feature in this PR can be used via inherited roles. An inherited role is a role which can be created by combining multiple singular roles. For example, if there are two roles `author` and `editor` configured in the graphql-engine, then we can create a inherited role with the name of `combined_author_editor` role which will combine the select permissions of the `author` and `editor` roles and then make GraphQL queries using the `combined_author_editor`. How are select permissions of different roles are combined? ------------------------------------------------------------ A select permission includes 5 things: 1. Columns accessible to the role 2. Row selection filter 3. Limit 4. Allow aggregation 5. Scalar computed fields accessible to the role Suppose there are two roles, `role1` gives access to the `address` column with row filter `P1` and `role2` gives access to both the `address` and the `phone` column with row filter `P2` and we create a new role `combined_roles` which combines `role1` and `role2`. Let's say the following GraphQL query is queried with the `combined_roles` role. ```graphql query { employees { address phone } } ``` This will translate to the following SQL query: ```sql select (case when (P1 or P2) then address else null end) as address, (case when P2 then phone else null end) as phone from employee where (P1 or P2) ``` The other parameters of the select permission will be combined in the following manner: 1. Limit - Minimum of the limits will be the limit of the inherited role 2. Allow aggregations - If any of the role allows aggregation, then the inherited role will allow aggregation 3. Scalar computed fields - same as table column fields, as in the above example APIs for inherited roles: ---------------------- 1. `add_inherited_role` `add_inherited_role` is the [metadata API](https://hasura.io/docs/1.0/graphql/core/api-reference/index.html#schema-metadata-api) to create a new inherited role. It accepts two arguments `role_name`: the name of the inherited role to be added (String) `role_set`: list of roles that need to be combined (Array of Strings) Example: ```json { "type": "add_inherited_role", "args": { "role_name":"combined_user", "role_set":[ "user", "user1" ] } } ``` After adding the inherited role, the inherited role can be used like single roles like earlier Note: An inherited role can only be created with non-inherited/singular roles. 2. `drop_inherited_role` The `drop_inherited_role` API accepts the name of the inherited role and drops it from the metadata. It accepts a single argument: `role_name`: name of the inherited role to be dropped Example: ```json { "type": "drop_inherited_role", "args": { "role_name":"combined_user" } } ``` Metadata --------- The derived roles metadata will be included under the `experimental_features` key while exporting the metadata. ```json { "experimental_features": { "derived_roles": [ { "role_name": "manager_is_employee_too", "role_set": [ "employee", "manager" ] } ] } } ``` Scope ------ Only postgres queries and subscriptions are supported in this PR. Important points: ----------------- 1. All columns exposed to an inherited role will be marked as `nullable`, this is done so that cell value nullification can be done. TODOs ------- - [ ] Tests - [ ] Test a GraphQL query running with a inherited role without enabling inherited roles in experimental features - [] Tests for aggregate queries, limit, computed fields, functions, subscriptions (?) - [ ] Introspection test with a inherited role (nullability changes in a inherited role) - [ ] Docs - [ ] Changelog Co-authored-by: Vamshi Surabhi <6562944+0x777@users.noreply.github.com> GitOrigin-RevId: 3b8ee1e11f5ceca80fe294f8c074d42fbccfec63
2021-03-08 14:14:13 +03:00
converted to a BoolExp, which will be used to check if the said column should
be nullified. For example,
Suppose there are two roles, role1 gives access only to the `addr` column with
row filter P1 and role2 gives access to both addr and phone column with row
filter P2. The `OR`ing of the predicates will have already been done while
the schema has been generated. The SQL generated will look like this:
select
(case when (P1 or P2) then addr else null end) as addr,
(case when P2 then phone else null end) as phone
from employee
where (P1 or P2)
-}
class PostgresAnnotatedFieldJSON (pgKind :: PostgresKind) where
annRowToJson :: FieldName -> [(FieldName, S.SQLExp)] -> (S.Alias, S.SQLExp)
instance PostgresAnnotatedFieldJSON 'Vanilla where
annRowToJson fieldAlias fieldExps =
-- postgres ignores anything beyond 63 chars for an iden
-- in this case, we'll need to use json_build_object function
-- json_build_object is slower than row_to_json hence it is only
-- used when needed
if any ((> 63) . T.length . getFieldNameTxt . fst) fieldExps
then withJsonBuildObj fieldAlias $ concatMap toJsonBuildObjectExps fieldExps
else withRowToJSON fieldAlias $ map toRowToJsonExtr fieldExps
where
toJsonBuildObjectExps (fieldName, fieldExp) =
[S.SELit $ getFieldNameTxt fieldName, fieldExp]
toRowToJsonExtr (fieldName, fieldExp) =
S.Extractor fieldExp $ Just $ S.toAlias fieldName
instance PostgresAnnotatedFieldJSON 'Citus where
annRowToJson fieldAlias fieldExps =
-- Due to the restrictions Citus imposes on joins between tables of various
-- distribution types we cannot use row_to_json and have to only rely on
-- json_build_object.
withJsonBuildObj fieldAlias $ concatMap toJsonBuildObjectExps fieldExps
where
toJsonBuildObjectExps (fieldName, fieldExp) =
[S.SELit $ getFieldNameTxt fieldName, fieldExp]
processAnnFields ::
forall pgKind m.
( MonadReader StringifyNumbers m,
MonadWriter JoinTree m,
Backend ('Postgres pgKind),
PostgresAnnotatedFieldJSON pgKind
) =>
Identifier ->
FieldName ->
SimilarArrayFields ->
AnnFields ('Postgres pgKind) ->
m (S.Alias, S.SQLExp)
processAnnFields sourcePrefix fieldAlias similarArrFields annFields = do
fieldExps <- forM annFields $ \(fieldName, field) ->
(fieldName,)
<$> case field of
AFExpression t -> pure $ S.SELit t
AFNodeId _ tn pKeys -> pure $ mkNodeId tn pKeys
AFColumn c -> toSQLCol c
AFObjectRelation objSel -> withWriteObjectRelation $ do
let AnnRelationSelectG relName relMapping annObjSel = objSel
AnnObjectSelectG objAnnFields tableFrom tableFilter = annObjSel
objRelSourcePrefix = mkObjectRelationTableAlias sourcePrefix relName
sourcePrefixes = mkSourcePrefixes objRelSourcePrefix
annFieldsExtr <- processAnnFields (_pfThis sourcePrefixes) fieldName HM.empty objAnnFields
let selectSource =
ObjectSelectSource
(_pfThis sourcePrefixes)
(S.FISimple tableFrom Nothing)
(toSQLBoolExp (S.QualTable tableFrom) tableFilter)
objRelSource = ObjectRelationSource relName relMapping selectSource
pure
( objRelSource,
HM.fromList [annFieldsExtr],
S.mkQIdenExp objRelSourcePrefix fieldName
)
AFArrayRelation arrSel -> do
let arrRelSourcePrefix = mkArrayRelationSourcePrefix sourcePrefix fieldAlias similarArrFields fieldName
arrRelAlias = mkArrayRelationAlias fieldAlias similarArrFields fieldName
processArrayRelation (mkSourcePrefixes arrRelSourcePrefix) fieldName arrRelAlias arrSel
pure $ S.mkQIdenExp arrRelSourcePrefix fieldName
AFComputedField _ _ (CFSScalar scalar caseBoolExpMaybe) -> do
computedFieldSQLExp <- fromScalarComputedField scalar
-- The computed field is conditionally outputed depending
-- on the presence of `caseBoolExpMaybe` and the value it
-- evaluates to. `caseBoolExpMaybe` will be set only in the
-- case of an inherited role.
