SKIP 4 — Transitioning to scikit-image 2.0

Author:

Juan Nunez-Iglesias <juan.nunez-iglesias@monash.edu>

Author:

Lars Grüter

Status:

Draft

Type:

Standards Track

Created:

2022-04-08

Resolved:

<null>

Resolution:

<null>

Version effective:

None

Abstract

scikit-image is preparing to release version 1.0. This was seen as an opportunity to clean up the API, including backwards incompatible changes. Some of these changes involve changing return values without changing function signatures, which can ordinarily only be done by adding an otherwise useless keyword argument (such as new_return_style=True) whose default value changes over several releases. The result is still a backwards incompatible change, but made over a longer time period.

Despite being in beta and in a 0.x series of releases, scikit-image is used extremely broadly, and any backwards incompatible changes are likely to be disruptive. Given the rejection of SKIP-3, this document proposes an alternative pathway to create a new API. The new pathway involves the following steps:

  • Any pending deprecations that were scheduled for v0.20 and v0.21 are finalised (the new API suggested by deprecation messages in v0.19 becomes the only API).

  • This is released as 1.0.

  • At this point, the branch main changes the package and import names to skimage2, and the API is free to evolve.

Further motivation for the API changes is explained below, and largely duplicated from SKIP-3.

Motivation and Scope

scikit-image has grown organically over the past 12+ years, with functionality being added by a broad community of contributors from different backgrounds. This has resulted in various parts of the API being inconsistent: for example, skimage.transform.warp inverts the order of coordinates, so that a translation of (45, 32) actually moves the values in a NumPy array by 32 along the 0th axis, and 45 along the 1st, but only in 2D. In 3D, a translation of (45, 32, 77) moves the values in each axis by the number in the corresponding position.

Additionally, as our user base has grown, it has become apparent that certain early API choices turned out to be more confusing than helpful. For example, scikit-image will automatically convert images to various data types, rescaling them in the process. A uint8 image in the range [0, 255] will automatically be converted to a float64 image in [0, 1]. This might initially seem reasonable, but, for consistency, uint16 images in [0, 65535] are rescaled to [0, 1] floats, and uint16 images with 12-bit range in [0, 4095], which are common in microscopy, are rescaled to [0, 0.0625]. These silent conversions have resulted in much user confusion.

Changing this convention would require adding a preserve_range= keyword argument to almost all scikit-image functions, whose default value would change from False to True over 4 versions. Eventually, the change would be backwards-incompatible, no matter how gentle we made the deprecation curve.

Other major functions, such as skimage.measure.regionprops, could use an API tweak, for example by returning a dictionary mapping labels to properties, rather than a list.

Given the accumulation of potential API changes that have turned out to be too burdensome and noisy to fix with a standard deprecation cycle, principally because they involve changing function outputs for the same inputs, it makes sense to make all those changes in a transition to version 2.0.

Although semantic versioning [6] technically allows API changes with major version bumps, we must acknowledge that (1) an enormous number of projects depend on scikit-image and would thus be affected by backwards incompatible changes, and (2) it is not yet common practice in the scientific Python community to put upper version bounds on dependencies, so it is very unlikely that anyone used scikit-image<1.* or scikit-image<2.* in their dependency list. This implies that releasing a version 2.0 of scikit-image with breaking API changes would disrupt a large number of users. Additionally, such wide-sweeping changes would invalidate a large number of StackOverflow and other user guides. Finally, releasing a new version with a large number of changes prevents users from gradually migrating to the new API: an old code base must be migrated wholesale because it is impossible to depend on both versions of the API. This would represent an enormous barrier of entry for many users.

Given the above, this SKIP proposes that we release a new package where we can apply everything we have learned from over a decade of development, without disrupting our existing user base.

Detailed description

It is beyond the scope of this document to list all of the proposed API changes for skimage2, many of which have yet to be decided upon. Indeed, the scope and ambition of the 2.0 transition could grow if this SKIP is accepted. This SKIP instead proposes a mechanism for managing the transition without breaking users’ code. A meta-issue tracking the proposed changes can be found on GitHub, scikit-image/scikit-image#5439 [7]. Some examples are briefly included below for illustrative purposes:

  • Stop rescaling input arrays when the dtype must be coerced to float.

  • Stop swapping coordinate axis order in different contexts, such as drawing or warping.

  • Allow automatic return of non-NumPy types, so long as they are coercible to NumPy with numpy.asarray.

  • Harmonizing similar parameters in different functions to have the same name; for example, we currently have random_seed, random_state, seed, or sample_seed in different functions, all to mean the same thing.

  • Changing measure.regionprops to return a dictionary instead of a list.

  • Combine functions that have the same purpose, such as watershed, slic, or felzenschwalb, into a common namespace. This would make it easier for new users to find out which functions they should try out for a specific task. It would also help the community grow around common APIs, where now scikit-image APIs are essentially unique for each function.

