Skip to main content

LazyImporter class to manage imports on attribute access.

Project description

ducktools: lazyimporter

Create an object to handle lazily importing from other modules.

Nearly every form of "lazyimporter" module name is taken on PyPI so this is namespaced.

Intended to help save on start time where some modules are only needed for specific functions while allowing information showing the import information to appear at the top of a module where expected.

This form of import works by creating a specific LazyImporter object that lazily imports modules or module attributes when the module or attribute is accessed on the object.

How to download

Download from PyPI: python -m pip install ducktools-lazyimporter

Example

Example using the packaging module.

__version__ = "v0.1.5"

from ducktools.lazyimporter import LazyImporter, FromImport

laz = LazyImporter([
    FromImport("packaging.version", "Version")
])

def is_newer_version(version_no: str) -> bool:
    """Check if a version number given indicates 
    a newer version than this package."""
    this_ver = laz.Version(__version__) 
    new_ver = laz.Version(version_no)
    return new_ver > this_ver

# Import will only occur when the function is called and 
# laz.Version is accessed
print(is_newer_version("v0.2.0"))

Hasn't this already been done

Yes.

But...

Most implementations rely on stdlib modules that are themselves slow to import (for example: typing, importlib.util, logging, inspect, ast). By contrast lazyimporter only uses modules that python imports on launch as part of site.

lazyimporter does not attempt to propagate laziness, only the modules provided to lazyimporter directly will be imported lazily. Any subdependencies of those modules will be imported eagerly as if the import statement is placed where the importer attribute is first accessed.

Use Case

There are two main use cases this is designed for.

Replacing in-line imports used in a module

Sometimes it is useful to use tools from a module that has a significant import time. If this is part of a function/method that won't necessarily always be used it is common to delay the import and place it inside the function/method.

Regular import within function:

def get_copy(obj):
    from copy import deepcopy
    return deepcopy(obj)

With a LazyImporter:

from ducktools.lazyimporter import LazyImporter, FromImport

laz = LazyImporter([FromImport("copy", "deepcopy")])

def get_copy(obj):
    return laz.deepcopy(obj)

While the LazyImporter is more verbose, it only invokes the import mechanism once when first accessed, while placing the import within the function invokes it every time the function is called. This can be a significant overhead if the function ends up used in a loop.

This also means that if the attribute is accessed anywhere it will be imported and in place wherever it is used.

Delaying the import of parts of a module's public API

Eager import:

from .submodule import useful_tool

__all__ = [..., "useful_tool"]

Lazy import:

from ducktools.lazyimporter import LazyImporter, FromImport, get_module_funcs

__all__ = [..., "useful_tool"]

laz = LazyImporter(
    [FromImport(".submodule", "useful_tool")],
    globs=globals(),  # If relative imports are used, globals() must be provided.
)
__getattr__, __dir__ = get_module_funcs(laz, __name__)

The import classes

In all of these instances modules is intended as the first argument to LazyImporter and all attributes would be accessed from the LazyImporter instance and not in the global namespace.

eg:

from ducktools.lazyimporter import LazyImporter, ModuleImport

modules = [ModuleImport("functools")]
laz = LazyImporter(modules)
laz.functools  # provides access to the module "functools"

ModuleImport

ModuleImport is used for your basic module style imports.

from ducktools.lazyimporter import ModuleImport

modules = [
    ModuleImport("module"),
    ModuleImport("other_module", "other_name"),
    ModuleImport("base_module.submodule", asname="short_name"),
]

is equivalent to

import module
import other_module as other_name
import base_module.submodule as short_name

when provided to a LazyImporter.

FromImport and MultiFromImport

FromImport is used for standard 'from' imports, MultiFromImport for importing multiple items from the same module. By using a MultiFromImport, when the first attribute is accessed, all will be assigned on the LazyImporter.

from ducktools.lazyimporter import FromImport, MultiFromImport

modules = [
    FromImport("dataclasses", "dataclass"),
    FromImport("functools", "partial", "partfunc"),
    MultiFromImport("collections", ["namedtuple", ("defaultdict", "dd")]),
]

is equivalent to

from dataclasses import dataclass
from functools import partial as partfunc
from collections import namedtuple, defaultdict as dd

when provided to a LazyImporter.

TryExceptImport

TryExceptImport is used for compatibility where a module may not be available and so a fallback module providing the same functionality should be used. For example when a newer version of python has a stdlib module that has replaced a third party module that was used previously.

from ducktools.lazyimporter import TryExceptImport

modules = [
    TryExceptImport("tomllib", "tomli", "tomllib"),
]

is equivalent to

try:
    import tomllib as tomllib
except ImportError:
    import tomli as tomllib

when provided to a LazyImporter.

How does it work

The following lazy importer:

from ducktools.lazyimporter import LazyImporter, FromImport

laz = LazyImporter([FromImport("functools", "partial")])

Generates an object that's roughly equivalent to this:

class SpecificLazyImporter:
    def __getattr__(self, name):
        if name == "partial":
            from functools import partial
            setattr(self, name, partial)
            return partial
        
        raise AttributeError(...)

laz = SpecificLazyImporter()

The first time the attribute is accessed the import is done and the output is stored on the instance, so repeated access immediately gets the desired object and the import mechanism is only invoked once.

(The actual __getattr__ function uses a dictionary lookup and delegates importing to the FromImport class. Names are all dynamic and imports are done through the __import__ function.)

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

ducktools-lazyimporter-0.4.0.tar.gz (21.4 kB view hashes)

Uploaded Source

Built Distribution

ducktools_lazyimporter-0.4.0-py3-none-any.whl (11.5 kB view hashes)

Uploaded Python 3

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page