Abstract

This PEP describes the procedure for changes to Python that are backwards-incompatible between the Python 2.X series and Python 3000. All such changes must be documented by an appropriate Python 3000 PEP and must be accompanied by code that can identify when pieces of Python 2.X code may be problematic in Python 3000.

Rationale

Python 3000 will introduce a number of backwards-incompatible changes to Python, mainly to streamline the language and to remove some previous design mistakes. But Python 3000 is not intended to be a new and completely different language from the Python 2.X series, and it is expected that much of the Python user community will make the transition to Python 3000 when it becomes available.

To encourage this transition, it is crucial to provide a clear and complete guide on how to upgrade Python 2.X code to Python 3000 code. Thus, for any backwards-incompatible change, two things are required:

• An official Python Enhancement Proposal (PEP)
• Code that can identify pieces of Python 2.X code that may be problematic in Python 3000

Python Enhancement Proposals

Every backwards-incompatible change must be accompanied by a PEP. This PEP should follow the usual PEP guidelines and explain the purpose and reasoning behind the backwards incompatible change. In addition to the usual PEP sections, all PEPs proposing backwards-incompatible changes must include an additional section: Compatibility Issues. This section should describe what is backwards incompatible about the proposed change to Python, and the major sorts of breakage to be expected.

While PEPs must still be evaluated on a case-by-case basis, a PEP may be inappropriate for Python 3000 if its Compatibility Issues section implies any of the following:

• Most or all instances of a Python 2.X construct are incorrect in Python 3000, and most or all instances of the Python 3000 construct are incorrect in Python 2.X.

  So for example, changing the meaning of the for-loop else-clause from “executed when the loop was not broken out of” to “executed when the loop had zero iterations” would mean that all Python 2.X for-loop else-clauses would be broken, and there would be no way to use a for-loop else-clause in a Python-3000-appropriate manner. Thus a PEP for such an idea would likely be rejected.

• Many instances of a Python 2.X construct are incorrect in Python 3000 and the PEP fails to demonstrate real-world use-cases for the changes.

  Backwards incompatible changes are allowed in Python 3000, but not to excess. A PEP that proposes backwards-incompatible changes should provide good examples of code that visibly benefits from the changes.

PEP-writing is time-consuming, so when a number of backwards-incompatible changes are closely related, they should be proposed in the same PEP. Such PEPs will likely have longer Compatibility Issues sections, however, since they must now describe the sorts of breakage expected from all the proposed changes.

Identifying Problematic Code

In addition to the PEP requirement, backwards incompatible changes to Python must also be accompanied by code to issue warnings for pieces of Python 2.X code that will behave differently in Python 3000. Such warnings will be enabled in Python 2.X using a new command-line switch: -3. All backwards incompatible changes should be accompanied by a patch for Python 2.X that, when -3 is specified, issues warnings for each construct that is being changed.

For example, if dict.keys() returns an iterator in Python 3000, the patch to the Python 2.X branch should do something like:

If -3 was specified, change dict.keys() to return a subclass of list that issues warnings whenever you use any methods other than __iter__().

Such a patch would mean that warnings are only issued when features that will not be present in Python 3000 are used, and almost all existing code should continue to work. (Code that relies on dict.keys() always returning a list and not a subclass should be pretty much non-existent.)

References

TBD

Copyright

This document has been placed in the public domain.