-- See [SQL generation for inherited role]
case caseBoolExpMaybe of
Nothing -> pure computedFieldSQLExp
Just caseBoolExp ->
let boolExp =
S.simplifyBoolExp $
toSQLBoolExp (S.QualifiedIdentifier baseTableIdentifier Nothing) $
_accColCaseBoolExpField <$> caseBoolExp
in pure $ S.SECond boolExp computedFieldSQLExp S.SENull
AFComputedField _ _ (CFSTable selectTy sel) -> withWriteComputedFieldTableSet $ do
let computedFieldSourcePrefix =
mkComputedFieldTableAlias sourcePrefix fieldName
(selectSource, nodeExtractors) <-
processAnnSimpleSelect
(mkSourcePrefixes computedFieldSourcePrefix)
fieldName
PLSQNotRequired
sel
let computedFieldTableSetSource = ComputedFieldTableSetSource fieldName selectSource
extractor =
asJsonAggExtr selectTy (S.toAlias fieldName) PLSQNotRequired $
orderByForJsonAgg selectSource
pure
( computedFieldTableSetSource,
extractor,
nodeExtractors,
S.mkQIdenExp computedFieldSourcePrefix fieldName
)
pure $ annRowToJson @pgKind fieldAlias fieldExps
where
mkSourcePrefixes newPrefix = SourcePrefixes newPrefix sourcePrefix
[Preview] Inherited roles for postgres read queries fixes #3868 docker image - `hasura/graphql-engine:inherited-roles-preview-48b73a2de` Note: To be able to use the inherited roles feature, the graphql-engine should be started with the env variable `HASURA_GRAPHQL_EXPERIMENTAL_FEATURES` set to `inherited_roles`. Introduction ------------ This PR implements the idea of multiple roles as presented in this [paper](https://www.microsoft.com/en-us/research/wp-content/uploads/2016/02/FGALanguageICDE07.pdf). The multiple roles feature in this PR can be used via inherited roles. An inherited role is a role which can be created by combining multiple singular roles. For example, if there are two roles `author` and `editor` configured in the graphql-engine, then we can create a inherited role with the name of `combined_author_editor` role which will combine the select permissions of the `author` and `editor` roles and then make GraphQL queries using the `combined_author_editor`. How are select permissions of different roles are combined? ------------------------------------------------------------ A select permission includes 5 things: 1. Columns accessible to the role 2. Row selection filter 3. Limit 4. Allow aggregation 5. Scalar computed fields accessible to the role Suppose there are two roles, `role1` gives access to the `address` column with row filter `P1` and `role2` gives access to both the `address` and the `phone` column with row filter `P2` and we create a new role `combined_roles` which combines `role1` and `role2`. Let's say the following GraphQL query is queried with the `combined_roles` role. ```graphql query { employees { address phone } } ``` This will translate to the following SQL query: ```sql select (case when (P1 or P2) then address else null end) as address, (case when P2 then phone else null end) as phone from employee where (P1 or P2) ``` The other parameters of the select permission will be combined in the following manner: 1. Limit - Minimum of the limits will be the limit of the inherited role 2. Allow aggregations - If any of the role allows aggregation, then the inherited role will allow aggregation 3. Scalar computed fields - same as table column fields, as in the above example APIs for inherited roles: ---------------------- 1. `add_inherited_role` `add_inherited_role` is the [metadata API](https://hasura.io/docs/1.0/graphql/core/api-reference/index.html#schema-metadata-api) to create a new inherited role. It accepts two arguments `role_name`: the name of the inherited role to be added (String) `role_set`: list of roles that need to be combined (Array of Strings) Example: ```json { "type": "add_inherited_role", "args": { "role_name":"combined_user", "role_set":[ "user", "user1" ] } } ``` After adding the inherited role, the inherited role can be used like single roles like earlier Note: An inherited role can only be created with non-inherited/singular roles. 2. `drop_inherited_role` The `drop_inherited_role` API accepts the name of the inherited role and drops it from the metadata. It accepts a single argument: `role_name`: name of the inherited role to be dropped Example: ```json { "type": "drop_inherited_role", "args": { "role_name":"combined_user" } } ``` Metadata --------- The derived roles metadata will be included under the `experimental_features` key while exporting the metadata. ```json { "experimental_features": { "derived_roles": [ { "role_name": "manager_is_employee_too", "role_set": [ "employee", "manager" ] } ] } } ``` Scope ------ Only postgres queries and subscriptions are supported in this PR. Important points: ----------------- 1. All columns exposed to an inherited role will be marked as `nullable`, this is done so that cell value nullification can be done. TODOs ------- - [ ] Tests - [ ] Test a GraphQL query running with a inherited role without enabling inherited roles in experimental features - [] Tests for aggregate queries, limit, computed fields, functions, subscriptions (?) - [ ] Introspection test with a inherited role (nullability changes in a inherited role) - [ ] Docs - [ ] Changelog Co-authored-by: Vamshi Surabhi <6562944+0x777@users.noreply.github.com> GitOrigin-RevId: 3b8ee1e11f5ceca80fe294f8c074d42fbccfec63
2021-03-08 14:14:13 +03:00
baseTableIdentifier = mkBaseTableAlias sourcePrefix
toSQLCol :: AnnColumnField ('Postgres pgKind) S.SQLExp -> m S.SQLExp
toSQLCol (AnnColumnField col typ asText colOpM caseBoolExpMaybe) = do
strfyNum <- ask
[Preview] Inherited roles for postgres read queries fixes #3868 docker image - `hasura/graphql-engine:inherited-roles-preview-48b73a2de` Note: To be able to use the inherited roles feature, the graphql-engine should be started with the env variable `HASURA_GRAPHQL_EXPERIMENTAL_FEATURES` set to `inherited_roles`. Introduction ------------ This PR implements the idea of multiple roles as presented in this [paper](https://www.microsoft.com/en-us/research/wp-content/uploads/2016/02/FGALanguageICDE07.pdf). The multiple roles feature in this PR can be used via inherited roles. An inherited role is a role which can be created by combining multiple singular roles. For example, if there are two roles `author` and `editor` configured in the graphql-engine, then we can create a inherited role with the name of `combined_author_editor` role which will combine the select permissions of the `author` and `editor` roles and then make GraphQL queries using the `combined_author_editor`. How are select permissions of different roles are