Copyright | (c) 2011-2016 Bryan O'Sullivan (c) 2011 MailRank, Inc. |
---|---|
License | BSD3 |
Maintainer | Bryan O'Sullivan <bos@serpentine.com> |
Stability | experimental |
Portability | portable |
Safe Haskell | None |
Language | Haskell2010 |
Types and functions for working efficiently with JSON data.
(A note on naming: in Greek mythology, Aeson was the father of Jason.)
- decode :: FromJSON a => ByteString -> Maybe a
- decode' :: FromJSON a => ByteString -> Maybe a
- eitherDecode :: FromJSON a => ByteString -> Either String a
- eitherDecode' :: FromJSON a => ByteString -> Either String a
- encode :: ToJSON a => a -> ByteString
- decodeStrict :: FromJSON a => ByteString -> Maybe a
- decodeStrict' :: FromJSON a => ByteString -> Maybe a
- eitherDecodeStrict :: FromJSON a => ByteString -> Either String a
- eitherDecodeStrict' :: FromJSON a => ByteString -> Either String a
- data Value
- data Encoding
- fromEncoding :: Encoding -> Builder
- type Array = Vector Value
- type Object = HashMap Text Value
- newtype DotNetTime = DotNetTime {}
- class FromJSON a where
- data Result a
- fromJSON :: FromJSON a => Value -> Result a
- class ToJSON a where
- toJSON :: a -> Value
- toEncoding :: a -> Encoding
- class KeyValue kv where
- class GFromJSON f where
- gParseJSON :: Options -> Value -> Parser (f a)
- class GToJSON f where
- class GToEncoding f where
- gToEncoding :: Options -> f a -> Encoding
- genericToJSON :: (Generic a, GToJSON (Rep a)) => Options -> a -> Value
- genericToEncoding :: (Generic a, GToEncoding (Rep a)) => Options -> a -> Encoding
- genericParseJSON :: (Generic a, GFromJSON (Rep a)) => Options -> Value -> Parser a
- defaultOptions :: Options
- withObject :: String -> (Object -> Parser a) -> Value -> Parser a
- withText :: String -> (Text -> Parser a) -> Value -> Parser a
- withArray :: String -> (Array -> Parser a) -> Value -> Parser a
- withNumber :: String -> (Number -> Parser a) -> Value -> Parser a
- withScientific :: String -> (Scientific -> Parser a) -> Value -> Parser a
- withBool :: String -> (Bool -> Parser a) -> Value -> Parser a
- data Series
- pairs :: Series -> Encoding
- foldable :: (Foldable t, ToJSON a) => t a -> Encoding
- (.:) :: FromJSON a => Object -> Text -> Parser a
- (.:?) :: FromJSON a => Object -> Text -> Parser (Maybe a)
- (.:!) :: FromJSON a => Object -> Text -> Parser (Maybe a)
- (.!=) :: Parser (Maybe a) -> a -> Parser a
- object :: [Pair] -> Value
- json :: Parser Value
- json' :: Parser Value
How to use this library
This section contains basic information on the different ways to work with data using this library. These range from simple but inflexible, to complex but flexible.
The most common way to use the library is to define a data type,
corresponding to some JSON data you want to work with, and then
write either a FromJSON
instance, a to ToJSON
instance, or both
for that type.
For example, given this JSON data:
{ "name": "Joe", "age": 12 }
we create a matching data type:
{-# LANGUAGE DeriveGeneric #-} import GHC.Generics data Person = Person { name :: Text , age :: Int } deriving (Generic, Show)
The LANGUAGE
pragma and Generic
instance let us write empty
FromJSON
and ToJSON
instances for which the compiler will
generate sensible default implementations.
instanceToJSON
Person where -- No need to provide atoJSON
implementation. -- For efficiency, we write a simpletoEncoding
implementation, as -- the default version usestoJSON
.toEncoding
=genericToEncoding
defaultOptions
instanceFromJSON
Person -- No need to provide aparseJSON
implementation.
