python-peps/pep-0667.rst

PEP: 667
Title: Consistent views of namespaces
Author: Mark Shannon <mark@hotpy.org>
Status: Draft
Type: Standards Track
Content-Type: text/x-rst
Created: 30-Jul-2021
Post-History: 20-Aug-2021


Abstract
========

In early versions of Python all namespaces, whether in functions,
classes or modules, were all implemented the same way: as a dictionary.

For performance reasons, the implementation of function namespaces was
changed. Unfortunately this meant that accessing these namespaces through
``locals()`` and ``frame.f_locals`` ceased to be consistent and some
odd bugs crept in over the years as threads, generators and coroutines
were added.

This PEP proposes making these namespaces consistent once more.
Modifications to ``frame.f_locals`` will always be visible in
the underlying variables. Modifications to local variables will
immediately be visible in ``frame.f_locals``, and they will be
consistent regardless of threading or coroutines.

The ``locals()`` function will act the same as it does now for class
and modules scopes. For function scopes it will return an instantaneous
snapshot of the underlying ``frame.f_locals``.

Motivation
==========

The current implementation of ``locals()`` and ``frame.f_locals`` is slow,
inconsistent and buggy.
We want to make it faster, consistent, and most importantly fix the bugs.

For example::

    class C:
        x = 1
        sys._getframe().f_locals['x'] = 2
        print(x)

prints ``2``

but::

    def f():
        x = 1
        sys._getframe().f_locals['x'] = 2
        print(x)
    f()

prints ``1``

This is inconsistent, and confusing.
With this PEP both examples would print ``2``.

Worse than that, the current behavior can result in strange bugs [1]_

There are no compensating advantages for the current behavior;
it is unreliable and slow.

Rationale
=========

The current implementation of ``frame.f_locals`` returns a dictionary
that is created on the fly from the array of local variables.
This can result in the array and dictionary getting out of sync with
each other. Writes to the ``f_locals`` may not show up as
modifications to local variables. Writes to local variables can get lost.

By making ``frame.f_locals`` return a view on the
underlying frame, these problems go away. ``frame.f_locals`` is always in
sync with the frame because it is a view of it, not a copy of it.

Specification
=============

Python
------

``frame.f_locals`` will return a view object on the frame that
implements the ``collections.abc.Mapping`` interface.

For module and class scopes ``frame.f_locals`` will be a dictionary,
for function scopes it will be a custom class.

``locals()`` will be defined as::

    def locals():
        frame = sys._getframe(1)
        f_locals = frame.f_locals
        if frame.is_function():
            f_locals = dict(f_locals)
        return f_locals

All writes to the ``f_locals`` mapping will be immediately visible
in the underlying variables. All changes to the underlying variables
will be immediately visible in the mapping. The ``f_locals`` object will
be a full mapping, and can have arbitrary key-value pairs added to it.

For example::

    def l():
        "Get the locals of caller"
        return sys._getframe(1).f_locals

    def test():
        if 0: y = 1 # Make 'y' a local variable
        x = 1
        l()['x'] = 2
        l()['y'] = 4
        l()['z'] = 5
        y
        print(locals(), x)

``test()`` will print ``{'x': 2, 'y': 4, 'z': 5} 2``

In Python 3.10, the above will fail with a ``NameError``,
as the definition of ``y`` by ``l()['y'] = 4`` is lost.

C-API
-----

Extensions to the API
'''''''''''''''''''''

Two new C-API functions will be added::

    PyObject *PyEval_Locals(void)
    PyObject *PyFrame_GetLocals(PyFrameObject *f)

``PyEval_Locals()`` is equivalent to: ``locals()``.

``PyFrame_GetLocals(f)`` is equivalent to: ``f.f_locals``.

Both functions will return a new reference.

Changes to existing APIs
''''''''''''''''''''''''

The C-API function ``PyEval_GetLocals()`` will be deprecated.
``PyEval_Locals()`` should be used instead.

The following three functions will become no-ops, and will be deprecated::

    PyFrame_FastToLocalsWithError()
    PyFrame_FastToLocals()
    PyFrame_LocalsToFast()

The above four deprecated functions will be removed in 3.13.

Behavior of f_locals for optimized functions
--------------------------------------------

Although ``f.f_locals`` behaves as if it were the namespace of the function,
there will be some observable differences.
For example, ``f.f_locals is f.f_locals`` may be ``False``.

However ``f.f_locals == f.f_locals`` will be ``True``, and
all changes to the underlying variables, by any means, will be
always be visible.

Backwards Compatibility
=======================

Python
------

The current implementation has many corner cases and oddities.
Code that works around those may need to be changed.
Code that uses ``locals()`` for simple templating, or print debugging,
will continue to work correctly. Debuggers and other tools that use
``f_locals`` to modify local variables, will now work correctly,
even in the presence of threaded code, coroutines and generators.

C-API
-----

PyEval_GetLocals
''''''''''''''''

Because ``PyEval_GetLocals()`` returns a borrowed reference, it requires
the dictionary to be cached on the frame, extending its lifetime and
forces memory to be allocated for the frame object on the heap as well.

Using ``PyEval_Locals()`` will be much more efficient
than ``PyEval_GetLocals()``.

