通过 haskell 管道传输 http 流

Piping an http stream through a haskell conduit

我正在尝试创建一个管道,它将通过管道源从 HTTP 流式传输数据。这是我目前所拥有的:

import qualified Network.HTTP.Client.Conduit as CC

getStream :: String -> IO (ConduitM () BS.ByteString IO ())
getStream url = do
  req <- parseUrl url
  return $  CC.withResponse req $ \res -> do
    responseBody res $= (awaitForever $ \bytes -> liftIO $ do
      putStrLn $ "Got " ++ show (BS.length bytes) ++ " but will ignore    them")

但是我得到了

No instance for (Control.Monad.Reader.Class.MonadReader env0 IO) …
      arising from a use of ‘CC.withResponse’
    In the expression: CC.withResponse req
    In the second argument of ‘($)’, namely
      ‘CC.withResponse req
       $ \ res
           -> do { responseBody res $= (awaitForever $ \ bytes -> ...) }’
    In a stmt of a 'do' block:
      return
      $ CC.withResponse req
        $ \ res
            -> do { responseBody res $= (awaitForever $ \ bytes -> ...) }

怎么会出现 MonadReader?这对我来说没有任何意义。

这个 the example in the Network.HTTP.Conduit docs 的变体怎么样:

{-# LANGUAGE OverloadedStrings #-}

module Lib2 () where

import Data.Conduit (($$+-), awaitForever)
import qualified Network.HTTP.Client.Conduit as CC
import Network.HTTP.Conduit (http, tlsManagerSettings, newManager)
import Control.Monad.IO.Class (liftIO)
import Control.Monad.Trans.Resource (runResourceT)
import Data.Conduit.Binary (sinkFile) -- Exported from the package conduit-extra

main2 :: IO ()
main2 = do
       request <- CC.parseUrl "http://google.com/"
       manager <- newManager tlsManagerSettings
       runResourceT $ do
           response <- http request manager
           CC.responseBody response $$+- (awaitForever $ \x -> liftIO $ putStrLn "Chunk")

原回答

getStream 的 return 类型错误。尝试删除类型签名并使用 FlexibleContexts,例如:

{-# LANGUAGE OverloadedStrings, FlexibleContexts #-}

module Lib () where

import Data.Conduit
import qualified Data.ByteString as BS
import qualified Network.HTTP.Client.Conduit as CC
import Control.Monad.IO.Class

getStream url = do
  req <- CC.parseUrl url
  CC.withResponse req $ \res -> do
   CC.responseBody res $= (awaitForever $ \x -> liftIO $ putStrLn "Got a chunk")

然后 :t getStream 报告:

getStream
  :: (monad-control-1.0.0.4:Control.Monad.Trans.Control.MonadBaseControl
        IO (ConduitM a c m),
      mtl-2.2.1:Control.Monad.Reader.Class.MonadReader env m, MonadIO m,
      CC.HasHttpManager env,
      exceptions-0.8.0.2:Control.Monad.Catch.MonadThrow m) =>
     String -> ConduitM a c m ()

这表明 return 类型具有 ConduitM ... 的形式,而不是 IO ....

这也显示了 MonadReader 如何进入画面...monad m 必须通过 reader 环境访问 HTTP 管理器,如以下约束所示:

CC.HasHttpManager env
MonadReader env m

这就是说 m 有一个 reader 某种类型的环境 env 它本身有一种访问 HTTP 管理器的方法。

特别是,m 不能只是普通的 IO monad,这就是错误消息所抱怨的。

在评论中回答问题

下面是如何从 HTTP 响应创建 Producer 的示例:

{-# LANGUAGE OverloadedStrings #-}

module Lib3 () where

import qualified Data.ByteString as BS
import qualified Network.HTTP.Client.Conduit as CC
import           Network.HTTP.Conduit (http, tlsManagerSettings, newManager)
import qualified Network.HTTP.Client          as Client (httpLbs, responseOpen, responseClose)
import           Data.Conduit (Producer, addCleanup)
import           Data.Conduit (awaitForever, await, ($$))
import qualified Network.HTTP.Client.Conduit  as HCC

import Control.Monad.IO.Class (liftIO, MonadIO)

getStream url = do
  request <- CC.parseUrl url
  manager <- newManager tlsManagerSettings
  response <- Client.responseOpen request manager
  let producer :: Producer IO BS.ByteString
      producer = HCC.bodyReaderSource $ CC.responseBody response
      cleanup _ = do liftIO $ putStrLn "(cleaning up)"; Client.responseClose response
      producerWithCleanup = addCleanup cleanup producer
  return $ response { CC.responseBody = producerWithCleanup }

test = do
  res <- getStream "http://google.com"
  let producer = CC.responseBody res
      consumer = awaitForever $ \_ -> liftIO $ putStrLn "Got a chunk"
  producer $$ consumer