We can now encode a value like so:
>>> encode (Person {name = "Joe", age = 12}) "{\"name\":\"Joe\",\"age\":12}"
Writing instances by hand
When necessary, we can write ToJSON
and FromJSON
instances by
hand. This is valuable when the JSON-on-the-wire and Haskell data
are different or otherwise need some more carefully managed
translation. Let's revisit our JSON data:
{ "name": "Joe", "age": 12 }
We once again create a matching data type, without bothering to add
a Generic
instance this time:
data Person = Person { name :: Text , age :: Int } deriving Show
To decode data, we need to define a FromJSON
instance:
{-# LANGUAGE OverloadedStrings #-} instance FromJSON Person where parseJSON (Object v) = Person <$> v .: "name" <*> v .: "age" -- A non-Object value is of the wrong type, so fail. parseJSON _ = empty
We can now parse the JSON data like so:
>>> decode "{\"name\":\"Joe\",\"age\":12}" :: Maybe Person Just (Person {name = "Joe", age = 12})
To encode data, we need to define a ToJSON
instance. Let's begin
with an instance written entirely by hand.
instance ToJSON Person where -- this generates aValue
toJSON
(Person name age) =object
["name".=
name, "age".=
age] -- this encodes directly to a bytestring BuildertoEncoding
(Person name age) =pairs
("name".=
name
<>
"age".=
age)
We can now encode a value like so:
>>> encode (Person {name = "Joe", age = 12}) "{\"name\":\"Joe\",\"age\":12}"
There are predefined FromJSON
and ToJSON
instances for many
types. Here's an example using lists and Int
s:
>>> decode "[1,2,3]" :: Maybe [Int] Just [1,2,3]
And here's an example using the Map
type to get a map of
Int
s.
>>> decode "{\"foo\":1,\"bar\":2}" :: Maybe (Map String Int) Just (fromList [("bar",2),("foo",1)])
Working with the AST
Sometimes you want to work with JSON data directly, without first
converting it to a custom data type. This can be useful if you want
to e.g. convert JSON data to YAML data, without knowing what the
contents of the original JSON data was. The Value
type, which is
an instance of FromJSON
, is used to represent an arbitrary JSON
AST (abstract syntax tree). Example usage:
>>> decode "{\"foo\": 123}" :: Maybe Value Just (Object (fromList [("foo",Number 123)]))
>>> decode "{\"foo\": [\"abc\",\"def\"]}" :: Maybe Value Just (Object (fromList [("foo",Array (fromList [String "abc",String "def"]))]))
Once you have a Value
you can write functions to traverse it and
make arbitrary transformations.
Decoding to a Haskell value
We can decode to any instance of FromJSON
:
λ> decode "[1,2,3]" :: Maybe [Int] Just [1,2,3]
Alternatively, there are instances for standard data types, so you
can use them directly. For example, use the Map
type to
get a map of Int
s.
λ> import Data.Map λ> decode "{\"foo\":1,\"bar\":2}" :: Maybe (Map String Int) Just (fromList [("bar",2),("foo",1)])
Decoding a mixed-type object
The above approach with maps of course will not work for mixed-type objects that don't follow a strict schema, but there are a couple of approaches available for these.
The Object
type contains JSON objects:
λ> decode "{\"name\":\"Dave\",\"age\":2}" :: Maybe Object Just (fromList) [("name",String "Dave"),("age",Number 2)]
You can extract values from it with a parser using parse
,
parseEither
or, in this example, parseMaybe
:
λ> do result <- decode "{\"name\":\"Dave\",\"age\":2}" flip parseMaybe result $ \obj -> do age <- obj .: "age" name <- obj .: "name" return (name ++ ": " ++ show (age*2)) Just "Dave: 4"
Considering that any type that implements FromJSON
can be used
here, this is quite a powerful way to parse JSON. See the
documentation in FromJSON
for how to implement this class for
your own data types.
The downside is that you have to write the parser yourself; the upside is that you have complete control over the way the JSON is parsed.
Encoding and decoding
Decoding is a two-step process.
- When decoding a value, the process is reversed: the bytes are
converted to a
Value
, then theFromJSON
class is used to convert to the desired type.
There are two ways to encode a value.
- Convert to a
Value
usingtoJSON
, then possibly further encode. This was the only method available in aeson 0.9 and earlier. - Directly encode (to what will become a
ByteString
) usingtoEncoding
. This is much more efficient (about 3x faster, and less memory intensive besides), but is only available in aeson 0.10 and newer.
For convenience, the encode
and decode
functions combine both
steps.
Direct encoding
In older versions of this library, encoding a Haskell value
involved converting to an intermediate Value
, then encoding that.
A "direct" encoder converts straight from a source Haskell value
to a ByteString
without constructing an intermediate Value
.
This approach is faster than toJSON
, and allocates less memory.
The toEncoding
method makes it possible to implement direct
encoding with low memory overhead.
