fix: recognize binary signatures and fail early

Fail early when receiving binary input with recognized signature:
- zip[-based]: including OpenDocument and Microsoft formats
- PDF
- CFBF-based: old Microsoft formats including .doc and .xls
- DjVu

3 files changed, 28 insertions, 4 deletions

diff --git a/src/Text/Pandoc/App/Input.hs b/src/Text/Pandoc/App/Input.hs
index ecb423634..87e71da6b 100644
--- a/src/Text/Pandoc/App/Input.hs
+++ b/src/Text/Pandoc/App/Input.hs
@@ -14,7 +14,7 @@ module Text.Pandoc.App.Input
   , readInput
   ) where
 
-import Control.Monad ((>=>))
+import Control.Monad ((>=>), when)
 import Control.Monad.Except (throwError, catchError)
 import Data.Text (Text)
 import Network.URI (URI (..), parseURI)
@@ -109,15 +109,29 @@ inputToText convTabs (fp, (bs,mt)) =
                            (toTextM fp bs)
                            (\case
                               PandocUTF8DecodingError{} -> do
-                                -- TODO check for binary file signatures
-                                -- here and exit with an error instead
-                                -- of treating as latin1e..
+                                when (hasKnownSignature bs) $
+                                  throwError $
+                                    PandocInputNotTextError (T.pack fp)
                                 report $ NotUTF8Encoded
                                   (if null fp
                                       then "input"
                                       else fp)
                                 return $ T.pack $ B8.unpack bs
                               e -> throwError e)
+ where
+  -- "50 4B 03 04" is zip file signature
+  isZip bs' = "\x50\x4B\x03\x04" `BS.isPrefixOf` bs'
+  -- "25 50 44 46 2D" is PDF file signature
+  isPDF bs' = "\x25\x50\x44\x46\x2D" `BS.isPrefixOf` bs'
+  -- "D0 CF 11 E0 A1 B1 1A E1" is Compound File Binary Format signature used in
+  -- variety of old Microsoft formats (.doc and .xls among others)
+  isCFBF bs' = "\xD0\xCF\x11\xE0\xA1\xB1\x1A\xE1" `BS.isPrefixOf` bs'
+  -- "41 54 26 54 46 4F 52 4D ?? ?? ?? ?? 44 4A 56" is DjVu signature
+  isDjVu bs' = case BS.stripPrefix "\x41\x54\x26\x54\x46\x4F\x52\x4D" bs' of
+    Nothing -> False
+    Just x -> BS.isPrefixOf "\x44\x4A\x56" $ BS.drop 4 x
+
+  hasKnownSignature bs' = any ($ bs') [isZip, isPDF, isCFBF, isDjVu]
 
 inputToLazyByteString :: (FilePath, (BS.ByteString, Maybe MimeType))
                       -> BL.ByteString
diff --git a/src/Text/Pandoc/Error.hs b/src/Text/Pandoc/Error.hs
index f9c9aea8d..e782df735 100644
--- a/src/Text/Pandoc/Error.hs
+++ b/src/Text/Pandoc/Error.hs
@@ -65,6 +65,7 @@ data PandocError = PandocIOError Text IOError
                  | PandocUnsupportedExtensionError Text Text
                  | PandocCiteprocError CiteprocError
                  | PandocBibliographyError Text Text
+                 | PandocInputNotTextError Text
                  deriving (Show, Typeable, Generic)
 
 instance Exception PandocError
@@ -143,6 +144,13 @@ renderError e =
       prettyCiteprocError e'
     PandocBibliographyError fp msg ->
       "Error reading bibliography file " <> fp <> ":\n" <> msg
+    PandocInputNotTextError fp ->
+      "Expected text as an input, but received binary data from " <>
+      (if T.null fp
+        then "stdin"
+        else "file " <> fp) <>
+      ".\nIf you intended to convert from binary format, verify that it's " <>
+      "supported and use\nexplicit -f FORMAT."
 
 
 -- | Handle PandocError by exiting with an error message.
@@ -184,6 +192,7 @@ handleError (Left e) =
       PandocUTF8DecodingError{} -> 92
       PandocIpynbDecodingError{} -> 93
       PandocUnsupportedCharsetError{} -> 94
+      PandocInputNotTextError{} -> 95
       PandocCouldNotFindDataFileError{} -> 97
       PandocCouldNotFindMetadataFileError{} -> 98
       PandocResourceNotFound{} -> 99
diff --git a/src/Text/Pandoc/Format.hs b/src/Text/Pandoc/Format.hs
index 433592531..5b87a2531 100644
--- a/src/Text/Pandoc/Format.hs
+++ b/src/Text/Pandoc/Format.hs
@@ -181,6 +181,7 @@ formatFromFilePath x =
     ".ctx"      -> defFlavor "context"
     ".db"       -> defFlavor "docbook"
     ".dj"       -> defFlavor "djot"
+    ".djvu"     -> defFlavor "djvu" -- so we get an "unknown reader" error
     ".doc"      -> defFlavor "doc"  -- so we get an "unknown reader" error
     ".docx"     -> defFlavor "docx"
     ".dokuwiki" -> defFlavor "dokuwiki"