diff --git a/csv-slurp.cabal b/csv-slurp.cabal
index 4a556f3..15d108e 100644
--- a/csv-slurp.cabal
+++ b/csv-slurp.cabal
@@ -32,6 +32,7 @@ library
     , bytestring
     , conduit >=1.3.4.2 && <1.4
     , text
+    , transformers
   default-language: Haskell2010
   autogen-modules: Paths_csv_slurp
 
@@ -51,5 +52,6 @@ test-suite csv-slurp-test
     , csv-slurp
     , hspec >=2.8.5 && <2.9
     , text
+    , transformers
   default-language: Haskell2010
   autogen-modules: Paths_csv_slurp
diff --git a/package.yaml b/package.yaml
index 3f71495..57573fb 100644
--- a/package.yaml
+++ b/package.yaml
@@ -26,6 +26,7 @@ dependencies:
 - bytestring
 - conduit >= 1.3.4.2 && < 1.4
 - text
+- transformers
 
 library:
   source-dirs: src
diff --git a/src/Data/CSV/Slurp.hs b/src/Data/CSV/Slurp.hs
index da7e3fb..cd272f6 100644
--- a/src/Data/CSV/Slurp.hs
+++ b/src/Data/CSV/Slurp.hs
@@ -29,13 +29,16 @@ module Data.CSV.Slurp (
   decodeRows,
   decodeRawRows,
   decodeUTF8,
+  toBytes,
 ) where
 
 import Conduit (ConduitT, mapC, (.|))
+import Control.Monad.Trans.State (StateT, evalStateT)
 import qualified Data.ByteString as BS
 import Data.Maybe (fromMaybe)
 import qualified Data.Text as T
 import Data.Text.Encoding (decodeUtf8')
+import Data.Word (Word8)
 
 -- | decode the rows from a stream of ByteStrings
 decodeRows :: Monad m => ConduitT BS.ByteString [T.Text] m ()
@@ -43,7 +46,7 @@ decodeRows = decodeRawRows .| mapC (map $ fromMaybe "" . decodeUTF8)
 
 -- | decode the rows returning raw ByteStrings instead of text
 decodeRawRows :: Monad m => ConduitT BS.ByteString [BS.ByteString] m ()
-decodeRawRows = return ()
+decodeRawRows = toBytes .| evalStateT decodeLoop newDecodeState
 
 -- | decode a raw ByteString into Text (if possible)
 decodeUTF8 :: BS.ByteString -> Maybe T.Text
@@ -51,4 +54,24 @@ decodeUTF8 bs = case decodeUtf8' bs of
   Left _    -> Nothing
   Right txt -> Just txt
 
+-- | convert a stream to ByteStrings to a string of bytes
+toBytes :: Monad m => ConduitT BS.ByteString Word8 m ()
+toBytes = return ()
+
+data DecodeState = DecodeState
+  { isQuoted  :: Bool
+  , collected :: BS.ByteString
+  } deriving (Eq, Show)
+
+newDecodeState :: DecodeState
+newDecodeState = DecodeState
+  { isQuoted  = False
+  , collected = ""
+  }
+
+decodeLoop
+  :: Monad m
+  => StateT DecodeState (ConduitT Word8 [BS.ByteString] m) ()
+decodeLoop = return ()
+
 --jl
diff --git a/test/Data/CSV/SlurpSpec.hs b/test/Data/CSV/SlurpSpec.hs
index 9b59ba5..d36d7da 100644
--- a/test/Data/CSV/SlurpSpec.hs
+++ b/test/Data/CSV/SlurpSpec.hs
@@ -31,6 +31,7 @@ import Data.CSV.Slurp
 spec :: Spec
 spec = describe "Data.CSV.Slurp" $ do
   decodeRowsSpec
+  decodeRawRowsSpec
   decodeUTF8Spec
 
 decodeRowsSpec :: Spec
@@ -60,6 +61,130 @@ decodeRowsSpec = describe "decodeRows" $ mapM_
     invalidIn = ["\"a"]
     validRes  = [["foo", "bar"], ["baz", "quux"]]
 