To complicate matters, the default implementation of toEncoding
uses toJSON
. Why? The toEncoding
method was added to this
library much more recently than toJSON
. Using toJSON
ensures
that packages written against older versions of this library will
compile and produce correct output, but they will not see any
speedup from direct encoding.
To write a minimal implementation of direct encoding, your type
must implement GHC's Generic
class, and your code should look
like this:
toEncoding
=genericToEncoding
defaultOptions
What if you have more elaborate encoding needs? For example, perhaps you need to change the names of object keys, omit parts of a value.
To encode to a JSON "object", use the pairs
function.
toEncoding
(Person name age) =pairs
("name".=
name
<>
"age".=
age)
Any container type that implements Foldable
can be encoded to a
JSON "array" using foldable
.
> import Data.Sequence as Seq > encode (Seq.fromList [1,2,3]) "[1,2,3]"
decode :: FromJSON a => ByteString -> Maybe a Source
Efficiently deserialize a JSON value from a lazy ByteString
.
If this fails due to incomplete or invalid input, Nothing
is
returned.
The input must consist solely of a JSON document, with no trailing data except for whitespace.
This function parses immediately, but defers conversion. See
json
for details.
decode' :: FromJSON a => ByteString -> Maybe a Source
Efficiently deserialize a JSON value from a lazy ByteString
.
If this fails due to incomplete or invalid input, Nothing
is
returned.
The input must consist solely of a JSON document, with no trailing data except for whitespace.
This function parses and performs conversion immediately. See
json'
for details.
eitherDecode :: FromJSON a => ByteString -> Either String a Source
Like decode
but returns an error message when decoding fails.
eitherDecode' :: FromJSON a => ByteString -> Either String a Source
Like decode'
but returns an error message when decoding fails.
encode :: ToJSON a => a -> ByteString Source
Efficiently serialize a JSON value as a lazy ByteString
.
This is implemented in terms of the ToJSON
class's toEncoding
method.
Variants for strict bytestrings
decodeStrict :: FromJSON a => ByteString -> Maybe a Source
Efficiently deserialize a JSON value from a strict ByteString
.
If this fails due to incomplete or invalid input, Nothing
is
returned.
The input must consist solely of a JSON document, with no trailing data except for whitespace.
This function parses immediately, but defers conversion. See
json
for details.
decodeStrict' :: FromJSON a => ByteString -> Maybe a Source
Efficiently deserialize a JSON value from a lazy ByteString
.
If this fails due to incomplete or invalid input, Nothing
is
returned.
The input must consist solely of a JSON document, with no trailing data except for whitespace.
This function parses and performs conversion immediately. See
json'
for details.
eitherDecodeStrict :: FromJSON a => ByteString -> Either String a Source
Like decodeStrict
but returns an error message when decoding fails.
eitherDecodeStrict' :: FromJSON a => ByteString -> Either String a Source
Like decodeStrict'
but returns an error message when decoding fails.
Core JSON types
A JSON value represented as a Haskell value.
An encoding of a JSON value.
fromEncoding :: Encoding -> Builder Source
Acquire the underlying bytestring builder.
Convenience types
newtype DotNetTime Source
A newtype wrapper for UTCTime
that uses the same non-standard
serialization format as Microsoft .NET, whose
System.DateTime
type is by default serialized to JSON as in the following example:
/Date(1302547608878)/
The number represents milliseconds since the Unix epoch.
DotNetTime | |
|
Type conversion
A type that can be converted from JSON, with the possibility of failure.
In many cases, you can get the compiler to generate parsing code for you (see below). To begin, let's cover writing an instance by hand.
There are various reasons a conversion could fail. For example, an
Object
could be missing a required key, an Array
could be of
the wrong size, or a value could be of an incompatible type.
The basic ways to signal a failed conversion are as follows:
empty
andmzero
work, but are terse and uninformativefail
yields a custom error messagetypeMismatch
produces an informative message for cases when the value encountered is not of the expected type
An example type and instance:
-- Allow ourselves to writeText
literals. {-# LANGUAGE OverloadedStrings #-} data Coord = Coord { x :: Double, y :: Double } instance FromJSON Coord where parseJSON (Object
v) = Coord<$>
v.:
"x"<*>
v.:
"y" -- We do not expect a non-Object
value here. -- We could usemzero
to fail, buttypeMismatch
-- gives a much more informative error message. parseJSON invalid =typeMismatch
"Coord" invalid
Instead of manually writing your FromJSON
instance, there are two options
to do it automatically:
- Data.Aeson.TH provides Template Haskell functions which will derive an instance at compile time. The generated instance is optimized for your type so will probably be more efficient than the following two options:
- The compiler can provide a default generic implementation for
parseJSON
.
