Add seconds to engine program id generation #3906
Conversation
- Clients were occasionally getting duplicate program ids when running successive results. - Root cause is unknown, but likely cause is that a sub-module was calling random.seed() somewhere. - This changes program id generation to include seconds in the id, so that successive calls will generate distinct ids even if the random number generator is seeded.
dstrain115
requested review from
cduck,
vtomole,
wcourtney and
a team
as code owners
| @@ -65,7 +65,7 @@ class ProtoVersion(enum.Enum): | |||
| def _make_random_id(prefix: str, length: int = 16): | |||
| random_digits = [random.choice(string.ascii_uppercase + string.digits) for _ in range(length)] | |||
| suffix = ''.join(random_digits) | |||
| suffix += datetime.date.today().strftime('%y%m%d') | |||
| suffix += datetime.date.today().strftime('%y%m%d-%H%M%S') | |||
There was a problem hiding this comment.
The problem with this second based approach that it assumes that if the faulty (assumingly seeding) client program is generating more than one program per second, you'll still get collisions. With UUID, neither seeding nor speed won't be a problem - though it will be longer (36 characters).
There was a problem hiding this comment.
A UUID based on what? I am not sure if that solves the problem.
Would you hash based on the time + circuit? If we use a random uuid, then that would be based on random() and also fail if seeded.
There was a problem hiding this comment.
Seems that https://docs.python.org/3/library/uuid.html#uuid.uuid4 is based on os.urandom(). It can't be seeded or manipulated.
Possibly, but you would have to generate more than one program in a second and also set random.seed() accidentally before generating each program. |
balopat
added
the
automerge
|
Let's experiment then! |
CirqBot
added
the
front_of_queue_automerge
|
Automerge cancelled: A required status check is not present. Missing statuses: ['Build docs', 'Build protos', 'Changed Notebooks Isolated Test against Cirq stable', 'Changed files test', 'Coverage check', 'Doc test', 'Format check', 'Lint check', 'Misc check', 'Notebook formatting', 'Pytest MacOS (3.7)', 'Pytest MacOS (3.8)', 'Pytest Ubuntu (3.7)', 'Pytest Ubuntu (3.8)', 'Pytest Windows (3.7)', 'Pytest Windows (3.8)', 'Type check'] |
CirqBot
removed
automerge
when running successive results.
was calling random.seed() somewhere.
so that successive calls will generate distinct ids even if the
random number generator is seeded.