如何序列化作为数据实例的数据结构？

Question

我想序列化一个部分由我自己定义的数据结构，部分是使用来自外部库（它是数据的实例）的数据结构。我想在不必定义序列化过程应该如何工作的情况下进行序列化，或者换句话说，以一种相当自动化的方式。

我一直在尝试通过谷歌搜索/询问找到的几种替代方法。他们都没有成功。图书馆 (genericserialize) 似乎是我想要的。不幸的是，即使对于简单的情况（如（True，False）），它也不起作用。它可能已经过时或/和不完整。二进制序列化要求外部数据结构派生二进制（它没有）。 Cereal 会出现类似的问题。当然是Read/Show，但同样，外部数据结构并没有派生Read，所以也不能使用。

也许我误解了什么。我认为当我第一次遇到这个问题时会更容易。在其他语言中，例如 Erlang，这是以非常简单的方式完成的。

你知道图书馆/技术或任何使这个过程尽可能简单的东西吗？

Answer 1

您可以使用Data.Binary。

为所有涉及的数据类型派生Generic。这将需要-XDeriveGeneric pragma。如果您使用的外部库中的数据类型不这样做，您可能必须有一个orphan instances 的模块并使用-XStandaloneDeriving。

完成后，您可以使用Binary 类已经支持从Generic 派生的事实。您需要启用-XDeriveAnyclass，然后您可以为所有涉及的数据类型添加Binary 实例（您可能再次需要为外部库中的类型提供独立实例）。

编辑

OP added a gist。然后，要编译它，我们需要添加以下语言编译指示：

{-# LANGUAGE DeriveDataTypeable,
             DeriveAnyClass,
             DeriveGeneric,
             StandaloneDeriving #-}

还有以下导入：

import Language.C.Syntax.AST
import Language.C.Data.Position
import Language.C.Data.Node
import Language.C.Data.Name
import Language.C.Data.Ident
import Language.C.Syntax.Constants

import Data.Data
import GHC.Generics
import Data.Binary

来自 OP 要点：

type CTypeSpecAnn = CTypeSpecifier NodeAnn 
type CDeclAnn = CDeclaration NodeAnn
type CDeclSpecAnn = CDeclarationSpecifier NodeAnn
type CDeclrAnn = CDeclarator NodeAnn
type CStatAnn = CStatement NodeAnn
type CExtDeclAnn = CExternalDeclaration NodeAnn
type CExprAnn = CExpression NodeAnn
type CBlockItemAnn = CCompoundBlockItem NodeAnn
type CTranslUnitAnn = CTranslationUnit NodeAnn

data TransState = 
    TransState 
    {
        free_node_id :: Int,
        freeVar :: Int,
        includes :: [String],
        fun_defs :: [(String, (CTypeSpecAnn, [CDeclAnn], ([CDeclSpecAnn], CDeclrAnn, [CDeclAnn], NodeAnn)), CStatAnn)],
        no_fun_defs :: [CExtDeclAnn],
        last_changed :: String,
        previous_changes :: ([(String, ((String,CStatAnn,CStatAnn), TransState, [(String, CStatAnn)]))], [(String, ((String,CExprAnn,CExprAnn), TransState, [(String, CStatAnn)]))]),
        applied_rules :: [String],
        applicable_rules :: Set String,
        trans_with_anns :: Bool
    }
    deriving (Show, Data, Typeable)

type NodeAnn = Annotation NodeInfo NodeProperties
data Annotation nI nP = Ann nI nP
    deriving (Show, Data, Typeable, Eq)

data NodeProperties = 
    NodeProperties
    { 
        _hasSideEffects :: PropertyInfo Bool,
        _readIn ::  PropertyInfo [String],
        _writeIn ::  PropertyInfo [String],
        _localSymbols :: PropertyInfo [String],
        _rangeInfo :: PropertyInfo [String],
        _isCanonical :: PropertyInfo Bool,
        _isPerfectNest :: PropertyInfo Bool,
        _hasLoops :: PropertyInfo Bool,
        _hasFunctionCalls :: PropertyInfo Bool,
        _hasControlFlowModifiers :: PropertyInfo Bool,
        _scalarDependences :: PropertyInfo [String],
        _polcaPragmas :: PropertyInfo [[String]],
        _allPragmas :: PropertyInfo [String]
    } 
    deriving (Show, Data, Typeable, Eq)

data PropertyInfo a = 
    PropertyInfo
    { 
        _pragmaType :: PragmaType,
        _definedBy  :: DefinerType,
        _value :: Maybe a
    } 
    deriving (Show, Data, Typeable, Eq)

makeLenses ''NodeProperties
makeLenses ''PropertyInfo

由于 OP 从未定义 PragmaType 和 DefinerType，我假设它们是 ()。

type PragmaType = ()
type DefinerType = ()

