Pre-pack outgoing TLMessage

This has several benefits:
- The message can be resent without re-calling bytes(),
  which for some requests may be expensive.
- Converting requests to bytes early lets us detect
  errors early, such as OverflowError on bad requests.
- Containers can't exceed 1044456 bytes so knowing their
  length is important. This can now be done in O(1).

But also several drawbacks:
- If the object is modified the bytes won't reflect this.
  This isn't an issue because it's only done for in msgs.
- Incoming messages can no longer be reconverted into
  bytes but this was never needed anyway.

telethon/network/mtprotosender.py

@@ -507,7 +507,6 @@ class MTProtoSender:
                       rpc_result.req_msg_id)
 
         if rpc_result.error:
-            # TODO Report errors if possible/enabled
             error = rpc_message_to_error(rpc_result.error)
             self._send_queue.put_nowait(self.state.create_message(
                 MsgsAck([message.msg_id])
@@ -517,10 +516,13 @@ class MTProtoSender:
                 message.future.set_exception(error)
             return
         elif message:
+            # TODO Would be nice to avoid accessing a per-obj read_result
+            # Instead have a variable that indicated how the result should
+            # be read (an enum) and dispatch to read the result, mostly
+            # always it's just a normal TLObject.
             with BinaryReader(rpc_result.body) as reader:
                 result = message.obj.read_result(reader)
 
-            # TODO Process entities
             if not message.future.cancelled():
                 message.future.set_result(result)
             return

telethon/network/mtprotostate.py

@@ -46,7 +46,8 @@ class MTProtoState:
             msg_id=self._get_new_msg_id(),
             seq_no=self._get_seq_no(isinstance(obj, TLRequest)),
             obj=obj,
-            after_id=after.msg_id if after else None
+            after_id=after.msg_id if after else None,
+            out=True  # Pre-convert the request into bytes
         )
 
     def update_message_id(self, message):

telethon/tl/core/tlmessage.py

@@ -21,9 +21,7 @@ class TLMessage(TLObject):
     sent `TLMessage`, and this result can be represented as a `Future`
     that will eventually be set with either a result, error or cancelled.
     """
-    def __init__(self, msg_id, seq_no, obj=None, after_id=0):
-        self.msg_id = msg_id
-        self.seq_no = seq_no
+    def __init__(self, msg_id, seq_no, obj, out=False, after_id=0):
         self.obj = obj
         self.container_msg_id = None
         self.future = asyncio.Future()
@@ -31,23 +29,56 @@ class TLMessage(TLObject):
         # After which message ID this one should run. We do this so
         # InvokeAfterMsgRequest is transparent to the user and we can
         # easily invoke after while confirming the original request.
+        # TODO Currently we don't update this if another message ID changes
         self.after_id = after_id
 
+        # There are two use-cases for the TLMessage, outgoing and incoming.
+        # Outgoing messages are meant to be serialized and sent across the
+        # network so it makes sense to pack them as early as possible and
+        # avoid this computation if it needs to be resent, and also shows
+        # serializing-errors as early as possible (foreground task).
+        #
+        # We assume obj won't change so caching the bytes is safe to do.
+        # Caching bytes lets us get the size in a fast way, necessary for
+        # knowing whether a container can be sent (<1MB) or not (too big).
+        #
+        # Incoming messages don't really need this body, but we save the
+        # msg_id and seq_no inside the body for consistency and raise if
+        # one tries to bytes()-ify the entire message (len == 12).
+        if not out:
+            self._body = struct.pack('<qi', msg_id, seq_no)
+        else:
+            if self.after_id is None:
+                body = GzipPacked.gzip_if_smaller(self.obj)
+            else:
+                body = GzipPacked.gzip_if_smaller(
+                    InvokeAfterMsgRequest(self.after_id, self.obj))
+
+            self._body = struct.pack('<qii', msg_id, seq_no, len(body)) + body
+
     def to_dict(self):
         return {
             '_': 'TLMessage',
             'msg_id': self.msg_id,
             'seq_no': self.seq_no,
             'obj': self.obj,
-            'container_msg_id': self.container_msg_id,
-            'after_id': self.after_id
+            'container_msg_id': self.container_msg_id
         }
 
-    def __bytes__(self):
-        if self.after_id is None:
-            body = GzipPacked.gzip_if_smaller(self.obj)
-        else:
-            body = GzipPacked.gzip_if_smaller(
-                InvokeAfterMsgRequest(self.after_id, self.obj))
+    @property
+    def msg_id(self):
+        return struct.unpack('<q', self._body[:8])[0]
 
-        return struct.pack('<qii', self.msg_id, self.seq_no, len(body)) + body
+    @msg_id.setter
+    def msg_id(self, value):
+        self._body = struct.pack('<q', value) + self._body[8:]
+
+    @property
+    def seq_no(self):
+        return struct.unpack('<i', self._body[8:12])[0]
+
+    def __bytes__(self):
+        if len(self._body) == 12:  # msg_id, seqno
+            raise TypeError('Incoming messages should not be bytes()-ed')
+
+        return self._body