graphql-engine/server/src-lib/Hasura/Backends/Postgres/SQL/Value.hs
Auke Booij 4c8ea8e865 Import pg-client-hs as PG
Result of executing the following commands:
```shell
# replace "as Q" imports with "as PG" (in retrospect this didn't need a regex)
git grep -lE 'as Q($|[^a-zA-Z])' -- '*.hs' | xargs sed -i -E 's/as Q($|[^a-zA-Z])/as PG\1/'
# replace " Q." with " PG."
git grep -lE ' Q\.' -- '*.hs' | xargs sed -i 's/ Q\./ PG./g'
# replace "(Q." with "(PG."
git grep -lE '\(Q\.' -- '*.hs' | xargs sed -i 's/(Q\./(PG./g'
# ditto, but for [, |, { and !
git grep -lE '\[Q\.' -- '*.hs' | xargs sed -i 's/\[Q\./\[PG./g'
git grep -l '|Q\.' -- '*.hs' | xargs sed -i 's/|Q\./|PG./g'
git grep -l '{Q\.' -- '*.hs' | xargs sed -i 's/{Q\./{PG./g'
git grep -l '!Q\.' -- '*.hs' | xargs sed -i 's/!Q\./!PG./g'
```
(Doing the `grep -l` before the `sed`, instead of `sed` on the entire codebase, reduces the number of `mtime` updates, and so reduces how many times a file gets recompiled while checking intermediate results.)

Finally, I manually removed a broken and unused `Arbitrary` instance in `Hasura.RQL.Network`. (It used an `import Test.QuickCheck.Arbitrary as Q` statement, which was erroneously caught by the first find-replace command.)

After this PR, `Q` is no longer used as an import qualifier. That was not the goal of this PR, but perhaps it's a useful fact for future efforts.

PR-URL: https://github.com/hasura/graphql-engine-mono/pull/5933
GitOrigin-RevId: 8c84c59d57789111d40f5d3322c5a885dcfbf40e
2022-09-20 19:55:51 +00:00

349 lines
12 KiB
Haskell

-- | Postgres SQL Value
--
-- Deals with Postgres scalar values, converting them to and from 'Text', and to
-- JSON 'Value'.
module Hasura.Backends.Postgres.SQL.Value
( PGScalarValue (..),
pgScalarValueToJson,
withConstructorFn,
parsePGValue,
scientificToInteger,
scientificToFloat,
textToScalarValue,
TxtEncodedVal (..),
txtEncodedVal,
binEncoder,
txtEncoder,
toPrepParam,
withScalarTypeAnn,
withTypeAnn,
)
where
import Data.Aeson
import Data.Aeson.Text qualified as AE
import Data.Aeson.Types qualified as AT
import Data.ByteString qualified as B
import Data.Int
import Data.Scientific
import Data.Text qualified as T
import Data.Text.Conversions qualified as TC
import Data.Text.Encoding qualified as TE
import Data.Text.Lazy qualified as TL
import Data.Time
import Data.UUID qualified as UUID
import Database.PG.Query qualified as PG
import Database.PG.Query.PTI qualified as PTI
import Database.PostgreSQL.LibPQ qualified as PQ
import Hasura.Backends.Postgres.SQL.DML qualified as S
import Hasura.Backends.Postgres.SQL.Types
import Hasura.Prelude
import Hasura.SQL.GeoJSON
import Hasura.SQL.Time
import Hasura.SQL.Types
import Hasura.SQL.Value (TxtEncodedVal (..))