To make this transition with a minimum amount of user disruption, this SKIP proposes releasing a new library, skimage2, that would replace the existing library, but only if users explicitly opt-in. Additionally, by releasing a new library, users could depend both on scikit-image (1.0) and on skimage2, allowing users to migrate their code progressively.

Implementation

The details of the proposal are as follows:

  • scikit-image 0.19 will be followed by scikit-image 1.0. Every deprecation message will be removed from 1.0, and the API will be considered the scikit-image 1.0 API.

  • After 1.0, the main branch will be changed to (a) change the import name to skimage2, (b) change the package name to skimage2, and (c) change the version number to 2.0-dev.

  • There will be no “scikit-image” package on PyPI with version 2.0. Users who pip install scikit-image will always get the 1.x version of the package. To install scikit-image 2.0, users will need to pip install skimage2, conda install skimage2, or similar.

  • After consensus has been reached on the new API, skimage2 will be released.

  • Following the release of skimage2, a release of scikit-image 1.1 is made. This release is identical to 1.0 (including bugfixes) but will advise users to either (a) upgrade to skimage2 or (b) pin the package to scikit-image<1.1 to avoid the warning.

  • scikit-image 1.0.x and 1.1.x will receive critical bug fixes for an unspecified period of time, depending on the severity of the bug and the amount of effort involved.

Backward compatibility

This proposal breaks backward compatibility in numerous places in the library. However, it does so in a new namespace, so that this proposal does not raise backward compatibility concerns for our users. That said, the authors will attempt to limit the number of backward incompatible changes to those likely to substantially improve the overall user experience. It is anticipated that porting skimage code to skimage2 will be a straightforward process and we will publish a user guide for making the transition by the time of the skimage2 release. Users will be notified about these resources - among other things - by a warning in scikit-image 1.1.

Alternatives

Releasing the new API in the same package using semantic versioning

This is SKIP-3, which was rejected after discussion with the community.

Continuous deprecation over multiple versions

This transition could occur gradually over many versions. For example, for functions automatically converting and rescaling float inputs, we could add a preserve_range keyword argument that would initially default to False, but the default value of False would be deprecated, with users getting a warning to switch to True. After the switch, we could (optionally) deprecate the argument, arriving, after a further two releases, at the same place: scikit-image no longer rescales data automatically, there are no unnecessary keyword arguments lingering all over the API.

Of course, this kind of operation would have to be done simultaneously over all of the above proposed changes.

Ultimately, the core team felt that this approach generates more work for both the scikit-image developers and the developers of downstream libraries, for dubious benefit: ultimately, later versions of scikit-image will still be incompatible with prior versions, although over a longer time scale.

A single package containing both versions

Since the import name is changing, it would be possible to make a single package with both the skimage and skimage2 namespaces shipping together, at least for some time. This option is attractive but it implies longer-term maintenance of the 1.0 namespace, for which we might lack maintainer time, or a long deprecation cycle for the 1.0 namespace, which would ultimately result in a lot of unhappy users getting deprecation messages from their scikit-image use.

Not making the proposed API changes

Another possibility is to reject backwards incompatible API changes outright, except in extreme cases. The core team feels that this is essentially equivalent to pinning the library at 0.19.

“scikit-image2” as the new package name

The authors acknowledge that the new names should be chosen with care to keep the disruption to scikit-image’s user base and community as small as possible. However, to protect users without upper version constraints from accidentally upgrading to the new API, the package name scikit-image must be changed. Changing the import name skimage is similarly advantageous because it allows using both APIs in the same environment.

This document suggests just skimage2 as the single new name for scikit-image’s API version 2.0, both for the import name and the name on PyPI, conda-forge and elsewhere. The following arguments were given in favor of this:

  • Only one new name is introduced with the project thereby keeping the number of associated names as low as possible.

  • With this change, the import and package name match.

  • Users might be confused whether they should install scikit-image2 or scikit-image-2. It was felt that skimage2 avoids this confusion.

  • Users who know what skimage is and see skimage2 in an install instruction somewhere, will likely be able to infer that it is a newer version of the package.

  • It is unlikely that users will be aware of the new API 2.0 but not of the new package name. A proposed release of scikit-image 1.1 might point users to skimage2 during the installation and update process and thereby clearly communicate the successors name.

The following arguments were made against naming the package skimage2:

  • According to the “Principle of least astonishment”, scikit-image2 might be considered the least surprising evolution of the package name.

  • It breaks with the convention that is followed by other scikits including scikit-image. (It was pointed out that this convention has not been true for some time and introducing a version number in the name is a precedent anyway.)

The earlier section “Related Work” describes how other projects dealt with similar problems.

Discussion

This SKIP is the result of discussion of SKIP-3. See the “Resolution” section of that document for further background on the motivation for this SKIP.

Resolution

References and Footnotes

All SKIPs should be declared as dedicated to the public domain with the CC0 license [1], as in Copyright, below, with attribution encouraged with CC0+BY [2].