Skip to content

CHANGELOG.md

Latest commit

 

History

History
192 lines (112 loc) · 7.73 KB

CHANGELOG.md

File metadata and controls

192 lines (112 loc) · 7.73 KB
Raw

Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog and this project adheres to Calendar Versioning.

The first number of the version is the year. The second number is incremented with each release, starting at 1 for each year. The third number is for emergencies when we need to start branches for older releases.

You can find our backwards-compatibility policy here.

Unreleased

26.1.0 - 2026-04-13

Changed

  • Passing timeout=0 now raises a warning since it counterintuitively disables all retries. If you explicitly want to disable the timeout, use timeout=None.

Fixed

  • _compute_backoff() no longer raises OverflowError when the attempt number exceeds 1024 with a float wait_exp_base. The exponential result is now clamped to wait_max on overflow. #137

25.2.0 - 2025-12-11

Removed

  • Support for Python 3.8 and 3.9.

Added

  • The type hints for our public API are now also verified using Pyrefly and ty. #124

  • stamina.retry() now retries wrapped generator functions and async generator functions.

    Warning: Being able to asend and athrow into wrapped async generators introduced nontrivial complexity in the implementation and is therefore provisional. If supporting these features causes problems, they may be removed again in a future version. #123

  • An on hook can now return a float or a datetime.timedelta to specify a custom backoff that overrides the default backoff. #103 #125

Fixed

  • Prevent unbounded stop condition when both attempts and timeout are non-None falsy values. #109

  • Default wait_exp_base parameter is now an integer to prevent an OverflowError after the 1023th retry. #104

  • Attempt.next_wait now returns the correct value. #115

25.1.0 - 2025-03-12

Added

  • cap argument to stamina.set_testing(). By default, the value passed as attempts is used strictly. When cap=True, it is used as an upper cap; that means that if the original attempts number is lower, it's not changed. #80

  • stamina.set_testing() can now be used as a context manager. #94

  • Instrumentation hooks can now can return context managers. If they do, they are entered when a retry is scheduled and exited right before the retry is attempted. #95

24.3.0 - 2024-08-27

Added

  • The on argument in all retry functions now can be a callable that takes an exception and returns a bool which decides whether or not a retry should be scheduled. #70

  • stamina.Attempt now has a next_wait attribute that contains the time the next backoff will wait, if the current attempt fails (sans jitter). #72

  • It is now possible to switch stamina into a testing mode using stamina.set_testing(). It disables backoffs and caps the number of retries. #73

24.2.0 - 2024-01-31

Added

  • stamina.RetryingCaller and stamina.AsyncRetryingCaller that allow even easier retries of single callables: stamina.RetryingCaller(attempts=5).on(ValueError)(do_something, "foo", bar=42) and stamina.RetryingCaller(attempts=5)(ValueError, do_something, "foo", bar=42) will call do_something("foo", bar=42) and retry on ValueError up to 5 times.

    stamina.RetryingCaller and stamina.AsyncRetryingCaller take the same arguments as stamina.retry(), except for on that can be bound separately.

    #56 #57

24.1.0 - 2024-01-03

Fixed

  • stamina doesn't retry successful blocks when it's deactivated anymore (yes, you read it right). #54

23.3.0 - 2023-12-05

Added

23.2.0 - 2023-10-30

Added

  • Instrumentation is now pluggable! You can define your own hooks that are run with retry details whenever a retry is scheduled. The documentation now has a whole chapter on instrumentation. #37

  • If structlog is not installed, the scheduled retry is now logged using the standard library logging module by default. #35

Changed

  • Tenacity's internal AttemptManager object is no longer exposed to the user. This was an oversight and never documented. stamina.retry_context() now yields instances of stamina.Attempt. #22

  • Initialization of instrumentation is now delayed. This means that if there's no retries, there's no startup overhead from importing structlog and prometheus-client. #34

  • Some key names in structlog log messages have been renamed to better reflect their meaning (slept → waited_so_far, attempt → retry_num, and error → caused_by). You can rename them back using structlog's structlog.processors.EventRenamer. #35

23.1.0 - 2023-07-04

Added

  • Official Python 3.12 support. #9
  • Async support. #10
  • Retries of arbitrary blocks using (async) for loops and context managers. #12
  • Proper documentation. #16
  • A backwards-compatibility policy.

Changed

  • The timeout, wait_initial, wait_max, and wait_jitter arguments can now also be of type datetime.timedelta.

22.2.0 - 2022-10-06

Added

  • Retries are now instrumented. If prometheus-client is installed, retries are counted using the Prometheus counter stamina_retries_total. If structlog is installed, they are logged using a structlog logger at warning level. These two instrumentations are independent from each other.

22.1.0 - 2022-10-02

Added

  • Initial release.