import PostgreSQL.Binary.Encoding qualified as PE
newtype RasterWKB = RasterWKB {getRasterWKB :: TC.Base16 B.ByteString}
deriving (Show, Eq)
instance FromJSON RasterWKB where
parseJSON = \case
String t -> case TC.fromText t of
Just v -> return $ RasterWKB v
Nothing ->
fail
"invalid hexadecimal representation of raster well known binary format"
_ -> fail "expecting String for raster"
instance ToJSON RasterWKB where
toJSON = toJSON . TC.toText . getRasterWKB
newtype Ltree = Ltree Text
deriving (Show, Eq)
instance ToJSON Ltree where
toJSON (Ltree t) = toJSON t
instance FromJSON Ltree where
parseJSON = \case
String t ->
if any T.null $ T.splitOn (T.pack ".") t
then fail message
else pure $ Ltree t
_ -> fail message
where
message = "Expecting label path: a sequence of zero or more labels separated by dots, for example L1.L2.L3"
-- @PGScalarValue@ represents any value that can be a column in a Postgres table
data PGScalarValue
= PGValInteger Int32
| PGValSmallInt Int16
| PGValBigInt Int64
| PGValFloat Float
| PGValDouble Double
| PGValNumeric Scientific
| PGValMoney Scientific
| PGValBoolean Bool
| PGValChar Char
| PGValVarchar Text
| PGValText Text
| PGValCitext Text
| PGValDate Day
| PGValTimeStamp LocalTime
| PGValTimeStampTZ UTCTime
| PGValTimeTZ ZonedTimeOfDay
| PGNull PGScalarType
| PGValJSON PG.JSON
| PGValJSONB PG.JSONB
| PGValGeo GeometryWithCRS
| PGValRaster RasterWKB
| PGValUUID UUID.UUID
| PGValLtree Ltree
| PGValLquery Text
| PGValLtxtquery Text
| PGValUnknown Text
| PGValArray [PGScalarValue]
deriving (Show, Eq)
pgScalarValueToJson :: PGScalarValue -> Value
pgScalarValueToJson = \case
PGValInteger i -> toJSON i
PGValSmallInt i -> toJSON i
PGValBigInt i -> toJSON i
PGValFloat f -> toJSON f
PGValDouble d -> toJSON d
PGValNumeric sc -> toJSON sc
PGValMoney m -> toJSON m
PGValBoolean b -> toJSON b
PGValChar t -> toJSON t
PGValVarchar t -> toJSON t
PGValText t -> toJSON t
PGValCitext t -> toJSON t
PGValDate d -> toJSON d
PGValTimeStamp u -> String $ formatTimestamp u
PGValTimeStampTZ u -> String $ formatTimestamp u
PGValTimeTZ (ZonedTimeOfDay tod tz) ->
toJSON (show tod ++ timeZoneOffsetString tz)
PGNull _ -> Null
PGValJSON (PG.JSON j) -> j
PGValJSONB (PG.JSONB j) -> j
PGValGeo o -> toJSON o
PGValRaster r -> toJSON r
PGValUUID u -> toJSON u
PGValLtree t -> toJSON t
PGValLquery t -> toJSON t
PGValLtxtquery t -> toJSON t
PGValUnknown t -> toJSON t
PGValArray a -> toJSON (map pgScalarValueToJson a)
textToScalarValue :: Maybe Text -> PGScalarValue
textToScalarValue = maybe (PGNull PGText) PGValText
withConstructorFn :: PGScalarType -> S.SQLExp -> S.SQLExp
withConstructorFn ty v
| isGeoType ty = S.SEFnApp "ST_GeomFromGeoJSON" [v] Nothing
| ty == PGRaster = S.SEFnApp "ST_RastFromHexWKB" [v] Nothing
| otherwise = v
-- FIXME: shouldn't this also use 'withConstructorFn'?
withScalarTypeAnn :: PGScalarType -> S.SQLExp -> S.SQLExp
withScalarTypeAnn colTy v = S.SETyAnn v . S.mkTypeAnn $ CollectableTypeScalar colTy
withTypeAnn :: CollectableType PGScalarType -> S.SQLExp -> S.SQLExp
withTypeAnn ty expr = flip S.SETyAnn (S.mkTypeAnn ty) $
case ty of
CollectableTypeScalar baseTy -> withConstructorFn baseTy expr
CollectableTypeArray _ -> expr
-- TODO: those two functions are useful outside of Postgres, and
-- should be moved to a common place of the code. Perhaps the Prelude?
scientificToInteger :: (Integral i, Bounded i) => Scientific -> AT.Parser i
scientificToInteger num =
toBoundedInteger num
`onNothing` fail
( "The value " ++ show num ++ " lies outside the "
++ "bounds or is not an integer. Maybe it is a "
++ "float, or is there integer overflow?"
)
scientificToFloat :: (RealFloat f) => Scientific -> AT.Parser f
scientificToFloat num =
toBoundedRealFloat num
`onLeft` \_ ->
fail
( "The value " ++ show num ++ " lies outside the "
++ "bounds. Is it overflowing the float bounds?"
)
parsePGValue :: PGScalarType -> Value -> AT.Parser PGScalarValue
parsePGValue ty val = case (ty, val) of
(_, Null) -> pure $ PGNull ty
(PGUnknown _, String t) -> pure $ PGValUnknown t
(PGRaster, _) -> parseTyped -- strictly parse raster value
(PGLtree, _) -> parseTyped
(_, String t) -> parseTyped <|> pure (PGValUnknown t)
(_, _) -> parseTyped
where
parseBoundedInt :: forall i. (Integral i, Bounded i) => Value -> AT.Parser i
parseBoundedInt = withScientific ("Integer expected for input type: " ++ show ty) scientificToInteger
parseBoundedFloat :: forall a. (RealFloat a) => Value -> AT.Parser a
parseBoundedFloat = withScientific ("Float expected for input type: " ++ show ty) scientificToFloat
parseTyped = case ty of
PGSmallInt -> PGValSmallInt <$> parseBoundedInt val
PGInteger -> PGValInteger <$> parseBoundedInt val
PGBigInt -> PGValBigInt <$> parseBoundedInt val
PGSerial -> PGValInteger <$> parseBoundedInt val
PGBigSerial -> PGValBigInt <$> parseBoundedInt val
PGFloat -> PGValFloat <$> parseBoundedFloat val
PGDouble -> PGValDouble <$> parseBoundedFloat val
PGNumeric -> PGValNumeric <$> parseJSON val
PGMoney -> PGValMoney <$> parseJSON val
PGBoolean -> PGValBoolean <$> parseJSON val
PGChar -> PGValChar <$> parseJSON val
PGVarchar -> PGValVarchar <$> parseJSON val
PGText -> PGValText <$> parseJSON val
PGCitext -> PGValCitext <$> parseJSON val
PGDate -> PGValDate <$> parseJSON val
PGTimeStamp -> PGValTimeStamp <$> parseJSON val
PGTimeStampTZ -> PGValTimeStampTZ <$> parseJSON val
PGTimeTZ -> PGValTimeTZ <$> parseJSON val
PGJSON -> PGValJSON . PG.JSON <$> parseJSON val
PGJSONB -> PGValJSONB . PG.JSONB <$> parseJSON val
PGGeometry -> PGValGeo <$> parseJSON val
PGGeography -> PGValGeo <$> parseJSON val
PGRaster -> PGValRaster <$> parseJSON val
PGUUID -> PGValUUID <$> parseJSON val
PGLtree -> PGValLtree <$> parseJSON val
PGLquery -> PGValLquery <$> parseJSON val
PGLtxtquery -> PGValLtxtquery <$> parseJSON val
PGUnknown tyName ->
fail $ "A string is expected for type: " ++ T.unpack tyName
PGCompositeScalar tyName ->
fail $ "A string is expected for type: " ++ T.unpack tyName
PGEnumScalar tyName ->
fail $ "A string is expected for type: " ++ T.unpack tyName
PGArray s -> parseJSON val >>= fmap PGValArray . traverse (parsePGValue s)
txtEncodedVal :: PGScalarValue -> TxtEncodedVal
txtEncodedVal = \case
PGValInteger i -> TELit $ tshow i
PGValSmallInt i -> TELit $ tshow i
PGValBigInt i -> TELit $ tshow i
PGValFloat f -> TELit $ tshow f
PGValDouble d -> TELit $ tshow d
PGValNumeric sc -> TELit $ tshow sc
-- PostgreSQL doesn't like scientific notation for money, so pass it
-- with 2 decimal places.
PGValMoney m -> TELit $ T.pack $ formatScientific Fixed (Just 2) m
PGValBoolean b -> TELit $ bool "false" "true" b
PGValChar t -> TELit $ T.singleton t
PGValVarchar t -> TELit t
PGValText t -> TELit t
PGValCitext t -> TELit t
PGValDate d -> TELit $ T.pack $ showGregorian d
PGValTimeStamp u -> TELit $ formatTimestamp u
PGValTimeStampTZ u -> TELit $ formatTimestamp u
PGValTimeTZ (ZonedTimeOfDay tod tz) ->
TELit $ T.pack (show tod ++ timeZoneOffsetString tz)
PGNull _ ->
TENull
PGValJSON (PG.JSON j) ->
TELit $
TL.toStrict $
AE.encodeToLazyText j
PGValJSONB (PG.JSONB j) ->
TELit $
TL.toStrict $
AE.encodeToLazyText j
PGValGeo o ->
TELit $
TL.toStrict $
AE.encodeToLazyText o
PGValRaster r -> TELit $ TC.toText $ getRasterWKB r
PGValUUID u -> TELit $ UUID.toText u
PGValLtree (Ltree t) -> TELit t
PGValLquery t -> TELit t
PGValLtxtquery t -> TELit t
PGValUnknown t -> TELit t
PGValArray ts -> TELit $ buildArrayLiteral ts
binEncoder :: PGScalarValue -> PG.PrepArg
binEncoder = \case
PGValInteger i -> PG.toPrepVal i
PGValSmallInt i -> PG.toPrepVal i
PGValBigInt i -> PG.toPrepVal i
PGValFloat f -> PG.toPrepVal f
PGValDouble d -> PG.toPrepVal d
PGValNumeric sc -> PG.toPrepVal sc
PGValMoney m -> PG.toPrepVal m
PGValBoolean b -> PG.toPrepVal b
PGValChar t -> PG.toPrepVal t
PGValVarchar t -> PG.toPrepVal t
PGValText t -> PG.toPrepVal t
PGValCitext t -> PG.toPrepVal t
PGValDate d -> PG.toPrepVal d
PGValTimeStamp u -> PG.toPrepVal u
PGValTimeStampTZ u -> PG.toPrepVal u
PGValTimeTZ (ZonedTimeOfDay t z) -> PG.toPrepValHelper PTI.timetz PE.timetz_int (t, z)
PGNull ty -> (pgTypeOid ty, Nothing)
PGValJSON u -> PG.toPrepVal u
PGValJSONB u -> PG.toPrepVal u
PGValGeo o -> PG.toPrepVal $ TL.toStrict $ AE.encodeToLazyText o
PGValRaster r -> PG.toPrepVal $ TC.toText $ getRasterWKB r
PGValUUID u -> PG.toPrepVal u
PGValLtree (Ltree t) -> PG.toPrepVal t
PGValLquery t -> PG.toPrepVal t
PGValLtxtquery t -> PG.toPrepVal t
PGValUnknown t -> (PTI.auto, Just (TE.encodeUtf8 t, PQ.Text))
PGValArray s -> (PTI.auto, Just (TE.encodeUtf8 $ buildArrayLiteral s, PQ.Text))
formatTimestamp :: FormatTime t => t -> Text
formatTimestamp = T.pack . formatTime defaultTimeLocale "%0Y-%m-%dT%T%QZ"
txtEncoder :: PGScalarValue -> S.SQLExp
txtEncoder colVal = case txtEncodedVal colVal of
TENull -> S.SENull
TELit t -> S.SELit t
-- arrays are sufficiently complicated, e.g. in the case of empty and unknown element arrays,
-- for us to default to text encoding in all cases, and defer to Postgres' handling of them
--
-- FIXME: this will fail if we ever introduce the box type as a @PGScalarValue@,
-- which uses a different seperator https://www.postgresql.org/docs/current/arrays.html#ARRAYS-INPUT
-- https://github.com/hasura/graphql-engine-mono/issues/4892
buildArrayLiteral :: [PGScalarValue] -> Text
buildArrayLiteral ts =
T.concat ["{", T.intercalate "," (map (inner . encodeElement) ts), "}"]
where
-- present text elements as json strings
escape = TL.toStrict . AE.encodeToLazyText
encodeElement = \case
PGValChar t -> TELit $ escape $ T.singleton t
PGValVarchar t -> TELit $ escape t
PGValText t -> TELit $ escape t
PGValCitext t -> TELit $ escape t
PGValLquery t -> TELit $ escape t
PGValLtxtquery t -> TELit $ escape t
PGValUnknown t -> TELit $ escape t
other -> txtEncodedVal other
inner = \case
TENull -> "null"
TELit t -> t
{- Note [Type casting prepared params]
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Prepared values are passed to Postgres via text encoding. Explicit type cast for prepared params
is needed to distinguish the column types. For example, the parameter for citext column type is
generated as ($i)::citext where 'i' is parameter position (integer).
Also see https://github.com/hasura/graphql-engine/issues/2818
-}
toPrepParam :: Int -> PGScalarType -> S.SQLExp
toPrepParam i ty =
-- See Note [Type casting prepared params] above
withScalarTypeAnn ty . withConstructorFn ty $ S.SEPrep i