diff --git a/ChangeLog.md b/ChangeLog.md --- a/ChangeLog.md +++ b/ChangeLog.md @@ -1,5 +1,9 @@ # Revision history for avro +## HEAD + +- Replace `entropy` library with `tf-random`, which is easier to use with ghcjs + ## 0.4.1.1 - Fixed bugs in handling of namespaces when parsing and printing avro types diff --git a/avro.cabal b/avro.cabal --- a/avro.cabal +++ b/avro.cabal @@ -4,15 +4,15 @@ -- -- see: https://github.com/sol/hpack -- --- hash: 0368e22fa6552aa4d9b3ada7a95e8210af8b40090f3244fe99d68ca30f606e8d +-- hash: ca29cdadb7a7a9edda357686a62ea84b9122a6bedb7fe04890a5d1fcc8719127 name: avro -version: 0.4.1.1 +version: 0.4.1.2 synopsis: Avro serialization support for Haskell description: Avro serialization and deserialization support for Haskell category: Data -homepage: https://github.com/GaloisInc/avro#readme -bug-reports: https://github.com/GaloisInc/avro/issues +homepage: https://github.com/haskell-works/avro#readme +bug-reports: https://github.com/haskell-works/avro/issues author: Thomas M. DuBuisson maintainer: Alexey Raga license: BSD3 @@ -29,6 +29,7 @@ test/data/record.avsc test/data/reused.avsc test/data/small.avsc + test/data/unions-no-namespace.avsc test/data/unions.avsc test/data/enums-object.json test/data/namespace-inference.json @@ -42,7 +43,7 @@ source-repository head type: git - location: https://github.com/GaloisInc/avro + location: https://github.com/haskell-works/avro flag dev description: Use development GHC flags @@ -98,7 +99,6 @@ , bytestring , containers , data-binary-ieee754 - , entropy , fail , hashable , mtl @@ -107,6 +107,7 @@ , semigroups , tagged , text + , tf-random , unordered-containers , vector if flag(templatehaskell) @@ -169,7 +170,6 @@ , bytestring , containers , directory - , entropy , extra , fail , hashable @@ -182,6 +182,7 @@ , semigroups , tagged , text + , tf-random , transformers , unordered-containers , vector diff --git a/src/Data/Avro/Encode.hs b/src/Data/Avro/Encode.hs --- a/src/Data/Avro/Encode.hs +++ b/src/Data/Avro/Encode.hs @@ -52,8 +52,9 @@ import qualified Data.Vector as V import qualified Data.Vector.Unboxed as U import Data.Word +import System.Random.TF.Init (initTFGen) +import System.Random.TF.Instances (randoms) import Prelude as P -import System.Entropy (getEntropy) import Data.Avro.EncodeRaw import Data.Avro.HasAvroSchema @@ -67,7 +68,7 @@ -- | Generates a new synchronization marker for encoding Avro containers newSyncBytes :: IO BL.ByteString -newSyncBytes = BL.fromStrict <$> getEntropy 16 +newSyncBytes = BL.pack . DL.take 16 . randoms <$> initTFGen -- |Encode chunks of objects into a container, using 16 random bytes for -- the synchronization markers. diff --git a/src/Data/Avro/JSON.hs b/src/Data/Avro/JSON.hs --- a/src/Data/Avro/JSON.hs +++ b/src/Data/Avro/JSON.hs @@ -96,11 +96,14 @@ fail "Invalid encoding of union: object with too many fields." | otherwise = let + canonicalize name + | Just _ <- Schema.primitiveType name = name + | otherwise = Schema.renderFullname $ Schema.parseFullname name branch = - head (HashMap.keys obj) + head $ HashMap.keys obj names = HashMap.fromList [(Schema.typeName t, t) | t <- NE.toList schemas] - in case HashMap.lookup branch names of + in case