Restore original behavior of `min_size` in `remove_small_*` during its deprecation by lagru · Pull Request #8003 · scikit-image/scikit-image

lagru · 2026-01-06T16:20:28Z

Description

The renaming of min_size/area_threshold to max_size in #7739 was actually accompanied by a change in behaviour for the same value, due to using > in the old nomenclature and >= in the new one.

Unfortunately our deprecation machinery was until now not smart enough to know whether a renamed parameter was provided under its old name or its new name. To address this, this PR:

Updates the deprecation machinery to detect when a parameter was given with the old name/position, and
Uses this updated machinery to modify min_size/area_threshold by 1 to ensure that the new behaviour matches the old one.

Checklist

A descriptive but concise pull request title
Docstrings for all functions
Unit tests
A gallery example in ./doc/examples for new features
Contribution guide is followed

Release note

For maintainers and optionally contributors, please refer to the instructions on how to document this PR for the release notes.

In `skimage.morphology`, fix an unintended change in deprecated behavior in
`remove_small_objects` and `remove_small_holes`. The deprecation of the 
parameters `min_size` and `area_threshold` introduced an off-by-one error when
using them.

to signal that a deprecated parameter received a value that was forwarded to the new parameter. This allows correcting for potentially changed behavior between deprecated and new parameter.

The deprecation accidentally changed the behavior of the deprecated `min_size`. However, the original behavior should stay available during the deprecation period! This fixes that. Furthermore, the new default must be `max_size=63` to be equivalent to the old `min_size=64`.

jmuhlich · 2026-01-06T22:01:31Z

I submitted a PR against your feature branch that fixes some failing tests. I thought it would be easier that way than to add a code review comment explaining the changes!

Fix deprecate_parameter tests for DEPRECATED_GOT_VALUE

jni

Clever. Maybe too clever 😂 but I'll allow it. 😂 Thanks @lagru!

jni · 2026-01-14T12:15:11Z

(And thank you @jmuhlich for the report and for contributing to this fix! 🙏)

jni · 2026-01-14T12:22:50Z

@lagru Want to update the description before merging? I like to have the final commit message describe the problem and fix more fully than just GH links. Example:

Fixes #8001

The renaming of min_size/area_threshold to max_size in #7739 was actually accompanied by a change in behaviour for the same value, due to using > in the old nomenclature and >= in the new one.

Unfortunately our deprecation machinery was until now not smart enough to know whether a renamed parameter was provided under its old name or its new name. To address this, this PR:

Updates the deprecation machinery to detect when a parameter was given with the old name/position, and

Uses this updated machinery to modify min_size/area_threshold by 1 to ensure that the new behaviour matches the old one.

lagru · 2026-01-14T12:34:41Z

Done. I'll just merge right now.

lagru · 2026-01-14T12:36:49Z

Feel welcome to edit the release note in the PR description further if you want. :)

lagru · 2026-01-14T12:38:12Z

Clever. Maybe too clever 😂 but I'll allow it. 😂

Yeah, I feel like that decorator is getting out of hand. If we had infinite time, I'd just refactor and split it for the different use cases.

stefanv

Darnit; I reviewed this a while ago and forgot to press "submit" 🤦

stefanv · 2026-01-13T21:17:43Z

src/skimage/_shared/utils.py

    This is useful if ``None`` is already used for a different purpose or just
    to highlight a deprecated parameter in the signature.
+
+    This is a sentinel and not meant to be initialized.


Suggested change

This is a sentinel and not meant to be initialized.

This is a sentinel and is not meant to be initialized.

stefanv · 2026-01-13T21:18:26Z

src/skimage/_shared/utils.py

+    """
+
+
+class DEPRECATED_GOT_VALUE(metaclass=PatchClassRepr):


Since this is internal, I'd opt for a descriptive name: DEPRECATED_BUT_RECEIVED_VALUE or DEPRECATED_BUT_SET_BY_USER.

Actually, looking at the code below, another way to spell this would simply be RECEIVED_VALUE or SET_BY_USER, since it is used on the module: deprecate_parameter.SET_BY_USER or deprecate_paramer.RECEIVED_VALUE. (Would read better as deprecated_parameter.)

stefanv · 2026-01-13T21:20:32Z

src/skimage/morphology/misc.py

 )
 def remove_small_objects(
-    ar, min_size=DEPRECATED, connectivity=1, *, max_size=64, out=None
+    ar, min_size=DEPRECATED, connectivity=1, *, max_size=63, out=None


Did this change inadvertently?

stefanv · 2026-01-14T20:28:25Z

No action needed on my late submission review; I'll make a PR for quick review, which you can choose to accept or not.

lagru added 2 commits January 6, 2026 17:12

Use DEPRECATED_GOT_VALUE in deprecate_param

388b8ca

to signal that a deprecated parameter received a value that was forwarded to the new parameter. This allows correcting for potentially changed behavior between deprecated and new parameter.

lagru added the 🩹 type: Bug fix Fixes unexpected or incorrect behavior label Jan 6, 2026

Fix deprecate_parameter tests for DEPRECATED_GOT_VALUE

8204ecd

lagru and others added 2 commits January 7, 2026 10:33

Merge pull request #1 from jmuhlich/pr-8003-fix1

79e9437

Fix deprecate_parameter tests for DEPRECATED_GOT_VALUE

Merge branch 'main' into fix-rm-small-obj-min-size-depr

96b9190

lagru marked this pull request as ready for review January 8, 2026 11:43

Fix remove_small_holes too

d3f76b2

lagru changed the title ~~Restore original behavior of min_size in remove_small_objects during its deprecation~~ Restore original behavior of min_size in remove_small_* during its deprecation Jan 13, 2026

jni approved these changes Jan 14, 2026

View reviewed changes

lagru merged commit dc02027 into scikit-image:main Jan 14, 2026
25 checks passed

stefanv added this to the 0.27 milestone Jan 14, 2026

lagru deleted the fix-rm-small-obj-min-size-depr branch January 14, 2026 12:36

stefanv reviewed Jan 14, 2026

View reviewed changes

ethanbb mentioned this pull request Feb 19, 2026

Exclude scikit-image 0.26.0 due to remove_small_* regression flatironinstitute/CaImAn#1571

Merged

1 task

	This is a sentinel and not meant to be initialized.
	This is a sentinel and is not meant to be initialized.

Uh oh!

Conversation

lagru commented Jan 6, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Description

Checklist

Release note

Uh oh!

jmuhlich commented Jan 6, 2026

Uh oh!

jni left a comment

Choose a reason for hiding this comment

Uh oh!

jni commented Jan 14, 2026

Uh oh!

jni commented Jan 14, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

lagru commented Jan 14, 2026

Uh oh!

Uh oh!

lagru commented Jan 14, 2026

Uh oh!

lagru commented Jan 14, 2026

Uh oh!

stefanv left a comment

Choose a reason for hiding this comment

Uh oh!

stefanv Jan 13, 2026

Choose a reason for hiding this comment

Uh oh!

stefanv Jan 13, 2026

Choose a reason for hiding this comment

Uh oh!

stefanv Jan 13, 2026

Choose a reason for hiding this comment

Uh oh!

stefanv Jan 13, 2026

Choose a reason for hiding this comment

Uh oh!

stefanv commented Jan 14, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

lagru commented Jan 6, 2026 •

edited

Loading

jni commented Jan 14, 2026 •

edited

Loading