Skip to content

Providers occupy too much memory #968

@neumond

Description

@neumond

Bug report

What's wrong

Open a REPL, create not-so-huge amount of providers:

>>> a = [mimesis.Business() for _ in range(10000)]

Not examine memory footprint of the process: in my case it's 1,1g.

How is that should be

Why would I need so many providers? I never knew they're such heavy objects, I was using them naively in tests:

@pytest.fixture
def user_factory():
  mgen = mimesis.Person()
  async def user_factory():
    name = mgen.name()
    ...
  return user_factory

Seems like pytest keeps references to each fixture value, even if the test has finished. In our case it's closure with mimesis provider. For each test we create a provider or several of them. This leads to huge memory consumption of test suite.

I agree that correct use would be

@pytest.fixture(scope='session')
def mimesis_person_gen():
  return mimesis.Person()

@pytest.fixture
def user_factory(mimesis_person_gen):
  ...

this fixes memory consumption issue. But there's no any mention in documentation that provider objects are that heavy and one should construct them as few as possible. Probably issuing a warning on constructing 1000th provider would help many people who try to find out why their tests consume so much memory.

Metadata

Metadata

Assignees

No one assigned

    Labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions