Attach with websocket and do manual line buffering

This works around the odd byte sequences we see at the beginning of
every chunk when attaching via the streaming HTTP endpoint and a plain
socket.

plum/cli/log_printer.py

@@ -29,7 +29,8 @@ class LogPrinter(object):
 
     def _make_log_generator(self, container, color_fn):
         prefix = color_fn(container.name + " | ")
-        return (prefix + line for line in self._readlines(self._attach(container)))
+        websocket = self._attach(container)
+        return (prefix + line for line in split_buffer(read_websocket(websocket), '\n'))
 
     def _attach(self, container):
         params = {
@@ -41,13 +42,25 @@ class LogPrinter(object):
         }
         params.update(self.attach_params)
         params = dict((name, 1 if value else 0) for (name, value) in params.items())
-        return container.attach_socket(params=params)
+        return container.attach_socket(params=params, ws=True)
 
-    def _readlines(self, socket):
+def read_websocket(websocket):
-        for line in iter(socket.makefile().readline, b''):
+    while True:
-            if not line.endswith('\n'):
+        data = websocket.recv()
-                line += '\n'
+        if data:
+            yield data
+        else:
+            break
 
-            yield line
+def split_buffer(reader, separator):
+    buffered = ''
 
-        socket.close()
+    for data in reader:
+        lines = (buffered + data).split(separator)
+        for line in lines[:-1]:
+            yield line + separator
+        if len(lines) > 1:
+            buffered = lines[-1]
+
+    if len(buffered) > 0:
+        yield buffered