Fix a string reference collision bug with AMF3

cpyamf/amf3.pxd

@@ -20,7 +20,7 @@ cdef class ClassDefinition(object):
 
 
 cdef class Context(codec.Context):
-    cdef codec.IndexedCollection strings
+    cdef codec.ByteStringReferenceCollection strings
     cdef dict classes
     cdef dict class_ref
     cdef dict proxied_objects

cpyamf/amf3.pyx

@@ -141,7 +141,7 @@ cdef class Context(codec.Context):
     """
 
     def __cinit__(self):
-        self.strings = codec.IndexedCollection(use_hash=1)
+        self.strings = codec.ByteStringReferenceCollection()
         self.classes = {}
         self.class_ref = {}
         self.proxied_objects = {}

cpyamf/codec.pxd

@@ -29,6 +29,17 @@ cdef class IndexedCollection(object):
     cpdef Py_ssize_t append(self, object obj) except -1
 
 
+cdef class ByteStringReferenceCollection(IndexedCollection):
+    """
+    There have been rare hash collisions within a single AMF payload causing
+    corrupt payloads.
+
+    Which strings cause collisions is dependent on the python runtime, each
+    platform might have a slightly different implementation which means that
+    testing is extremely difficult.
+    """
+
+
 cdef class Context(object):
     """
     C based version of ``pyamf.BaseContext``

cpyamf/codec.pyx

@@ -120,7 +120,7 @@ cdef class IndexedCollection(object):
 
         return <object>self.data[ref]
 
-    cdef inline object _ref(self, object obj):
+    cdef object _ref(self, object obj):
         if self.use_hash:
             return hash(obj)
 
@@ -198,6 +198,25 @@ cdef class IndexedCollection(object):
         return n
 
 
+cdef class ByteStringReferenceCollection(IndexedCollection):
+    """
+    There have been rare hash collisions within a single AMF payload causing
+    corrupt payloads.
+
+    Which strings cause collisions is dependent on the python runtime, each
+    platform might have a slightly different implementation which means that
+    testing is extremely difficult.
+    """
+
+    cdef object _ref(self, object obj):
+        return obj
+
+    def __copy__(self):
+        cdef ByteStringReferenceCollection n = ByteStringReferenceCollection()
+
+        return n
+
+
 cdef class Context(object):
     """
     I hold the AMF context for en/decoding streams.

pyamf/amf3.py

@@ -600,13 +600,13 @@ class Context(codec.Context):
     I hold the AMF3 context for en/decoding streams.
 
     @ivar strings: A list of string references.
-    @type strings: C{list}
+    @type strings: L{codec.ByteStringReferenceCollection}
     @ivar classes: A list of L{ClassDefinition}.
     @type classes: C{list}
     """
 
     def __init__(self):
-        self.strings = codec.IndexedCollection(use_hash=True)
+        self.strings = codec.ByteStringReferenceCollection()
         self.classes = {}
         self.class_ref = {}
 

pyamf/codec.py

@@ -120,6 +120,30 @@ def __repr__(self):
             id(self))
 
 
+class ByteStringReferenceCollection(IndexedCollection):
+    """
+    There have been rare hash collisions within a single AMF payload causing
+    corrupt payloads.
+
+    Which strings cause collisions is dependent on the python runtime, each
+    platform might have a slightly different implementation which means that
+    testing is extremely difficult.
+    """
+
+    def __init__(self, *args, **kwargs):
+        super(ByteStringReferenceCollection, self).__init__(use_hash=False)
+
+    def getReferenceTo(self, byte_string):
+        return self.dict.get(byte_string, -1)
+
+    def append(self, byte_string):
+        self.list.append(byte_string)
+        idx = len(self.list) - 1
+        self.dict[byte_string] = idx
+
+        return idx
+
+
 class Context(object):
     """
     The base context for all AMF [de|en]coding.
@@ -215,13 +239,12 @@ def getStringForBytes(self, s):
 
         @since: 0.6
         """
-        h = hash(s)
-        u = self._unicodes.get(h, None)
+        u = self._unicodes.get(s, None)
 
         if u is not None:
             return u
 
-        u = self._unicodes[h] = s.decode('utf-8')
+        u = self._unicodes[s] = s.decode('utf-8')
 
         return u
 
@@ -232,13 +255,12 @@ def getBytesForString(self, u):
 
         @since: 0.6
         """
-        h = hash(u)
-        s = self._unicodes.get(h, None)
+        s = self._unicodes.get(u, None)
 
         if s is not None:
             return s
 
-        s = self._unicodes[h] = u.encode('utf-8')
+        s = self._unicodes[u] = u.encode('utf-8')
 
         return s