repository·master·Indexed 26 days ago
https://github.com/lk-geimfari/mimesisA high-performance Python library for generating realistic fake data across 47 different locales. Designed for populating test databases, mocking APIs, and creating sample datasets, Mimesis provides localized providers (Address, Finance, Person, etc.), universal providers (Internet, Numeric, Science, etc.), and schema-based generation. It is optimized for speed, significantly outperforming Faker in benchmarks, and offers integration with factory_boy.
Mimesis is a Python library designed to generate fake but realistic data across multiple languages and locales. It is suitable for populating test databases, mocking API responses, generating JSON/XML fixtures, creating sample datasets, and anonymizing production data.
Supported data types include:
Mimesis is a high-performance Python data generator used to populate databases, create complex JSON/XML files, anonymize production data, and generate Pandas dataframes.
Key capabilities include:
mimesis.builder.SchemaBuilder.Mimesis is designed for high-performance data generation and consistently outperforms Faker in speed across all tested operations. In benchmarks conducted on an Apple M1 Pro, Mimesis demonstrated a 100% win rate across 47 tested operations, providing an average speedup of approximately 23×.
Key performance characteristics include:
In summary, Mimesis operates at the nanosecond–microsecond scale, whereas Faker frequently operates at the microsecond–millisecond scale.
Mimesis is significantly faster than Faker across all tested providers. In benchmark tests involving 47 operations, Mimesis outperformed Faker in 100% of cases, achieving an overall speedup of approximately 23.18x.
Key performance highlights include:
generate_100_names): ~25.03x speedup.Mimesis uses data providers to generate various types of data (food, people, addresses, etc.). Providers are categorized into two types:
Person). If no locale is specified, Locale.EN is used by default.Code).Warning: Data providers are heavy objects because they load JSON data into memory. Avoid constructing an excessive number of provider instances.
Note: Attempting to pass a locale argument to a locale-independent provider will raise a TypeError.
To use Mimesis effectively, familiarize yourself with these core architectural terms:
mimesis.enums.Locale).key parameter (see mimesis.keys).The Generic provider provides a single interface to access all Mimesis providers for a specific locale. It automatically handles the distinction between locale-dependent and locale-independent providers.
from mimesis import Generic
from mimesis.locales import Locale
g = Generic(locale=Locale.ES)
g.datetime.month()
# Output: 'Agosto'
g.code.imei()
# Output: '353918052107063'
g.food.fruit()
# Output: 'Limón'To reproduce or refresh the performance and memory benchmarks for Mimesis, use the benchmark scripts located in the benchmarks directory of the repository.
Note that absolute timings and memory peaks will vary depending on your hardware and Python interpreter, but the relative performance gains (speedup and memory efficiency) are the primary metrics of interest.
Mimesis supports multiple locales to generate data in different languages and for different countries. The default locale is English (United States) (Locale.EN). To use a specific locale, import Locale from mimesis.locales and pass it to the provider's constructor using the locale argument.
from mimesis import Address
from mimesis.locales import Locale
de = Address(locale=Locale.DE)
ru = Address(locale=Locale.RU)
de.region()
# Output: 'Brandenburg'
ru.federal_subject()
# Output: 'Алтайский край'To implement domain-specific methods, subclass BaseProvider.
Requirements:
BaseProvider.Meta class containing at least a name attribute (this is the attribute name used to access the provider via Generic, e.g., generic.some_provider).Meta.auto_register.Registration Options (Meta.auto_register):
False (Recommended for apps/tests): You must manually attach the provider to Generic instances using add_provider() or add_providers(). This prevents leaking providers into unrelated objects.True (Default, for plugins/libraries): The provider is automatically registered in the ProviderRegistry at class definition time and becomes available to all new Generic and Field objects created after import.from mimesis.providers.base import BaseProvider
class SomeProvider(BaseProvider):
class Meta:
name = "some_provider"
auto_register = False
@staticmethod
def hello() -> str:
return "Hello!"You can also use override_locale with mimesis.Generic instances to temporarily change the locale for specific sub-providers (e.g., generic.text).
from mimesis import Generic
from mimesis.locales import Locale
generic = Generic(locale=Locale.EN)
generic.text.word()
# Output: 'anyone'
with generic.text.override_locale(Locale.FR):
generic.text.word()
# Output: 'mieux'
generic.text.word()
# Output: 'responsibilities'To define custom logic for fields that aren't standard Mimesis providers, define a field_handlers list within a Params inner class inside your factory. Each handler is a tuple containing the field name and a callable (e.g., a lambda) that accepts rand (the Mimesis instance) and **kwargs.
import factory
from mimesis.plugins.factory import FactoryField
class FactoryWithCustomFieldHandlers(factory.Factory):
class Meta(object):
model = Guest
class Params(object):
field_handlers = [
("num", lambda rand, **kwargs: rand.randint(1, 99)),
("nick", lambda rand, **kwargs: rand.choice(["john", "alice"])),
]
age = FactoryField("num")
nickname = FactoryField("nick")