combined? ------------------------------------------------------------ A select permission includes 5 things: 1. Columns accessible to the role 2. Row selection filter 3. Limit 4. Allow aggregation 5. Scalar computed fields accessible to the role Suppose there are two roles, `role1` gives access to the `address` column with row filter `P1` and `role2` gives access to both the `address` and the `phone` column with row filter `P2` and we create a new role `combined_roles` which combines `role1` and `role2`. Let's say the following GraphQL query is queried with the `combined_roles` role. ```graphql query { employees { address phone } } ``` This will translate to the following SQL query: ```sql select (case when (P1 or P2) then address else null end) as address, (case when P2 then phone else null end) as phone from employee where (P1 or P2) ``` The other parameters of the select permission will be combined in the following manner: 1. Limit - Minimum of the limits will be the limit of the inherited role 2. Allow aggregations - If any of the role allows aggregation, then the inherited role will allow aggregation 3. Scalar computed fields - same as table column fields, as in the above example APIs for inherited roles: ---------------------- 1. `add_inherited_role` `add_inherited_role` is the [metadata API](https://hasura.io/docs/1.0/graphql/core/api-reference/index.html#schema-metadata-api) to create a new inherited role. It accepts two arguments `role_name`: the name of the inherited role to be added (String) `role_set`: list of roles that need to be combined (Array of Strings) Example: ```json { "type": "add_inherited_role", "args": { "role_name":"combined_user", "role_set":[ "user", "user1" ] } } ``` After adding the inherited role, the inherited role can be used like single roles like earlier Note: An inherited role can only be created with non-inherited/singular roles. 2. `drop_inherited_role` The `drop_inherited_role` API accepts the name of the inherited role and drops it from the metadata. It accepts a single argument: `role_name`: name of the inherited role to be dropped Example: ```json { "type": "drop_inherited_role", "args": { "role_name":"combined_user" } } ``` Metadata --------- The derived roles metadata will be included under the `experimental_features` key while exporting the metadata. ```json { "experimental_features": { "derived_roles": [ { "role_name": "manager_is_employee_too", "role_set": [ "employee", "manager" ] } ] } } ``` Scope ------ Only postgres queries and subscriptions are supported in this PR. Important points: ----------------- 1. All columns exposed to an inherited role will be marked as `nullable`, this is done so that cell value nullification can be done. TODOs ------- - [ ] Tests - [ ] Test a GraphQL query running with a inherited role without enabling inherited roles in experimental features - [] Tests for aggregate queries, limit, computed fields, functions, subscriptions (?) - [ ] Introspection test with a inherited role (nullability changes in a inherited role) - [ ] Docs - [ ] Changelog Co-authored-by: Vamshi Surabhi <6562944+0x777@users.noreply.github.com> GitOrigin-RevId: 3b8ee1e11f5ceca80fe294f8c074d42fbccfec63
2021-03-08 14:14:13 +03:00
let sqlExpression =
withColumnOp colOpM $
S.mkQIdenExp baseTableIdentifier col
[Preview] Inherited roles for postgres read queries fixes #3868 docker image - `hasura/graphql-engine:inherited-roles-preview-48b73a2de` Note: To be able to use the inherited roles feature, the graphql-engine should be started with the env variable `HASURA_GRAPHQL_EXPERIMENTAL_FEATURES` set to `inherited_roles`. Introduction ------------ This PR implements the idea of multiple roles as presented in this [paper](https://www.microsoft.com/en-us/research/wp-content/uploads/2016/02/FGALanguageICDE07.pdf). The multiple roles feature in this PR can be used via inherited roles. An inherited role is a role which can be created by combining multiple singular roles. For example, if there are two roles `author` and `editor` configured in the graphql-engine, then we can create a inherited role with the name of `combined_author_editor` role which will combine the select permissions of the `author` and `editor` roles and then make GraphQL queries using the `combined_author_editor`. How are select permissions of different roles are combined? ------------------------------------------------------------ A select permission includes 5 things: 1. Columns accessible to the role 2. Row selection filter 3. Limit 4. Allow aggregation 5. Scalar computed fields accessible to the role Suppose there are two roles, `role1` gives access to the `address` column with row filter `P1` and `role2` gives access to both the `address` and the `phone` column with row filter `P2` and we create a new role `combined_roles` which combines `role1` and `role2`. Let's say the following GraphQL query is queried with the `combined_roles` role. ```graphql query { employees { address phone } } ``` This will translate to the following SQL query: ```sql select (case when (P1 or P2) then address else null end) as address, (case when P2 then phone else null end) as phone from employee where (P1 or P2) ``` The other parameters of the select permission will be combined in the following manner: 1. Limit - Minimum of the limits will be the limit of the inherited role 2. Allow aggregations - If any of the role allows aggregation, then the inherited role will allow aggregation 3. Scalar computed fields - same as table column fields, as in the above example APIs for inherited roles: ---------------------- 1. `add_inherited_role` `add_inherited_role` is the [metadata API](https://hasura.io/docs/1.0/graphql/core/api-reference/index.html#schema-metadata-api) to create a new inherited role. It accepts two arguments `role_name`: the name of the inherited role to be added (String) `role_set`: list of roles that need to be combined (Array of Strings) Example: ```json { "type": "add_inherited_role", "args": { "role_name":"combined_user", "role_set":[ "user", "user1" ] } } ``` After adding the inherited role, the inherited role can be used like single roles like earlier Note: An inherited role can only be created with non-inherited/singular roles. 2. `drop_inherited_role` The `drop_inherited_role` API accepts the name of the inherited role and drops it from the metadata. It accepts a single argument: `role_name`: name of the inherited role to be dropped Example: ```json { "type": "drop_inherited_role", "args": { "role_name":"combined_user" } } ``` Metadata --------- The derived roles metadata will be included under the `experimental_features` key while exporting the metadata. ```json { "experimental_features": { "derived_roles": [ { "role_name": "manager_is_employee_too", "role_set": [ "employee", "manager" ] } ] } } ``` Scope ------ Only postgres queries and subscriptions are supported in this PR. Important points: ----------------- 1. All columns exposed to an inherited role will be marked as `nullable`, this is done so that cell value nullification can be done. TODOs ------- - [ ] Tests - [ ] Test a GraphQL query running with a inherited role without enabling inherited roles in experimental features - [] Tests for aggregate queries, limit, computed fields, functions, subscriptions (?) - [ ] Introspection test with a inherited role (nullability changes in a inherited role) - [ ] Docs - [ ] Changelog Co-authored-by: Vamshi Surabhi <6562944+0x777@users.noreply.github.com> GitOrigin-RevId: 3b8ee1e11f5ceca80fe294f8c074d42fbccfec63