This code::

    locals = PyEval_GetLocals();
    if (locals == NULL) {
        goto error_handler;
    }
    Py_INCREF(locals);

should be replaced with::

    locals = PyEval_Locals();
    if (locals == NULL) {
        goto error_handler;
    }

PyFrame_FastToLocals, etc.
''''''''''''''''''''''''''

These functions were designed to convert the internal "fast" representation 
of the locals variables of a function to a dictionary, and vice versa.

Calls to them are no longer required. C code that directly accesses the
``f_locals`` field of a frame should be modified to call
``PyFrame_GetLocals()`` instead::

    PyFrame_FastToLocals(frame);
    PyObject *locals = frame.f_locals;
    Py_INCREF(locals);

becomes::

    PyObject *locals = PyFrame_GetLocals(frame);
    if (frame == NULL)
        goto error_handler;

Implementation
==============

Each read of ``frame.f_locals`` will create a new proxy object that gives
the appearance of being the mapping of local (including cell and free)
variable names to the values of those local variables.

A possible implementation is sketched out below.
All attributes that start with an underscore are invisible and
cannot be accessed directly.
They serve only to illustrate the proposed design.

::

    NULL: Object # NULL is a singleton representing the absence of a value.

    class CodeType:

        _name_to_offset_mapping_impl: dict | NULL
        _cells: frozenset # Set of indexes of cell and free variables
        ...

        def __init__(self, ...):
            self._name_to_offset_mapping_impl = NULL
            self._variable_names = deduplicate(
                self.co_varnames + self.co_cellvars + self.co_freevars
            )
            ...

        @property
        def _name_to_offset_mapping(self):
            "Mapping of names to offsets in local variable array."
            if self._name_to_offset_mapping_impl is NULL:
                self._name_to_offset_mapping_impl = {
                    name: index for (index, name) in enumerate(self._variable_names)
                }
            return self._name_to_offset_mapping_impl

    class FrameType:

        _locals : array[Object] # The values of the local variables, items may be NULL.
        _extra_locals: dict | NULL # Dictionary for storing extra locals not in _locals.

        def __init__(self, ...):
            self._extra_locals = NULL
            ...

        @property
        def f_locals(self):
            return FrameLocalsProxy(self)

    class FrameLocalsProxy:
        "Implements collections.MutableMapping."

        __slots__ "_frame"

        def __init__(self, frame:FrameType):
            self._frame = frame

        def __getitem__(self, name):
            f = self._frame
            co = f.f_code
            if name in co._name_to_offset_mapping:
                index = co._name_to_offset_mapping[name]
                val = f._locals[index]
                if val is NULL:
                    raise KeyError(name)
                if index in co._cells
                    val = val.cell_contents
                    if val is NULL:
                        raise KeyError(name)
                return val
            else:
                if f._extra_locals is NULL:
                    raise KeyError(name)
                return f._extra_locals[name]

        def __setitem__(self, name, value):
            f = self._frame
            co = f.f_code
            if name in co._name_to_offset_mapping:
                index = co._name_to_offset_mapping[name]
                kind = co._local_kinds[index]
                if index in co._cells
                    cell = f._locals[index]
                    cell.cell_contents = val
                else:
                    f._locals[index] = val
            else:
                if f._extra_locals is NULL:
                    f._extra_locals = {}
                f._extra_locals[name] = val

        def __iter__(self):
            f = self._frame
            co = f.f_code
            yield from iter(f._extra_locals)
            for index, name in enumerate(co._variable_names):
                val = f._locals[index]
                if val is NULL:
                    continue
                if index in co._cells:
                    val = val.cell_contents
                    if val is NULL:
                        continue
                yield name

        def __contains__(self, item):
            f = self._frame
            if item in f._extra_locals:
                return True
            return item in co._variable_names

        def __len__(self):
            f = self._frame
            co = f.f_code
            res = 0
            for index, _ in enumerate(co._variable_names):
                val = f._locals[index]
                if val is NULL:
                    continue
                if index in co._cells:
                    if val.cell_contents is NULL:
                        continue
                res += 1
            return len(self._extra_locals) + res

C API
-----

``PyEval_GetLocals()`` will be implemented roughly as follows::

    PyObject *PyEval_GetLocals(void) {
        PyFrameObject * = ...; // Get the current frame.
        Py_CLEAR(frame->_locals_cache);
        frame->_locals_cache = PyEval_Locals();
        return frame->_locals_cache;
    }

As with all functions that return a borrowed reference, care must be taken to
ensure that the reference is not used beyond the lifetime of the object.

Comparison with PEP 558
=======================

This PEP and :pep:`558` share a common goal:
to make the semantics of  ``locals()`` and ``frame.f_locals()``
intelligible, and their operation reliable.


The key difference between this PEP and :pep:`558` is that
:pep:`558` keeps an internal copy of the local variables,
whereas this PEP does not.

:pep:`558` does not specify exactly when the internal copy is
updated, making the behavior of :pep:`558` impossible to reason about.


Open Issues
===========

An alternative way to define ``locals()`` would be simply as::

    def locals():
        return sys._getframe(1).f_locals

This would be simpler and easier to understand. However,
there would be backwards compatibility issues when ``locals`` is assigned
to a local variable or when passed to ``eval`` or ``exec``.

References
==========

.. [1] https://bugs.python.org/issue30744

Copyright
=========

This document is placed in the public domain or under the
CC0-1.0-Universal license, whichever is more permissive.

..
    Local Variables:
    mode: indented-text
    indent-tabs-mode: nil
    sentence-end-double-space: t
    fill-column: 70
    coding: utf-8
    End: