Python Performance and constructors

# Python Performance and constructors

'''This is the type of data I get, is a dictionary (I think is derived from a JSON object)'''
query_message = {'id': '000000000000000000',
 'from': {'id': 000000000, 'first_name': 'AAAAAAAAAAAAA', 'username': 'AAAAAAAAAA', 'language_code': 'aa-AA'},
 'message': {'message_id': 0000, 'from': {'id': 000000000,'first_name': 'AAAAAAAAAAAAAA', 'username': 'AAAAAAAAAAAA'},'chat': {'id': 000000000, 'first_name': 'AAAAAAAAAAAAA', 'username': 'AAAAAAAAAA', 'type': 'private'}, 'date': 1501131446, 'document': {'file_name': 'AAAAAAA.mp4', 'mime_type': 'video/mp4', 'thumb': {'file_id': 'AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA', 'file_size': 3262, 'file_path': 'thumbnails/20859.jpg', 'width': 90, 'height': 67}, 'file_id': 'AAAAAAAAAAAAAAAAAAAAAAAAAAAAAA', 'file_size': 766649}, 'caption': 'AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA'},
 'chat_instance': 'AAAAAAAAAAAAAAAAAAAA',
 'data': 'AAAAAAAAA'}

''' I constructed my class to parse it this way:
 since I will get many of this objects, and I don't want to process them all
 I thought to skip some steps
 I would like tho to keep the option to look at the arguments later on
 now with glance the class is built without the optional arguments
'''

class CbkQuery:

    def __init__(self, query_msg, glance = False, opt = True):

        #required arguments
        self.id = query_msg['id']
        self.person = Person(query_msg['from'])
        self.chat_instance = query_msg['chat_instance']

        if glance:
            pass
        else:
            self.query_msg = query_msg
            if opt:
                # optional arguments
                self.data = self.query_msg.get('data')
                self.message = self.query_msg.get('message')
                if self.message is not None:
                    self.message = Message(self.message)
                self.inline_message_id = self.query_msg.get('inline_message_id')
                self.game_short_name = self.query_msg.get('game_short_name')

''' Since I did this gimmick in order to have a fast reading, I wanted to time it
and see how much I save by "glancing" the class
'''

# test 1M times each constructor possibility
print("\n MY MESSAGE ")

lg.startTimer()
for i in range(1000000):
    qmsg = CbkQuery(msg, glance=True, opt=False)
lg.log("glance processing took: ", True)
# glance processing took:
# Elapsed time for subprocess: 00:00:01.749

lg.startTimer()
for i in range(100000):
    qmsg = CbkQuery(msg, glance=False, opt=False)
lg.log("store query, processing took: ", True)
# store query, processing took:
# Elapsed time for subprocess: 00:00:00.182

lg.startTimer()
for i in range(100000):
    qmsg = CbkQuery(msg)
lg.log("build complete class processing took: ", True)
# build complete class processing took:
# Elapsed time for subprocess: 00:00:01.322

''' surprise surprise, glancing the class takes longer, the fastest is to store the "unprocessed message in the class"
can somebody explain me why? is it because storing the unprocessed message actually makes a shallow copy instead of a copy?
'''