A demo of pickle usage vs. random usage respective to existing and proposed

pickle_random_testing.py

import pickle
import os.path
import random


# Old pickle queue logic (summarized)

pickle_cache_file = "pickle.queue"

reset_list_count = 0
assignment_rounds = 50
assignee_list_rotation_events = 0

assignment_counts_pickle = {
    "user1@mozilla.com": 0,
    "user2@mozilla.com": 0,
    "user3@mozilla.com": 0
}

# If a queue is not initialized, create one
if not os.path.isfile(pickle_cache_file):
    assign_hash = ['user1@mozilla.com',
                   'user2@mozilla.com', 'user3@mozilla.com']
    assign_list = assign_hash[:]
    with open(pickle_cache_file, 'wb') as f:
        pickle.dump((assign_list, assign_hash), f)

# Emulate a specific round
for iteration in range(1, assignment_rounds):
    # Incremeent reset list counter
    reset_list_count += 1

    with open(pickle_cache_file, 'rb') as f:
        assign_list, assign_hash = pickle.load(f)

    # Emulate a list reset at arbitrary times, to simulate an assignee edit
    if reset_list_count % random.randint(1, 28) == 0:
        assignee_list_rotation_events += 1
        assign_hash = ['user1@mozilla.com',
                       'user2@mozilla.com', 'user3@mozilla.com']
        assign_list = assign_hash[:]

    assignee = assign_list.pop()
    assign_list.insert(0, assignee)

    # Add a count to whomever was assigned
    assignment_counts_pickle[assignee] += 1

    with open(pickle_cache_file, 'wb') as f:
        pickle.dump((assign_list, assign_hash), f)

print("Pickle Strategy... (with {} rotation edits)".format(
    str(assignee_list_rotation_events)))
print(assignment_counts_pickle)

# New random.choice logic (summarized)
assign_hash = ['user1@mozilla.com',
               'user2@mozilla.com', 'user3@mozilla.com']

assignment_counts_random = {
    "user1@mozilla.com": 0,
    "user2@mozilla.com": 0,
    "user3@mozilla.com": 0
}

for iteration in range(1, assignment_rounds):
    assignee = random.choice(assign_hash)
    assignment_counts_random[assignee] += 1

print("Random Strategy...")
print(assignment_counts_random)

$ python ~/Desktop/pickle_test.py
Pickle Strategy... (with 10 rotation edits)
{'user1@mozilla.com': 13, 'user2@mozilla.com': 16, 'user3@mozilla.com': 20}
Random Strategy...
{'user1@mozilla.com': 20, 'user2@mozilla.com': 13, 'user3@mozilla.com': 16}

Note that results are not deterministic, so you will see differences depending on exactly when the list rotation is reset.

TLDR: I don't think either are as deterministic as we want in 50 assignment sample.

I ran a case just on pickle itself:

import pickle


try:
    with open('test', 'rb') as f:
       u = pickle.load(f)
except: 
    u = ['z', 'a', 'b', 'c']

print ('current', u)

with open ('test', 'wb') as f:
    pickle.dump(u, f)

/tmp % for i in $(seq 1 10); do python ./a.py; done                                                                                                                                   :(
current ['z', 'a', 'b', 'c']
current ['z', 'a', 'b', 'c']
current ['z', 'a', 'b', 'c']
current ['z', 'a', 'b', 'c']
current ['z', 'a', 'b', 'c']
current ['z', 'a', 'b', 'c']
current ['z', 'a', 'b', 'c']
current ['z', 'a', 'b', 'c']
current ['z', 'a', 'b', 'c']
current ['z', 'a', 'b', 'c']

ie it seems to keep the serialization the same (even without assign_hash)

maybe you're just testing https://gist.github.com/claudijd/15ccfe5f72841ff4bc2143728ae2ffd6#file-pickle_random_testing-py-L37 in this case

with rotation logic:

import pickle


try:
    with open('test', 'rb') as f:
       u = pickle.load(f)
except: 
    u = ['z', 'a', 'b', 'c']

print ('current', u)

# rotate
a = u.pop()
u.insert(0, a)

with open ('test', 'wb') as f:
    pickle.dump(u, f)

/tmp % for i in $(seq 1 10); do python ./a.py; done
current ['z', 'a', 'b', 'c']
current ['c', 'z', 'a', 'b']
current ['b', 'c', 'z', 'a']
current ['a', 'b', 'c', 'z']
current ['z', 'a', 'b', 'c']
current ['c', 'z', 'a', 'b']
current ['b', 'c', 'z', 'a']
current ['a', 'b', 'c', 'z']
current ['z', 'a', 'b', 'c']
current ['c', 'z', 'a', 'b']