然后我们有大量的deriving instance（我们必须递归地找到所有类型依赖项并将Generic 和Binary 添加到它们）。

deriving instance Generic (CDeclaration a)
deriving instance Binary a => Binary (CDeclaration a)
deriving instance Generic (CTypeSpecifier a)
deriving instance Binary a => Binary (CTypeSpecifier a)
deriving instance Generic (CDeclarationSpecifier a)
deriving instance Binary a => Binary (CDeclarationSpecifier a)
deriving instance Generic (CDeclarator a)
deriving instance Binary a => Binary (CDeclarator a)
deriving instance Generic (CStatement a)
deriving instance Binary a => Binary (CStatement a)
deriving instance Generic (CExternalDeclaration a)
deriving instance Binary a => Binary (CExternalDeclaration a)
deriving instance Generic (CExpression a)
deriving instance Binary a => Binary (CExpression a)
deriving instance Generic (CCompoundBlockItem a)
deriving instance Binary a => Binary (CCompoundBlockItem a)
deriving instance Generic (CTranslationUnit a)
deriving instance Binary a => Binary (CTranslationUnit a)
deriving instance Generic (CInitializer a)
deriving instance Binary a => Binary (CInitializer a)
deriving instance Generic (CStructureUnion a)
deriving instance Binary a => Binary (CStructureUnion a)
deriving instance Generic (CTypeQualifier a)
deriving instance Binary a => Binary (CTypeQualifier a)
deriving instance Generic (CStringLiteral a)
deriving instance Binary a => Binary (CStringLiteral a)
deriving instance Generic (CAttribute a)
deriving instance Binary a => Binary (CAttribute a)
deriving instance Generic (CPartDesignator a)
deriving instance Binary a => Binary (CPartDesignator a)
deriving instance Generic (CFunctionDef a)
deriving instance Binary a => Binary (CFunctionDef a)
deriving instance Generic (CAssemblyStatement a)
deriving instance Binary a => Binary (CAssemblyStatement a)
deriving instance Generic (CAssemblyOperand a)
deriving instance Binary a => Binary (CAssemblyOperand a)
deriving instance Generic (CConstant a)
deriving instance Binary a => Binary (CConstant a)
deriving instance Generic (CEnumeration a)
deriving instance Binary a => Binary (CEnumeration a)
deriving instance Generic (CStorageSpecifier a)
deriving instance Binary a => Binary (CStorageSpecifier a)
deriving instance Generic (CDerivedDeclarator a)
deriving instance Binary a => Binary (CDerivedDeclarator a)
deriving instance Generic (CBuiltinThing a)
deriving instance Binary a => Binary (CBuiltinThing a)
deriving instance Generic (CArraySize a)
deriving instance Binary a => Binary (CArraySize a)
deriving instance Generic (Flags a)
deriving instance Binary a => Binary (Flags a)
deriving instance Generic NodeInfo
deriving instance Binary NodeInfo
deriving instance Generic Name
deriving instance Binary Name
deriving instance Generic Ident
deriving instance Binary Ident
deriving instance Generic CString
deriving instance Binary CString
deriving instance Generic CStructTag
deriving instance Binary CStructTag
deriving instance Generic CFloat
deriving instance Binary CFloat
deriving instance Generic CChar
deriving instance Binary CChar
deriving instance Generic CUnaryOp
deriving instance Binary CUnaryOp
deriving instance Generic CBinaryOp
deriving instance Binary CBinaryOp
deriving instance Generic CInteger
deriving instance Binary CInteger
deriving instance Generic CAssignOp
deriving instance Binary CAssignOp
deriving instance Generic CIntFlag
deriving instance Binary CIntFlag
deriving instance Generic CIntRepr
deriving instance Binary CIntRepr

deriving instance Binary TransState
deriving instance Generic TransState
deriving instance Generic (Annotation a b)
deriving instance (Binary a, Binary b) => Binary (Annotation a b)
deriving instance Generic NodeProperties
deriving instance Binary NodeProperties
deriving instance Generic (PropertyInfo a)
deriving instance Binary a => Binary (PropertyInfo a)

附带说明，这些实例的后半部分可以在类型本身的数据声明中声明（因为它们是在当前文件中声明的）。一个问题是我们仍然没有Binary 的Binary 实例@。我们无法自动派生它，因为它的数据构造函数尚未导出。然而，没有什么能阻止我们手动编写Binary 实例。

instance Binary Position where
    put p | isNoPos p = putWord8 0 
          | isBuiltinPos p = putWord8 1
          | isInternalPos p = putWord8 2
          | isSourcePos p = putWord8 3 >> put (posOffset p) >> put (posFile p) >> put (posRow p) >> put (posColumn p)

    get = do
        marker <- getWord8
        case marker of
            0 -> return nopos
            1 -> return builtinPos
            2 -> return internalPos
            3 -> position <$> get <*> get <*> get <*> get