2021-03-08 14:14:13 +03:00
finalSQLExpression =
-- Check out [SQL generation for inherited role]
case caseBoolExpMaybe of
Nothing -> sqlExpression
[Preview] Inherited roles for postgres read queries fixes #3868 docker image - `hasura/graphql-engine:inherited-roles-preview-48b73a2de` Note: To be able to use the inherited roles feature, the graphql-engine should be started with the env variable `HASURA_GRAPHQL_EXPERIMENTAL_FEATURES` set to `inherited_roles`. Introduction ------------ This PR implements the idea of multiple roles as presented in this [paper](https://www.microsoft.com/en-us/research/wp-content/uploads/2016/02/FGALanguageICDE07.pdf). The multiple roles feature in this PR can be used via inherited roles. An inherited role is a role which can be created by combining multiple singular roles. For example, if there are two roles `author` and `editor` configured in the graphql-engine, then we can create a inherited role with the name of `combined_author_editor` role which will combine the select permissions of the `author` and `editor` roles and then make GraphQL queries using the `combined_author_editor`. How are select permissions of different roles are combined? ------------------------------------------------------------ A select permission includes 5 things: 1. Columns accessible to the role 2. Row selection filter 3. Limit 4. Allow aggregation 5. Scalar computed fields accessible to the role Suppose there are two roles, `role1` gives access to the `address` column with row filter `P1` and `role2` gives access to both the `address` and the `phone` column with row filter `P2` and we create a new role `combined_roles` which combines `role1` and `role2`. Let's say the following GraphQL query is queried with the `combined_roles` role. ```graphql query { employees { address phone } } ``` This will translate to the following SQL query: ```sql select (case when (P1 or P2) then address else null end) as address, (case when P2 then phone else null end) as phone from employee where (P1 or P2) ``` The other parameters of the select permission will be combined in the following manner: 1. Limit - Minimum of the limits will be the limit of the inherited role 2. Allow aggregations - If any of the role allows aggregation, then the inherited role will allow aggregation 3. Scalar computed fields - same as table column fields, as in the above example APIs for inherited roles: ---------------------- 1. `add_inherited_role` `add_inherited_role` is the [metadata API](https://hasura.io/docs/1.0/graphql/core/api-reference/index.html#schema-metadata-api) to create a new inherited role. It accepts two arguments `role_name`: the name of the inherited role to be added (String) `role_set`: list of roles that need to be combined (Array of Strings) Example: ```json { "type": "add_inherited_role", "args": { "role_name":"combined_user", "role_set":[ "user", "user1" ] } } ``` After adding the inherited role, the inherited role can be used like single roles like earlier Note: An inherited role can only be created with non-inherited/singular roles. 2. `drop_inherited_role` The `drop_inherited_role` API accepts the name of the inherited role and drops it from the metadata. It accepts a single argument: `role_name`: name of the inherited role to be dropped Example: ```json { "type": "drop_inherited_role", "args": { "role_name":"combined_user" } } ``` Metadata --------- The derived roles metadata will be included under the `experimental_features` key while exporting the metadata. ```json { "experimental_features": { "derived_roles": [ { "role_name": "manager_is_employee_too", "role_set": [ "employee", "manager" ] } ] } } ``` Scope ------ Only postgres queries and subscriptions are supported in this PR. Important points: ----------------- 1. All columns exposed to an inherited role will be marked as `nullable`, this is done so that cell value nullification can be done. TODOs ------- - [ ] Tests - [ ] Test a GraphQL query running with a inherited role without enabling inherited roles in experimental features - [] Tests for aggregate queries, limit, computed fields, functions, subscriptions (?) - [ ] Introspection test with a inherited role (nullability changes in a inherited role) - [ ] Docs - [ ] Changelog Co-authored-by: Vamshi Surabhi <6562944+0x777@users.noreply.github.com> GitOrigin-RevId: 3b8ee1e11f5ceca80fe294f8c074d42fbccfec63
2021-03-08 14:14:13 +03:00
Just caseBoolExp ->
let boolExp =
S.simplifyBoolExp $
toSQLBoolExp (S.QualifiedIdentifier baseTableIdentifier Nothing) $
_accColCaseBoolExpField <$> caseBoolExp
in S.SECond boolExp sqlExpression S.SENull
pure $ toJSONableExp strfyNum typ asText finalSQLExpression
fromScalarComputedField :: ComputedFieldScalarSelect ('Postgres pgKind) S.SQLExp -> m S.SQLExp
fromScalarComputedField computedFieldScalar = do
strfyNum <- ask
pure $
toJSONableExp strfyNum (ColumnScalar ty) False $
withColumnOp colOpM $
S.SEFunction $ S.FunctionExp fn (fromTableRowArgs sourcePrefix args) Nothing
where
ComputedFieldScalarSelect fn args ty colOpM = computedFieldScalar
withColumnOp :: Maybe (ColumnOp ('Postgres pgKind)) -> S.SQLExp -> S.SQLExp
withColumnOp colOpM sqlExp = case colOpM of
Nothing -> sqlExp
Just (ColumnOp opText cExp) -> S.mkSQLOpExp opText sqlExp cExp
mkNodeId :: QualifiedTable -> PrimaryKeyColumns ('Postgres pgKind) -> S.SQLExp
mkNodeId (QualifiedObject tableSchema tableName) pkeyColumns =
let columnInfoToSQLExp pgColumnInfo =
toJSONableExp LeaveNumbersAlone (ciType pgColumnInfo) False $
S.mkQIdenExp (mkBaseTableAlias sourcePrefix) $ ciColumn pgColumnInfo
in -- See Note [Relay Node id].