+decodeRawRowsSpec :: Spec
+decodeRawRowsSpec = describe "decodeRawRows" $ mapM_
+  ( \(label, input, expected) -> context label $ do
+    result <- runConduit $ sourceList input .| decodeRawRows .| consume
+    let
+      expLen = length expected
+      resLen = length result
+    xit ("should have " ++ show expLen ++ " rows") $
+      resLen `shouldBe` expLen
+    mapM_
+      ( \(n, expected', result') -> context ("row " ++ show n) $
+        xit ("should be " ++ show result') $
+          expected' `shouldBe` result'
+      ) $ zip3 [(0::Int)..] expected result
+  )
+
+  --  label,         input,      expected
+  [ ( "unquoted",    unquotedIn,  normalRes    )
+  , ( "quoted",      quotedIn,    normalRes    )
+  , ( "mixed",       mixedIn,     normalRes    )
+  , ( "CR only",     crOnlyIn,    normalRes    )
+  , ( "LF only",     lfOnlyIn,    normalRes    )
+  , ( "has quote",   quoteIn,     quoteRes     )
+  , ( "has CR",      crIn,        crRes        )
+  , ( "has LF",      lfIn,        lfRes        )
+  , ( "has CRLF",    crlfIn,      crlfRes      )
+  , ( "odd chunk",   oddChunkIn,  normalRes    )
+  , ( "no newline",  noNewlineIn, normalRes    )
+  , ( "malformed",   malformedIn, malformedRes )
+  ]
+
+  where
+
+    unquotedIn =
+      [ "foo,bar\r\n"
+      , "baz,quux\r\n"
+      ]
+
+    quotedIn =
+      [ "\"foo\",\"bar\"\r\n"
+      ,  "\"baz\",\"quux\"\r\n"
+      ]
+
+    mixedIn =
+      [ "\"foo\",bar\r\n"
+      , "baz,\"quux\"\r\n"
+      ]
+
+    crOnlyIn =
+      [ "foo,bar\r"
+      , "baz,quux\r"
+      ]
+
+    lfOnlyIn =
+      [ "foo,bar\n"
+      , "baz,quux\n"
+      ]
+
+    quoteIn =
+      [ "\"a\"\"b\",bar\r\n"
+      , "baz,quux\r\n"
+      ]
+
+    crIn =
+      [ "\"a\rb\",bar\r\n"
+      , "baz,quux\r\n"
+      ]
+
+    lfIn =
+      [ "\"a\nb\",bar\r\n"
+      , "baz,quux\r\n"
+      ]
+
+    crlfIn =
+      [ "\"a\r\nb\",bar\r\n"
+      , "baz,quux\r\n"
+      ]
+
+    oddChunkIn =
+      [ "foo,"
+      , "bar\r\nbaz,"
+      , "quux\r\n"
+      ]
+
+    noNewlineIn =
+      [ "foo,bar\r\n"
+      , "baz,quux"
+      ]
+
+    malformedIn =
+      [ "a\"b,bar\r\n"
+      , "baz,quux\r\n"
+      ]
+
+    normalRes =
+      [ ["foo", "bar"]
+      , ["baz", "quux"]
+      ]
+
+    quoteRes =
+      [ ["a\"b", "bar"]
+      , ["baz", "quux"]
+      ]
+
+    crRes =
+      [ ["a\rb", "bar"]
+      , ["baz", "quux"]
+      ]
+
+    lfRes =
+      [ ["a\nb", "bar"]
+      , ["baz", "quux"]
+      ]
+
+    crlfRes =
+      [ ["a\r\nb", "bar"]
+      , ["baz", "quux"]
+      ]
+
+    malformedRes =
+      [ ["", "bar"]
+      , ["baz", "quux"]
+      ]
+
 decodeUTF8Spec :: Spec
 decodeUTF8Spec = describe "decodeUTF8" $ mapM_
   ( \(label, input, expected) -> context label $