1 files changed, 55 insertions, 1 deletions
diff --git a/dedup/hashing.py b/dedup/hashing.py
index 002eda8..a8a46c7 100644
--- a/dedup/hashing.py
+++ b/dedup/hashing.py
@@ -1,3 +1,5 @@
+import itertools
+
 class HashBlacklist(object):
     """Turn a hashlib-like object into a hash that returns None for some
     blacklisted hashes instead of the real hash value.
@@ -29,6 +31,54 @@ class HashBlacklist(object):
     def copy(self):
         return HashBlacklist(self.hashobj.copy(), self.blacklist)
 
+class HashBlacklistContent(object):
+    """Turn a hashlib-like object into a hash that returns None for some
+    blacklisted content instead of the real hash value. Unlike HashBlacklist,
+    not the output of the hash is considered, but its input."""
+
+    def __init__(self, hashobj, blacklist=(), maxlen=None):
+        """
+        @param hashobj: a hashlib-like object
+        @param blacklist: an object providing __contains__.
+            hash inputs which are contained in the blacklist
+            are turned into None values
+        @param maxlen: the maximum length of a blacklisted input.
+            Defaults to max(map(len, blacklist)), so if it is absent,
+            the blacklist must support iteration.
+        """
+        self.hashobj = hashobj
+        self.blacklist = blacklist
+        if maxlen is None:
+            # the chain avoids passing the empty sequence to max
+            maxlen = max(itertools.chain((0,), itertools.imap(len, blacklist)))
+        self.maxlen = maxlen
+        self.stored = ""
+
+    @property
+    def name(self):
+        return self.hashobj.name
+
+    def update(self, data):
+        if self.stored is not None:
+            self.stored += data
+            if len(self.stored) > self.maxlen:
+                self.stored = None
+        self.hashobj.update(data)
+
+    def digest(self):
+        if self.stored is not None and self.stored in self.blacklist:
+            return None
+        return self.hashobj.digest()
+
+    def hexdigest(self):
+        if self.stored is not None and self.stored in self.blacklist:
+            return None
+        return self.hashobj.hexdigest()
+
+    def copy(self):
+        return HashBlacklistContent(self.hashobj.copy(), self.blacklist,
+                                    self.maxlen)
+
 class DecompressedHash(object):
     """Apply a decompression function before the hash. This class provides the
     hashlib interface (update, hexdigest, copy) excluding digest and name."""
@@ -49,9 +99,13 @@ class DecompressedHash(object):
 
     def hexdigest(self):
         if not hasattr(self.decompressor, "flush"):
+            if self.decompressor.unused_data:
+                raise ValueError("decompressor did not consume all data")
             return self.hashobj.hexdigest()
         tmpdecomp = self.decompressor.copy()
         data = tmpdecomp.flush()
+        if tmpdecomp.unused_data:
+            raise ValueError("decompressor did not consume all data")
         tmphash = self.hashobj.copy()
         tmphash.update(data)
         return tmphash.hexdigest()
@@ -61,7 +115,7 @@ class DecompressedHash(object):
 
 class SuppressingHash(object):
     """A hash that silences exceptions from the update and hexdigest methods of
-    a hashlib-like object. If an exception has occured, hexdigest always
+    a hashlib-like object. If an exception has occurred, hexdigest always
     returns None."""
     def __init__(self, hashobj, exceptions=()):
         """