HashMap.lookup (canonicalize branch) names of Just t -> do nested <- parseAvroJSON union env t (obj ! branch) return (Avro.Union schemas t nested) diff --git a/src/Data/Avro/Schema.hs b/src/Data/Avro/Schema.hs --- a/src/Data/Avro/Schema.hs +++ b/src/Data/Avro/Schema.hs @@ -22,10 +22,12 @@ , Field(..), Order(..) , TypeName(..) , renderFullname + , parseFullname , mkEnum, mkUnion , validateSchema -- * Lower level utilities , typeName + , primitiveType , buildTypeEnvironment , extractBindings @@ -294,6 +296,27 @@ NamedType name -> renderFullname name Union (x:|_) _ -> typeName x _ -> renderFullname $ name bt + +-- | If the given string is the name of a primitive type, return the +-- type, otherwise return 'Nothing'. +-- +-- @ +-- λ> primitiveType "string" +-- Just String +-- λ> primitiveType "foo" +-- Nothing +-- @ +primitiveType :: Text -> Maybe Type +primitiveType = \case + "null" -> Just Null + "boolean" -> Just Boolean + "int" -> Just Int + "long" -> Just Long + "float" -> Just Float + "double" -> Just Double + "bytes" -> Just Bytes + "string" -> Just String + _ -> Nothing data Field = Field { fldName :: Text , fldAliases :: [Text] diff --git a/test/Avro/JSONSpec.hs b/test/Avro/JSONSpec.hs --- a/test/Avro/JSONSpec.hs +++ b/test/Avro/JSONSpec.hs @@ -24,6 +24,7 @@ deriveAvro "test/data/reused.avsc" deriveAvro "test/data/small.avsc" deriveAvro "test/data/unions.avsc" +deriveAvro "test/data/unions-no-namespace.avsc" spec :: Spec spec = describe "Avro.JSONSpec: JSON serialization/parsing" $ do @@ -95,6 +96,20 @@ it "should parse (unions)" $ do parseJSON unionsJsonA `shouldBe` pure unionsExampleA parseJSON unionsJsonB `shouldBe` pure unionsExampleB + + let unionsNoNamespaceA = UnionsNoNamespace (Left TypeA) + unionsNoNamespaceB = UnionsNoNamespace (Right TypeB) + it "should roundtrip (unions-no-namespace)" $ do + parseJSON (Aeson.encode (toJSON unionsNoNamespaceA)) `shouldBe` + pure unionsNoNamespaceA + parseJSON (Aeson.encode (toJSON unionsNoNamespaceB)) `shouldBe` + pure unionsNoNamespaceB + let noNamespace = "test/data/unions-no-namespace-object.json" + objectA = "{ \"unionField\" : { \"TypeA\" : {} } }" + objectB = "{ \"unionField\" : { \"TypeB\" : {} } }" + it "should parse (unions-no-namespace)" $ do + parseJSON objectA `shouldBe` pure unionsNoNamespaceA + parseJSON objectB `shouldBe` pure unionsNoNamespaceB getFileName :: FilePath -> IO FilePath getFileName p = do diff --git a/test/data/unions-no-namespace.avsc b/test/data/unions-no-namespace.avsc new file mode 100644 --- /dev/null +++ b/test/data/unions-no-namespace.avsc @@ -0,0 +1,24 @@ +{ + "type" : "record", + "name" : "UnionsNoNamespace", + "doc" : "An example schema that has a union whose components are types with no namespace. This was causing problems with our JSON parsing code.", + "fields" : [ + { + "name" : "unionField", + "type" : [ + { + "type" : "record", + "name" : "TypeA", + "doc" : "TypeA is in the null namespace but has no leading dot in the schema.", + "fields" : [] + }, + { + "type" : "record", + "name" : ".TypeB", + "doc" : "Note the leading dot—semantically this is the same as having no namespace, but it caused parsing issues in the past.", + "fields" : [] + } + ] + } + ] +}