encodeBase64 $
flip S.SETyAnn S.textTypeAnn $
S.applyJsonBuildArray $
[ S.intToSQLExp $ nodeIdVersionInt currentNodeIdVersion,
S.SELit (getSchemaTxt tableSchema),
S.SELit (toTxt tableName)
]
<> map columnInfoToSQLExp (toList pkeyColumns)
injectJoinCond ::
-- | Join condition
S.BoolExp ->
-- | Where condition
S.BoolExp ->
-- | New where frag
S.WhereFrag
injectJoinCond joinCond whereCond =
S.WhereFrag $ S.simplifyBoolExp $ S.BEBin S.AndOp joinCond whereCond
mkJoinCond :: S.Alias -> HashMap PGCol PGCol -> S.BoolExp
mkJoinCond baseTablepfx colMapn =
foldl' (S.BEBin S.AndOp) (S.BELit True) $
flip map (HM.toList colMapn) $ \(lCol, rCol) ->
S.BECompare S.SEQ (S.mkQIdenExp baseTablepfx lCol) (S.mkSIdenExp rCol)
generateSQLSelect ::
-- | Pre join condition
S.BoolExp ->
SelectSource ->
SelectNode ->
S.Select
generateSQLSelect joinCondition selectSource selectNode =
S.mkSelect
{ S.selExtr = [S.Extractor e $ Just a | (a, e) <- HM.toList extractors],
S.selFrom = Just $ S.FromExp [joinedFrom],
S.selOrderBy = nodeOrderBy,
S.selLimit = S.LimitExp . S.intToSQLExp <$> _ssLimit nodeSlicing,
S.selOffset = S.OffsetExp . S.int64ToSQLExp <$> _ssOffset nodeSlicing,
S.selDistinct = nodeDistinctOn
}
where
SelectSource sourcePrefix fromItem whereExp sortAndSlice = selectSource
SelectNode extractors joinTree = selectNode
JoinTree objectRelations arrayRelations arrayConnections computedFields = joinTree
ApplySortingAndSlicing
(baseOrderBy, baseSlicing, baseDistinctOn)
(nodeOrderBy, nodeSlicing, nodeDistinctOn) = applySortingAndSlicing sortAndSlice
-- this is the table which is aliased as "sourcePrefix.base"
baseSelect =
S.mkSelect
{ S.selExtr = [S.Extractor (S.SEStar Nothing) Nothing],
S.selFrom = Just $ S.FromExp [fromItem],
S.selWhere = Just $ injectJoinCond joinCondition whereExp,
S.selOrderBy = baseOrderBy,
S.selLimit = S.LimitExp . S.intToSQLExp <$> _ssLimit baseSlicing,
S.selOffset = S.OffsetExp . S.int64ToSQLExp <$> _ssOffset baseSlicing,
S.selDistinct = baseDistinctOn
}
baseSelectAlias = S.Alias $ mkBaseTableAlias sourcePrefix
baseFromItem = S.mkSelFromItem baseSelect baseSelectAlias
-- function to create a joined from item from two from items
leftOuterJoin current new =
S.FIJoin $
S.JoinExpr current S.LeftOuter new $
S.JoinOn $ S.BELit True
-- this is the from eexp for the final select
joinedFrom :: S.FromItem
joinedFrom =
foldl' leftOuterJoin baseFromItem $
map objectRelationToFromItem (HM.toList objectRelations)
<> map arrayRelationToFromItem (HM.toList arrayRelations)
<> map arrayConnectionToFromItem (HM.toList arrayConnections)
<> map computedFieldToFromItem (HM.toList computedFields)
objectRelationToFromItem ::
(ObjectRelationSource, SelectNode) -> S.FromItem
objectRelationToFromItem (objectRelationSource, node) =
let ObjectRelationSource _ colMapping objectSelectSource = objectRelationSource
alias = S.Alias $ _ossPrefix objectSelectSource
source = objectSelectSourceToSelectSource objectSelectSource
select = generateSQLSelect (mkJoinCond baseSelectAlias colMapping) source node
in S.mkLateralFromItem select alias
arrayRelationToFromItem ::
(ArrayRelationSource, MultiRowSelectNode) -> S.FromItem
arrayRelationToFromItem (arrayRelationSource, arraySelectNode) =
let ArrayRelationSource _ colMapping source = arrayRelationSource
alias = S.Alias $ _ssPrefix source
select =
generateSQLSelectFromArrayNode source arraySelectNode $
mkJoinCond baseSelectAlias colMapping
in S.mkLateralFromItem select alias
arrayConnectionToFromItem ::
(ArrayConnectionSource, MultiRowSelectNode) -> S.FromItem
arrayConnectionToFromItem (arrayConnectionSource, arraySelectNode) =
let selectWith = connectionToSelectWith baseSelectAlias arrayConnectionSource arraySelectNode
alias = S.Alias $ _ssPrefix $ _acsSource arrayConnectionSource
in S.FISelectWith (S.Lateral True) selectWith alias
computedFieldToFromItem ::
(ComputedFieldTableSetSource, MultiRowSelectNode) -> S.FromItem
computedFieldToFromItem (computedFieldTableSource, node) =
let ComputedFieldTableSetSource _ source = computedFieldTableSource
internalSelect = generateSQLSelect (S.BELit True) source $ _mrsnSelectNode node
alias = S.Alias $ _ssPrefix source
select =
S.mkSelect
{ S.selExtr = _mrsnTopExtractors node,
S.selFrom = Just $ S.FromExp [S.mkSelFromItem internalSelect alias]
}
in S.mkLateralFromItem select alias
generateSQLSelectFromArrayNode ::
SelectSource ->
MultiRowSelectNode ->
S.BoolExp ->
S.Select
generateSQLSelectFromArrayNode selectSource arraySelectNode joinCondition =
S.mkSelect
{ S.selExtr = topExtractors,
S.selFrom = Just $ S.FromExp [selectFrom]
}
where
MultiRowSelectNode topExtractors selectNode = arraySelectNode
selectFrom =
S.mkSelFromItem
(generateSQLSelect joinCondition selectSource selectNode)
$ S.Alias $ _ssPrefix selectSource
mkAggregateSelect ::
forall pgKind.
( Backend ('Postgres pgKind),
PostgresAnnotatedFieldJSON pgKind
) =>
AnnAggregateSelect ('Postgres pgKind) ->
S.Select
mkAggregateSelect annAggSel =
let ((selectSource, nodeExtractors, topExtractor), joinTree) =
runWriter $
flip runReaderT strfyNum $
processAnnAggregateSelect sourcePrefixes rootFieldName annAggSel
selectNode = SelectNode nodeExtractors joinTree
arrayNode = MultiRowSelectNode [topExtractor] selectNode
in prefixNumToAliases $
generateSQLSelectFromArrayNode selectSource arrayNode $ S.BELit True
where
strfyNum = _asnStrfyNum annAggSel
rootFieldName = FieldName "root"
rootIdentifier = toIdentifier rootFieldName
sourcePrefixes = SourcePrefixes rootIdentifier rootIdentifier
mkSQLSelect ::
forall pgKind.
