Added new stream type which will request its size from its stream. This triggers full decompression, currently, but allows work to be delayed even further.If people want a stream, they usually read it anyway. Then it doesn't matter which attriubute they query first

Author: Byron

base.py

@@ -135,6 +135,23 @@ def read(self, size=-1):
 	@property
 	def stream(self):
 		return self[3]
+		
+	#} END stream reader interface
+	
+	
+class ODeltaStream(OStream):
+	"""Uses size info of its stream, delaying reads"""
+	
+	def __new__(cls, sha, type, size, stream, *args, **kwargs):
+		"""Helps with the initialization of subclasses"""
+		return tuple.__new__(cls, (sha, type, size, stream))
+	
+	#{ Stream Reader Interface
+	
+	@property
+	def size(self):
+		return self[3].size
+		
 	#} END stream reader interface
 
 

pack.py

@@ -34,6 +34,7 @@
 						OStream,
 						OPackInfo,
 						OPackStream,
+						ODeltaStream,
 						ODeltaPackInfo,
 						ODeltaPackStream,
 					)
@@ -584,14 +585,9 @@ def _object(self, sha, as_stream, index=-1):
 			# To prevent it from applying the deltas when querying the size, 
 			# we extract it from the delta stream ourselves
 			streams = self.collect_streams_at_offset(offset)
-			buf = streams[0].read(512)
-			offset, src_size = msb_size(buf)
-			offset, target_size = msb_size(buf, offset)
-			
-			streams[0].stream.seek(0)				# assure it can be read by the delta reader
 			dstream = DeltaApplyReader.new(streams)
 
-			return OStream(sha, dstream.type, target_size, dstream) 
+			return ODeltaStream(sha, dstream.type, None, dstream) 
 		else:
 			if type_id not in delta_types:
 				return OInfo(sha, type_id_to_type_map[type_id], uncomp_size)