To use the second, simply add a deriving
clause to your
datatype and declare a Generic
FromJSON
instance for your datatype without giving
a definition for parseJSON
.
For example, the previous example can be simplified to just:
{-# LANGUAGE DeriveGeneric #-}
import GHC.Generics
data Coord = Coord { x :: Double, y :: Double } deriving Generic
instance FromJSON Coord
If DefaultSignatures
doesn't give exactly the results you want,
you can customize the generic decoding with only a tiny amount of
effort, using genericParseJSON
with your preferred Options
:
instance FromJSON Coord where parseJSON =genericParseJSON
defaultOptions
Nothing
The result of running a Parser
.
Monad Result Source | |
Functor Result Source | |
Applicative Result Source | |
Foldable Result Source | |
Traversable Result Source | |
Alternative Result Source | |
MonadPlus Result Source | |
MonadFail Result Source | |
Eq a => Eq (Result a) Source | |
Show a => Show (Result a) Source | |
Monoid (Result a) Source | |
NFData a => NFData (Result a) Source | |
Semigroup (Result a) Source |
fromJSON :: FromJSON a => Value -> Result a Source
Convert a value from JSON, failing if the types do not match.
A type that can be converted to JSON.
An example type and instance:
-- Allow ourselves to writeText
literals. {-# LANGUAGE OverloadedStrings #-} data Coord = Coord { x :: Double, y :: Double } instance ToJSON Coord where toJSON (Coord x y) =object
["x".=
x, "y".=
y] toEncoding (Coord x y) =pairs
("x".=
x<>
"y".=
y)
Instead of manually writing your ToJSON
instance, there are two options
to do it automatically:
- Data.Aeson.TH provides Template Haskell functions which will derive an instance at compile time. The generated instance is optimized for your type so will probably be more efficient than the following two options:
- The compiler can provide a default generic implementation for
toJSON
.
To use the second, simply add a deriving
clause to your
datatype and declare a Generic
ToJSON
instance for your datatype without giving
definitions for toJSON
or toEncoding
.
For example, the previous example can be simplified to a more minimal instance:
{-# LANGUAGE DeriveGeneric #-} import GHC.Generics data Coord = Coord { x :: Double, y :: Double } derivingGeneric
instance ToJSON Coord where toEncoding =genericToEncoding
defaultOptions
Why do we provide an implementation for toEncoding
here? The
toEncoding
function is a relatively new addition to this class.
To allow users of older versions of this library to upgrade without
having to edit all of their instances or encounter surprising
incompatibilities, the default implementation of toEncoding
uses
toJSON
. This produces correct results, but since it performs an
intermediate conversion to a Value
, it will be less efficient
than directly emitting an Encoding
. Our one-liner definition of
toEncoding
above bypasses the intermediate Value
.
If DefaultSignatures
doesn't give exactly the results you want,
you can customize the generic encoding with only a tiny amount of
effort, using genericToJSON
and genericToEncoding
with your
preferred Options
:
instance ToJSON Coord where toJSON =genericToJSON
defaultOptions
toEncoding =genericToEncoding
defaultOptions
Nothing
Convert a Haskell value to a JSON-friendly intermediate type.
toEncoding :: a -> Encoding Source
Encode a Haskell value as JSON.
The default implementation of this method creates an
intermediate Value
using toJSON
. This provides
source-level compatibility for people upgrading from older
versions of this library, but obviously offers no performance
advantage.
To benefit from direct encoding, you must provide an
implementation for this method. The easiest way to do so is by
having your types implement Generic
using the DeriveGeneric
extension, and then have GHC generate a method body as follows.
instance ToJSON Coord where toEncoding =genericToEncoding
defaultOptions
Generic JSON classes and options
class GFromJSON f where Source
Class of generic representation types (Rep
) that can be converted from JSON.
gParseJSON :: Options -> Value -> Parser (f a) Source
This method (applied to defaultOptions
) is used as the
default generic implementation of parseJSON
.