( Backend ('Postgres pgKind),
PostgresAnnotatedFieldJSON pgKind
) =>
JsonAggSelect ->
AnnSimpleSelect ('Postgres pgKind) ->
S.Select
allow custom mutations through actions (#3042) * basic doc for actions * custom_types, sync and async actions * switch to graphql-parser-hs on github * update docs * metadata import/export * webhook calls are now supported * relationships in sync actions * initialise.sql is now in sync with the migration file * fix metadata tests * allow specifying arguments of actions * fix blacklist check on check_build_worthiness job * track custom_types and actions related tables * handlers are now triggered on async actions * default to pgjson unless a field is involved in relationships, for generating definition list * use 'true' for action filter for non admin role * fix create_action_permission sql query * drop permissions when dropping an action * add a hdb_role view (and relationships) to fetch all roles in the system * rename 'webhook' key in action definition to 'handler' * allow templating actions wehook URLs with env vars * add 'update_action' /v1/query type * allow forwarding client headers by setting `forward_client_headers` in action definition * add 'headers' configuration in action definition * handle webhook error response based on status codes * support array relationships for custom types * implement single row mutation, see https://github.com/hasura/graphql-engine/issues/3731 * single row mutation: rename 'pk_columns' -> 'columns' and no-op refactor * use top level primary key inputs for delete_by_pk & account select permissions for single row mutations * use only REST semantics to resolve the webhook response * use 'pk_columns' instead of 'columns' for update_by_pk input * add python basic tests for single row mutations * add action context (name) in webhook payload * Async action response is accessible for non admin roles only if the request session vars equals to action's * clean nulls, empty arrays for actions, custom types in export metadata * async action mutation returns only the UUID of the action * unit tests for URL template parser * Basic sync actions python tests * fix output in async query & add async tests * add admin secret header in async actions python test * document async action architecture in Resolve/Action.hs file * support actions returning array of objects * tests for list type response actions * update docs with actions and custom types metadata API reference * update actions python tests as per #f8e1330 Co-authored-by: Tirumarai Selvan <tirumarai.selvan@gmail.com> Co-authored-by: Aravind Shankar <face11301@gmail.com> Co-authored-by: Rakesh Emmadi <12475069+rakeshkky@users.noreply.github.com>
2020-02-13 20:38:23 +03:00
mkSQLSelect jsonAggSelect annSel =
let permLimitSubQuery = PLSQNotRequired
((selectSource, nodeExtractors), joinTree) =
runWriter $
flip runReaderT strfyNum $
processAnnSimpleSelect sourcePrefixes rootFldName permLimitSubQuery annSel
selectNode = SelectNode nodeExtractors joinTree
topExtractor =
asJsonAggExtr jsonAggSelect rootFldAls permLimitSubQuery $
orderByForJsonAgg selectSource
arrayNode = MultiRowSelectNode [topExtractor] selectNode
in prefixNumToAliases $
generateSQLSelectFromArrayNode selectSource arrayNode $ S.BELit True
where
strfyNum = _asnStrfyNum annSel
rootFldIdentifier = toIdentifier rootFldName
sourcePrefixes = SourcePrefixes rootFldIdentifier rootFldIdentifier
rootFldName = FieldName "root"
rootFldAls = S.Alias $ toIdentifier rootFldName
mkConnectionSelect ::
forall pgKind.
( Backend ('Postgres pgKind),
PostgresAnnotatedFieldJSON pgKind
) =>
ConnectionSelect ('Postgres pgKind) Void S.SQLExp ->
S.SelectWithG S.Select
mkConnectionSelect connectionSelect =
let ((connectionSource, topExtractor, nodeExtractors), joinTree) =
runWriter $
flip runReaderT strfyNum $
processConnectionSelect
sourcePrefixes
rootFieldName
(S.Alias rootIdentifier)
mempty
connectionSelect
selectNode =
MultiRowSelectNode [topExtractor] $
SelectNode nodeExtractors joinTree
in prefixNumToAliasesSelectWith $
connectionToSelectWith (S.Alias rootIdentifier) connectionSource selectNode
where
strfyNum = _asnStrfyNum $ _csSelect connectionSelect
rootFieldName = FieldName "root"
rootIdentifier = toIdentifier rootFieldName
sourcePrefixes = SourcePrefixes rootIdentifier rootIdentifier
-- | First element extractor expression from given record set
-- For example:- To get first "id" column from given row set,
-- the function generates the SQL expression AS `(array_agg("id"))[1]`
mkFirstElementExp :: S.SQLExp -> S.SQLExp
mkFirstElementExp expIdentifier =
-- For Example
S.SEArrayIndex (S.SEFnApp "array_agg" [expIdentifier] Nothing) (S.intToSQLExp 1)
-- | Last element extractor expression from given record set.
-- For example:- To get first "id" column from given row set,
-- the function generates the SQL expression AS `(array_agg("id"))[array_length(array_agg("id"), 1)]`
mkLastElementExp :: S.SQLExp -> S.SQLExp
mkLastElementExp expIdentifier =
let arrayExp = S.SEFnApp "array_agg" [expIdentifier] Nothing
in S.SEArrayIndex arrayExp $
S.SEFnApp "array_length" [arrayExp, S.intToSQLExp 1] Nothing
cursorIdentifier :: Identifier
cursorIdentifier = Identifier "__cursor"
startCursorIdentifier :: Identifier
startCursorIdentifier = Identifier "__start_cursor"
endCursorIdentifier :: Identifier
endCursorIdentifier = Identifier "__end_cursor"
hasPreviousPageIdentifier :: Identifier
hasPreviousPageIdentifier = Identifier "__has_previous_page"
hasNextPageIdentifier :: Identifier
hasNextPageIdentifier = Identifier "__has_next_page"
pageInfoSelectAliasIdentifier :: Identifier
pageInfoSelectAliasIdentifier = Identifier "__page_info"
cursorsSelectAliasIdentifier :: Identifier
cursorsSelectAliasIdentifier = Identifier "__cursors_select"
encodeBase64 :: S.SQLExp -> S.SQLExp
encodeBase64 =
removeNewline . bytesToBase64Text . convertToBytes
where
convertToBytes e =
S.SEFnApp "convert_to" [e, S.SELit "UTF8"] Nothing
bytesToBase64Text e =
S.SEFnApp "encode" [e, S.SELit "base64"] Nothing
removeNewline e =
S.SEFnApp "regexp_replace" [e, S.SELit "\\n", S.SELit "", S.SELit "g"] Nothing
processConnectionSelect ::
forall pgKind m.
