Lazy Attributes for Python Objects

1.6 · active · verified Tue Apr 14

The `lazy` library (version 1.6) provides a Python descriptor that allows object attributes to be computed only when they are first accessed, and then caches the result for subsequent retrievals. It enables efficient resource usage by deferring potentially expensive computations until absolutely necessary. The project maintains an active, though not rapid, release cadence, with the latest update in 2023.

Warnings

gotcha Lazy attributes are computed only once and then cached. If the underlying data or state that the lazy attribute depends on changes after its first access, the attribute will continue to return the stale cached value.
Fix: Ensure the underlying data is immutable or clear and re-initialize the object if fresh computation is required.
gotcha Any side effects within the method decorated with `@lazy` (e.g., logging, modifying other object state) will occur only on the first access when the attribute is initially computed. Subsequent accesses will retrieve the cached value without re-executing the method.
Fix: Avoid relying on side effects for subsequent accesses or ensure side effects are idempotent and safe to run once.
gotcha The library's `@lazy` decorator may not be inherently thread-safe. If multiple threads concurrently access a lazy attribute for the very first time, it could lead to the underlying computation running multiple times or result in a race condition for setting the cached value.
Fix: Implement explicit locking or ensure thread synchronization if the lazy attribute is accessed by multiple threads before its initial computation.

Install

pip install lazy Install latest version

Imports

lazy
```
from lazy import lazy
```

Quickstart

Demonstrates the basic usage of the `@lazy` decorator on a method, turning it into a lazy-loaded, cached attribute. The computation ('computing bar') runs only on the first access.

from lazy import lazy

class Foo(object):
    @lazy
    def bar(self):
        print("computing bar")
        return 42

f = Foo()
print(f.bar)
print(f.bar)

view raw JSON →