Class of generic representation types (Rep
) that can be converted to
JSON.
class GToEncoding f where Source
gToEncoding :: Options -> f a -> Encoding Source
This method (applied to defaultOptions
) can be used as the
default generic implementation of toEncoding
.
genericToJSON :: (Generic a, GToJSON (Rep a)) => Options -> a -> Value Source
A configurable generic JSON creator. This function applied to
defaultOptions
is used as the default for toJSON
when the type
is an instance of Generic
.
genericToEncoding :: (Generic a, GToEncoding (Rep a)) => Options -> a -> Encoding Source
A configurable generic JSON encoder. This function applied to
defaultOptions
is used as the default for toEncoding
when the type
is an instance of Generic
.
genericParseJSON :: (Generic a, GFromJSON (Rep a)) => Options -> Value -> Parser a Source
A configurable generic JSON decoder. This function applied to
defaultOptions
is used as the default for parseJSON
when the
type is an instance of Generic
.
defaultOptions :: Options Source
Default encoding Options
:
Options
{fieldLabelModifier
= id ,constructorTagModifier
= id ,allNullaryToStringTag
= True ,omitNothingFields
= False ,sumEncoding
=defaultTaggedObject
,unwrapUnaryRecords
= False }
Inspecting Value
s
Value
swithObject :: String -> (Object -> Parser a) -> Value -> Parser a Source
withObject expected f value
applies f
to the Object
when value
is an Object
and fails using
otherwise.typeMismatch
expected
withText :: String -> (Text -> Parser a) -> Value -> Parser a Source
withText expected f value
applies f
to the Text
when value
is a String
and fails using
otherwise.typeMismatch
expected
withArray :: String -> (Array -> Parser a) -> Value -> Parser a Source
withArray expected f value
applies f
to the Array
when value
is an Array
and fails using
otherwise.typeMismatch
expected
withNumber :: String -> (Number -> Parser a) -> Value -> Parser a Source
Deprecated: Use withScientific instead
withNumber expected f value
applies f
to the Number
when value
is a Number
.
and fails using
otherwise.typeMismatch
expected
withScientific :: String -> (Scientific -> Parser a) -> Value -> Parser a Source
withScientific expected f value
applies f
to the Scientific
number when value
is a Number
.
and fails using
otherwise.typeMismatch
expected
withBool :: String -> (Bool -> Parser a) -> Value -> Parser a Source
withBool expected f value
applies f
to the Bool
when value
is a Bool
and fails using
otherwise.typeMismatch
expected
Constructors and accessors
A series of values that, when encoded, should be separated by
commas. Since 0.11.0.0, the .=
operator is overloaded to create
either (Text, Value)
or Series
. You can use Series when
encoding directly to a bytestring builder as in the following
example:
toEncoding (Person name age) = pairs ("name" .= name <> "age" .= age)
(.:) :: FromJSON a => Object -> Text -> Parser a Source
Retrieve the value associated with the given key of an Object
.
The result is empty
if the key is not present or the value cannot
be converted to the desired type.
This accessor is appropriate if the key and value must be present
in an object for it to be valid. If the key and value are
optional, use .:?
instead.
(.:?) :: FromJSON a => Object -> Text -> Parser (Maybe a) Source
Retrieve the value associated with the given key of an Object
.
The result is Nothing
if the key is not present, or empty
if
the value cannot be converted to the desired type.
This accessor is most useful if the key and value can be absent
from an object without affecting its validity. If the key and
value are mandatory, use .:
instead.
(.!=) :: Parser (Maybe a) -> a -> Parser a Source
Helper for use in combination with .:?
to provide default
values for optional JSON object fields.
This combinator is most useful if the key and value can be absent
from an object without affecting its validity and we know a default
value to assign in that case. If the key and value are mandatory,
use .:
instead.
Example usage:
v1 <- o.:?
"opt_field_with_dfl" .!= "default_val" v2 <- o.:
"mandatory_field" v3 <- o.:?
"opt_field2"
Parsing
Parse a top-level JSON value.
The conversion of a parsed value to a Haskell value is deferred until the Haskell value is needed. This may improve performance if only a subset of the results of conversions are needed, but at a cost in thunk allocation.
This function is an alias for value
. In aeson 0.8 and earlier, it
parsed only object or array types, in conformance with the
now-obsolete RFC 4627.
Parse a top-level JSON value.
This is a strict version of json
which avoids building up thunks
during parsing; it performs all conversions immediately. Prefer
this version if most of the JSON data needs to be accessed.
This function is an alias for value'
. In aeson 0.8 and earlier, it
parsed only object or array types, in conformance with the
now-obsolete RFC 4627.