PEP 661 – Sentinel Values

PEP: 661
Title: Sentinel Values
Author: Tal Einat <tal at python.org>
Discussions-To: https://discuss.python.org/t/pep-661-sentinel-values/9126
Status: Draft
Type: Standards Track
Created: 06-Jun-2021
Post-History: 06-Jun-2021

Abstract
Motivation
Rationale
Specification
Reference Implementation
Rejected Ideas
Additional Notes
References
Copyright

TL;DR: See the Specification and Reference Implementation.

Abstract

Unique placeholder values, commonly known as “sentinel values”, are useful in Python programs for several things, such as default values for function arguments where None is a valid input value. These cases are common enough for several idioms for implementing such “sentinels” to have arisen over the years, but uncommon enough that there hasn’t been a clear need for standardization. However, the common implementations, including some in the stdlib, suffer from several significant drawbacks.

This PEP suggests adding a utility for defining sentinel values, to be used in the stdlib and made publicly available as part of the stdlib.

Note: Changing all existing sentinels in the stdlib to be implemented this way is not deemed necessary, and whether to do so is left to the discretion of each maintainer.

Motivation

In May 2021, a question was brought up on the python-dev mailing list 1 about how to better implement a sentinel value for traceback.print_exception. The existing implementation used the following common idiom:

_sentinel = object()

However, this object has an uninformative and overly verbose repr, causing the function’s signature to be overly long and hard to read:

>>> help(traceback.print_exception)
Help on function print_exception in module traceback:

print_exception(exc, /, value=<object object at
0x000002825DF09650>, tb=<object object at 0x000002825DF09650>,
limit=None, file=None, chain=True)

Additionally, two other drawbacks of many existing sentinels were brought up in the discussion:

Not having a distinct type, hence it being impossible to define strict type signatures functions with sentinels as default values
Incorrect behavior after being copied or unpickled, due to a separate instance being created and thus comparisons using is failing

In the ensuing discussion, Victor Stinner supplied a list of currently used sentinel values in the Python standard library 2. This showed that the need for sentinels is fairly common, that there are various implementation methods used even within the stdlib, and that many of these suffer from at least one of the aforementioned drawbacks.

The discussion did not lead to any clear consensus on whether a standard implementation method is needed or desirable, whether the drawbacks mentioned are significant, nor which kind of implementation would be good.

A poll was created on discuss.python.org 3 to get a clearer sense of the community’s opinions. The poll’s results were not conclusive, with 40% voting for “The status-quo is fine / there’s no need for consistency in this”, but most voters voting for one or more standardized solutions. Specifically, 37% of the voters chose “Consistent use of a new, dedicated sentinel factory / class / meta-class, also made publicly available in the stdlib”.

With such mixed opinions, this PEP was created to facilitate making a decision on the subject.

Rationale

The criteria guiding the chosen implementation were:

The sentinel objects should behave as expected by a sentinel object: When compared using the is operator, it should always be considered identical to itself but never to any other object.
It should be simple to define as many distinct sentinel values as needed.
The sentinel objects should have a clear and short repr.
The sentinel objects should each have a distinct type, usable in type annotations to define strict type signatures.
The sentinel objects should behave correctly after copying and/or unpickling.
Creating a sentinel object should be a simple, straightforward one-liner.
Works using CPython and PyPy3. Will hopefully also work with other implementations.

After researching existing idioms and implementations, and going through many different possible implementations, an implementation was written which meets all of these criteria (see Reference Implementation).

Specification

A new sentinel function will be added to a new sentinels module. It will accept a single required argument, the name of the sentinel object, and a single optional argument, the repr of the object.

>>> NotGiven = sentinel('NotGiven')
>>> NotGiven
<NotGiven>
>>> MISSING = sentinel('MISSING', repr='mymodule.MISSING')
>>> MISSING
mymodule.MISSING

Checking if a value is such a sentinel should be done using the is operator, as is recommended for None. Equality checks using == will also work as expected, returning True only when the object is compared with itself.

The name should be set to the name of the variable used to reference the object, as in the examples above. Otherwise, the sentinel object won’t be able to survive copying or pickling+unpickling while retaining the above described behavior. Note, that when defined in a class scope, the name must be the fully-qualified name of the variable in the module, for example:

class MyClass:
    NotGiven = sentinel('MyClass.NotGiven')

Type annotations for sentinel values will use typing.Literal 9. For example:

def foo(value: int | Literal[NotGiven]) -> None:
    ...

