Skip to main content

More Imports! - Delayed importing

Project description

More Imports! - Delayed importing

A few methods to make late importing cleaner

Branch Status
master Build Status
dev Build Status

Problem

Splitting code into modules is nice, but it can result in cyclic dependencies.

foos.py

from bars import bar

def foo():
    bar()

bars.py

from foos import foo

def bar():
    foo()

We are not concerned with the infinite recursion; this is only for demonstrating cyclic dependencies.

More Imports!

Solution #1: Use expect/export pattern

All your cyclic dependencies are covered with this one pattern: Break cycles by expecting a name in the first module, and let the second module export to the first when the value is available

foos.py

from mo_imports import expect

bar = expect("bar")

def foo():
    bar()

bars.py

from mo_imports import export
from foos import foo

def bar():
    foo()

export("bars", bar)

Benefits

  • every expect is verified to match with an export (and visa-versa)
  • using an expected variable before export raises an error
  • code is run only once, at module load time, not later
  • methods do not run import code
  • all "imports" are at the top of the file

Solution #2: Use delay_import

Provide a proxy which is responsible for import upon first use of the module variable.

foos.py

from mo_imports import delay_import

bar = delay_import("bars.bar")

def foo():
    bar()

bars.py

from foos import foo

def bar():
    foo()

Benefits

  • cleaner code
  • costly imports are delayed until first use

WARNING

Requires any of __call__, __getitem__, __getattr__ to be called to trigger the import. This means sentinals, placeholders, and default values can NOT be imported using delay_import()

Other solutions

If you do not use mo-imports your import cycles can be broken using one of the following common patterns:

Bad Solution: Keep in single file

You can declare yet-another-module that holds the cycles

foosbars.py

    def foo():
        bar()

    def bar():
        foo()

but this breaks the code modularity

Bad Solution: Use end-of-file imports

During import, setup of the first module is paused while it imports a second. A bottom-of-file import will ensure the first module is mostly setup to be used by the second.

foos.py

def foo():
    bar()

from bars import bar

bars.py

def bar():
    foo()

from foos import foo

Linters do not like this pattern: You may miss imports, since these are hiding at the bottom.

Bad Solution: Inline import

Import the name only when it is needed

foos.py

def foo():
    from bars import bar
    bar()

bars.py

def bar():
    from foos import foo
    foo()

This is fine for rarely run code, but there is an undesirable overhead because import is checked everytime the method is run. You may miss imports because they are hiding inline rather than at the top of the file.

Bad Solution: Use the _late_import() pattern

When other bad solutions do not work work, then importing late is the remaining option

foos.py

from bars import bar

def foo():
    bar()

bars.py

foo = None

def _late_import():
    global foo
    from foos import foo
    _ = foo

def bar():
    if not foo:
        _late_import()
    foo()

Placeholders variables are added, which linters complain about type. There is the added _late_import() method. You risk it is not run everywhere as needed. This has less overhead than an inline import, but there is still a check.

More on importing

Importing a complex modular library is still hard; the complexity comes from the the order other modules declare their imports; you have no control over which of your modules will be imported first. For example, one module may

from my_lib.bars import bar
from my_lib.foos import foo

another module may choose the opposite order

from my_lib.foos import foo
from my_lib.bars import bar

Ordering imports

With cyclic dependencies, ordering the imports can get tricky. Here are some rules

  • choose your principle modules and the order you want them imported.
  • your remaining modules are assumed to be imported in alphabetical order (as most linters prefer)
  • use top level __init__.py to control the order of imports
  • encourage third party modules to use this top level module. for example
    from my_lib import foo, bar
    
  • finally, use mo_imports to break cycles

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

mo-imports-7.530.24035.tar.gz (12.8 kB view details)

Uploaded Source

Built Distribution

mo_imports-7.530.24035-py3-none-any.whl (11.6 kB view details)

Uploaded Python 3

File details

Details for the file mo-imports-7.530.24035.tar.gz.

File metadata

  • Download URL: mo-imports-7.530.24035.tar.gz
  • Upload date:
  • Size: 12.8 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/4.0.1 CPython/3.9.6

File hashes

Hashes for mo-imports-7.530.24035.tar.gz
Algorithm Hash digest
SHA256 fe9c52328dd8b7dcdc25c2f1257a7ad96c2959b2218d9a9586419b3292fafa12
MD5 e3c8c917045fa321bb58734d8768c5d6
BLAKE2b-256 032df6f60c02b343daa6bcd2e78a55ca0409dcb828e0d11ea6fee32b51737a8b

See more details on using hashes here.

File details

Details for the file mo_imports-7.530.24035-py3-none-any.whl.

File metadata

File hashes

Hashes for mo_imports-7.530.24035-py3-none-any.whl
Algorithm Hash digest
SHA256 7f8b9d5a2b41c61946af60c09172ac235da37cb7ae073c6807ada35874b15739
MD5 66c3ae1e2ff21c4204b942e7380b64e2
BLAKE2b-256 2a4f5c9e76ecf4b66fd05362e6fd1303b5d4f504aa0c8058a3da1353f5bc469c

See more details on using hashes here.

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