PEP 661 -- Sentinel Values

[taleinat]

Please consider PEP 661 -- Sentinel Values
https://peps.python.org/pep-0661/

• The PEP has been discussed in threads listed in its Post-History header
• The PEP was announced on Discuss (link in Post-History)
• The PEP includes all relevant Suggested Sections
• The PEP includes endorsements from the projects/groups/people it helps
• The PEP has a CODEOWNERS entry

[JelleZijlstra]

Since this PEP proposes a change to the type system (an extension of the values allowed in Literal[]), the Typing Council is also looking at it. We'll discuss internally and share our opinion here if the Steering Council thinks it's helpful.

[encukou]

Note that there is some more discussion on Discourse, and the PEP might need to be changed.

[JelleZijlstra]

On behalf of the Typing Council:

The PEP currently proposes to use Literal[SentinelName] as the type annotation to describe the type of a sentinel object. We prefer to support SentinelName directly without the need to wrap it in a Literal . This would require the sentinel base class to supply a __or__ method to support unions, which would be straightforward. The most common sentinel value used in Python is None, and it is special-cased in the type system to not require Literal[None] in a type annotation. Since type checkers (both static and runtime) will already need to special-case sentinels, we wouldn’t gain any simplicity by forcing users to wrap sentinels in Literal[].

A suggestion came up to allow both forms in type annotations, Literal[SentinelName] and SentinelName. We prefer to keep only the latter so we don’t introduce redundant ways to describe the same type.

(Note from me personally: The change we're recommending here aligns with something Tal already said he was planning to change in the PEP, leading to Petr's message above.)