Reference Implementation

The reference implementation is found in a dedicated GitHub repo 4. A simplified version follows:

def sentinel(name, repr=None):
    """Create a unique sentinel object."""
    repr = repr or f'<{name}>'

    module = _get_parent_frame().f_globals.get('__name__', '__main__')
    class_name = _get_class_name(name, module)
    class_namespace = {
        '__repr__': lambda self: repr,
    }
    cls = type(class_name, (), class_namespace)
    cls.__module__ = module
    _get_parent_frame().f_globals[class_name] = cls

    sentinel = cls()
    cls.__new__ = lambda cls_: sentinel

    return sentinel

def _get_class_name(sentinel_qualname, module_name):
    return '__'.join(['_sentinel_type',
                      module_name.replace('.', '_'),
                      sentinel_qualname.replace('.', '_')])

Note that a dedicated class is created automatically for each sentinel object. This class is assigned to the namespace of the module from which the sentinel() call was made, or to that of the sentinels module itself as a fallback. These classes have long names comprised of several parts to ensure their uniqueness. However, these names usually wouldn’t be used, since type annotations should use Literal[] as described above, and identity checks should be preferred over type checks.

Rejected Ideas

Use `NotGiven = object()`

This suffers from all of the drawbacks mentioned in the Rationale section.

Add a single new sentinel value, e.g. `MISSING` or `Sentinel`

Since such a value could be used for various things in various places, one could not always be confident that it would never be a valid value in some use cases. On the other hand, a dedicated and distinct sentinel value can be used with confidence without needing to consider potential edge-cases.

Additionally, it is useful to be able to provide a meaningful name and repr for a sentinel value, specific to the context where it is used.

Finally, this was a very unpopular option in the poll 3, with only 12% of the votes voting for it.

Use the existing `Ellipsis` sentinel value

This is not the original intended use of Ellipsis, though it has become increasingly common to use it to define empty class or function blocks instead of using pass.

Also, similar to a potential new single sentinel value, Ellipsis can’t be as confidently used in all cases, unlike a dedicated, distinct value.

Use a single-valued enum

The suggested idiom is:

class NotGivenType(Enum):
    NotGiven = 'NotGiven'
NotGiven = NotGivenType.NotGiven

Besides the excessive repetition, the repr is overly long: <NotGivenType.NotGiven: 'NotGiven'>. A shorter repr can be defined, at the expense of a bit more code and yet more repetition.

Finally, this option was the least popular among the nine options in the poll 3, being the only option to receive no votes.

A sentinel class decorator

The suggested interface:

@sentinel(repr='<NotGiven>')
class NotGivenType: pass
NotGiven = NotGivenType()

While this allowed for a very simple and clear implementation, the interface is too verbose, repetitive, and difficult to remember.

Using class objects

Since classes are inherently singletons, using a class as a sentinel value makes sense and allows for a simple implementation.

The simplest version of this idiom is:

class NotGiven: pass

To have a clear repr, one could define __repr__:

class NotGiven:
    def __repr__(self):
        return '<NotGiven>'

… or use a meta-class:

class NotGiven(metaclass=SentinelMeta): pass

However, all such implementations don’t have a dedicated type for the sentinel, which is considered desirable for strict typing. A dedicated type could be created by a meta-class or class decorator, but at that point the implementation would become much more complex and loses its advantages over the chosen implementation.

Additionally, using classes this way is unusual and could be confusing.

Define a recommended “standard” idiom, without supplying an implementation

Most common exiting idioms have significant drawbacks. So far, no idiom has been found that is clear and concise while avoiding these drawbacks.

Also, in the poll on this subject 3, the options for recommending an idiom were unpopular, with the highest-voted option being voted for by only 25% of the voters.

Additional Notes

This PEP and the initial implementation are drafted in a dedicated GitHub repo 4.
The support for copying/unpickling works when defined in a module’s scope or a (possibly nested) class’s scope. Note that in the latter case, the name provided as the first parameter must be the fully-qualified name of the variable in the module:
```
class MyClass:
    NotGiven = sentinel('MyClass.NotGiven', repr='<NotGiven>')
```