( MonadReader StringifyNumbers m,
MonadWriter JoinTree m,
Backend ('Postgres pgKind),
PostgresAnnotatedFieldJSON pgKind
) =>
SourcePrefixes ->
FieldName ->
S.Alias ->
HM.HashMap PGCol PGCol ->
ConnectionSelect ('Postgres pgKind) Void S.SQLExp ->
m
( ArrayConnectionSource,
S.Extractor,
HM.HashMap S.Alias S.SQLExp
)
processConnectionSelect sourcePrefixes fieldAlias relAlias colMapping connectionSelect = do
(selectSource, orderByAndDistinctExtrs, maybeOrderByCursor) <-
processSelectParams
sourcePrefixes
fieldAlias
similarArrayFields
selectFrom
permLimitSubQuery
tablePermissions
tableArgs
let mkCursorExtractor = (S.Alias cursorIdentifier,) . (`S.SETyAnn` S.textTypeAnn)
cursorExtractors = case maybeOrderByCursor of
Just orderByCursor -> [mkCursorExtractor orderByCursor]
Nothing ->
-- Extract primary key columns from base select along with cursor expression.
-- Those columns are required to perform connection split via a WHERE clause.
mkCursorExtractor primaryKeyColumnsObjectExp : primaryKeyColumnExtractors
(topExtractorExp, exps) <- flip runStateT [] $ processFields selectSource
let topExtractor = S.Extractor topExtractorExp $ Just $ S.Alias fieldIdentifier
allExtractors = HM.fromList $ cursorExtractors <> exps <> orderByAndDistinctExtrs
arrayConnectionSource =
ArrayConnectionSource
relAlias
colMapping
(mkSplitBoolExp <$> maybeSplit)
maybeSlice
selectSource
pure
( arrayConnectionSource,
topExtractor,
allExtractors
)
where
ConnectionSelect _ primaryKeyColumns maybeSplit maybeSlice select = connectionSelect
AnnSelectG fields selectFrom tablePermissions tableArgs _ = select
fieldIdentifier = toIdentifier fieldAlias
thisPrefix = _pfThis sourcePrefixes
permLimitSubQuery = PLSQNotRequired
primaryKeyColumnsObjectExp =
S.applyJsonBuildObj $
flip concatMap (toList primaryKeyColumns) $
\pgColumnInfo ->
[ S.SELit $ getPGColTxt $ ciColumn pgColumnInfo,
toJSONableExp LeaveNumbersAlone (ciType pgColumnInfo) False $
S.mkQIdenExp (mkBaseTableAlias thisPrefix) $ ciColumn pgColumnInfo
]
primaryKeyColumnExtractors =
flip map (toList primaryKeyColumns) $
\pgColumnInfo ->
let pgColumn = ciColumn pgColumnInfo
in ( S.Alias $ mkBaseTableColumnAlias thisPrefix pgColumn,
S.mkQIdenExp (mkBaseTableAlias thisPrefix) pgColumn
)
mkSplitBoolExp (firstSplit NE.:| rest) =
S.BEBin S.OrOp (mkSplitCompareExp firstSplit) $ mkBoolExpFromRest firstSplit rest
where
mkBoolExpFromRest previousSplit =
S.BEBin S.AndOp (mkEqualityCompareExp previousSplit) . \case
[] -> S.BELit False
(thisSplit : remainingSplit) -> mkSplitBoolExp (thisSplit NE.:| remainingSplit)
mkSplitCompareExp (ConnectionSplit kind v (OrderByItemG obTyM obCol _)) =
let obAlias = mkAnnOrderByAlias thisPrefix fieldAlias similarArrayFields obCol
obTy = fromMaybe S.OTAsc obTyM
compareOp = case (kind, obTy) of
(CSKAfter, S.OTAsc) -> S.SGT
(CSKAfter, S.OTDesc) -> S.SLT
(CSKBefore, S.OTAsc) -> S.SLT
(CSKBefore, S.OTDesc) -> S.SGT
in S.BECompare compareOp (S.SEIdentifier $ toIdentifier obAlias) v
mkEqualityCompareExp (ConnectionSplit _ v orderByItem) =
let obAlias =
mkAnnOrderByAlias thisPrefix fieldAlias similarArrayFields $
obiColumn orderByItem
in S.BECompare S.SEQ (S.SEIdentifier $ toIdentifier obAlias) v
similarArrayFields = HM.unions $
flip map (map snd fields) $ \case
ConnectionTypename {} -> mempty
ConnectionPageInfo {} -> mempty
ConnectionEdges edges -> HM.unions $
flip map (map snd edges) $ \case
EdgeTypename {} -> mempty
EdgeCursor {} -> mempty
EdgeNode annFields ->
mkSimilarArrayFields annFields $ _saOrderBy tableArgs
mkSimpleJsonAgg rowExp ob =
let jsonAggExp = S.SEFnApp "json_agg" [rowExp] ob
in S.SEFnApp "coalesce" [jsonAggExp, S.SELit "[]"] Nothing
processFields ::
forall n.
( MonadReader StringifyNumbers n,
MonadWriter JoinTree n,
MonadState [(S.Alias, S.SQLExp)] n
) =>
SelectSource ->
n S.SQLExp
processFields selectSource =
fmap (S.applyJsonBuildObj . concat) $
forM fields $
\(FieldName fieldText, field) ->
(S.SELit fieldText :) . pure
<$> case field of
ConnectionTypename t -> pure $ withForceAggregation S.textTypeAnn $ S.SELit t
ConnectionPageInfo pageInfoFields -> pure $ processPageInfoFields pageInfoFields
ConnectionEdges edges ->
fmap (flip mkSimpleJsonAgg (orderByForJsonAgg selectSource) . S.applyJsonBuildObj . concat) $
forM edges $
\(FieldName edgeText, edge) ->
(S.SELit edgeText :) . pure
<$> case edge of
EdgeTypename t -> pure $ S.SELit t
EdgeCursor -> pure $ encodeBase64 $ S.SEIdentifier (toIdentifier cursorIdentifier)
EdgeNode annFields -> do
let edgeFieldName =
FieldName $
getFieldNameTxt fieldAlias <> "." <> fieldText <> "." <> edgeText
edgeFieldIdentifier = toIdentifier edgeFieldName
annFieldsExtrExp <- processAnnFields thisPrefix edgeFieldName similarArrayFields annFields
modify' (<> [annFieldsExtrExp])
pure $ S.SEIdentifier edgeFieldIdentifier
processPageInfoFields infoFields =
S.applyJsonBuildObj $
flip concatMap infoFields $
\(FieldName fieldText, field) -> (:) (S.SELit fieldText) $ pure case field of
PageInfoTypename t -> withForceAggregation S.textTypeAnn $ S.SELit t
PageInfoHasNextPage ->
withForceAggregation S.boolTypeAnn $
mkSingleFieldSelect (S.SEIdentifier hasNextPageIdentifier) pageInfoSelectAliasIdentifier
PageInfoHasPreviousPage ->
withForceAggregation S.boolTypeAnn $
mkSingleFieldSelect (S.SEIdentifier hasPreviousPageIdentifier) pageInfoSelectAliasIdentifier
PageInfoStartCursor ->
withForceAggregation S.textTypeAnn $
encodeBase64 $ mkSingleFieldSelect (S.SEIdentifier startCursorIdentifier) cursorsSelectAliasIdentifier
PageInfoEndCursor ->
withForceAggregation S.textTypeAnn $
encodeBase64 $ mkSingleFieldSelect (S.SEIdentifier endCursorIdentifier) cursorsSelectAliasIdentifier
where
mkSingleFieldSelect field fromIdentifier =
S.SESelect
S.mkSelect
{ S.selExtr = [S.Extractor field Nothing],
S.selFrom = Just $ S.FromExp [S.FIIdentifier fromIdentifier]
}
connectionToSelectWith ::
S.Alias ->
ArrayConnectionSource ->
MultiRowSelectNode ->
S.SelectWithG S.Select
connectionToSelectWith baseSelectAlias arrayConnectionSource arraySelectNode =
let extractionSelect =
S.mkSelect
{ S.selExtr = topExtractors,
S.selFrom = Just $ S.FromExp [S.FIIdentifier finalSelectIdentifier]
}
in S.SelectWith fromBaseSelections extractionSelect
where
ArrayConnectionSource _ columnMapping maybeSplit maybeSlice selectSource =
arrayConnectionSource
MultiRowSelectNode topExtractors selectNode = arraySelectNode
baseSelectIdentifier = Identifier "__base_select"
splitSelectIdentifier = Identifier "__split_select"
sliceSelectIdentifier = Identifier "__slice_select"
finalSelectIdentifier = Identifier "__final_select"
rowNumberIdentifier = Identifier "__row_number"
rowNumberExp = S.SEUnsafe "(row_number() over (partition by 1))"
startRowNumberIdentifier = Identifier "__start_row_number"
endRowNumberIdentifier = Identifier "__end_row_number"
startCursorExp = mkFirstElementExp $ S.SEIdentifier cursorIdentifier
endCursorExp = mkLastElementExp $ S.SEIdentifier cursorIdentifier
startRowNumberExp = mkFirstElementExp $ S.SEIdentifier rowNumberIdentifier
endRowNumberExp = mkLastElementExp $ S.SEIdentifier rowNumberIdentifier
fromBaseSelections =
let joinCond = mkJoinCond baseSelectAlias columnMapping
baseSelectFrom =
S.mkSelFromItem
(generateSQLSelect joinCond selectSource selectNode)
$ S.Alias $ _ssPrefix selectSource
select =
S.mkSelect
{ S.selExtr =
[ S.selectStar,
S.Extractor rowNumberExp $ Just $ S.Alias rowNumberIdentifier
],
S.selFrom = Just $ S.FromExp [baseSelectFrom]
}
in (S.Alias baseSelectIdentifier, select) : fromSplitSelection
mkStarSelect fromIdentifier =
S.mkSelect
{ S.selExtr = [S.selectStar],
S.selFrom = Just $ S.FromExp [S.FIIdentifier fromIdentifier]
}
fromSplitSelection = case maybeSplit of
Nothing -> fromSliceSelection baseSelectIdentifier
Just splitBool ->
let select =
(mkStarSelect baseSelectIdentifier) {S.selWhere = Just $ S.WhereFrag splitBool}
in (S.Alias splitSelectIdentifier, select) : fromSliceSelection splitSelectIdentifier
fromSliceSelection prevSelect = case maybeSlice of
Nothing -> fromFinalSelect prevSelect
Just slice ->
let select = case slice of
SliceFirst limit ->
(mkStarSelect prevSelect)
{ S.selLimit = (Just . S.LimitExp . S.intToSQLExp) limit
}
SliceLast limit ->
let mkRowNumberOrderBy obType =
let orderByItem =
S.OrderByItem (S.SEIdentifier rowNumberIdentifier) (Just obType) Nothing
in S.OrderByExp $ orderByItem NE.:| []
sliceLastSelect =
(mkStarSelect prevSelect)
{ S.selLimit = (Just . S.LimitExp . S.intToSQLExp) limit,
S.selOrderBy = Just $ mkRowNumberOrderBy S.OTDesc
}
sliceLastSelectFrom =
S.mkSelFromItem sliceLastSelect $ S.Alias sliceSelectIdentifier
in S.mkSelect
{ S.selExtr = [S.selectStar],
S.selFrom = Just $ S.FromExp [sliceLastSelectFrom],
S.selOrderBy = Just $ mkRowNumberOrderBy S.OTAsc
}
in (S.Alias sliceSelectIdentifier, select) : fromFinalSelect sliceSelectIdentifier
fromFinalSelect prevSelect =
let select = mkStarSelect prevSelect
in (S.Alias finalSelectIdentifier, select) : fromCursorSelection
fromCursorSelection =
let extrs =
[ S.Extractor startCursorExp $ Just $ S.Alias startCursorIdentifier,
S.Extractor endCursorExp $ Just $ S.Alias endCursorIdentifier,
S.Extractor startRowNumberExp $ Just $ S.Alias startRowNumberIdentifier,
S.Extractor endRowNumberExp $ Just $ S.Alias endRowNumberIdentifier
]
select =
S.mkSelect
{ S.selExtr = extrs,
S.selFrom = Just $ S.FromExp [S.FIIdentifier finalSelectIdentifier]
}
in (S.Alias cursorsSelectAliasIdentifier, select) : fromPageInfoSelection
fromPageInfoSelection =
let hasPrevPage =
S.SEBool $
S.mkExists (S.FIIdentifier baseSelectIdentifier) $
S.BECompare S.SLT (S.SEIdentifier rowNumberIdentifier) $
S.SESelect $
S.mkSelect
{ S.selFrom = Just $ S.FromExp [S.FIIdentifier cursorsSelectAliasIdentifier],
S.selExtr = [S.Extractor (S.SEIdentifier startRowNumberIdentifier) Nothing]
}
hasNextPage =
S.SEBool $
S.mkExists (S.FIIdentifier baseSelectIdentifier) $
S.BECompare S.SGT (S.SEIdentifier rowNumberIdentifier) $
S.SESelect $
S.mkSelect
{ S.selFrom = Just $ S.FromExp [S.FIIdentifier cursorsSelectAliasIdentifier],
S.selExtr = [S.Extractor (S.SEIdentifier endRowNumberIdentifier) Nothing]
}
select =
S.mkSelect
{ S.selExtr =
[ S.Extractor hasPrevPage $ Just $ S.Alias hasPreviousPageIdentifier,
S.Extractor hasNextPage $ Just $ S.Alias hasNextPageIdentifier
]
}
in pure (S.Alias pageInfoSelectAliasIdentifier, select)