PEP 379 – Adding an Assignment Expression
- Jervis Whitley <jervisau at gmail.com>
- Standards Track
- 2.7, 3.2
This PEP adds a new assignment expression to the Python language to make it possible to assign the result of an expression in almost any place. The new expression will allow the assignment of the result of an expression at first use (in a comparison for example).
Motivation and Summary
Issue1714448 “if something as x:”  describes a feature to allow
assignment of the result of an expression in an if statement to a
name. It supposed that the
as syntax could be borrowed for this
purpose. Many times it is not the expression itself that is
interesting, rather one of the terms that make up the
expression. To be clear, something like this:
if (f_result() == [1, 2, 3]) as res:
seems awfully limited, when this:
if (f_result() as res) == [1, 2, 3]:
is probably the desired result.
See the Examples section near the end.
A new expression is proposed with the (nominal) syntax:
EXPR -> VAR
This single expression does the following:
- Evaluate the value of
EXPR, an arbitrary expression;
- Assign the result to
VAR, a single assignment target; and
- Leave the result of
EXPRon the Top of Stack (TOS)
-> or (
RARROW) has been used to illustrate the concept that
the result of
EXPR is assigned to
The translation of the proposed syntax is:
VAR = (EXPR) (EXPR)
The assignment target can be either an attribute, a subscript or name:
f() -> name # where 'name' exists previously. f() -> name.attr # again 'name' exists prior to this expression. f() -> name
This expression should be available anywhere that an expression is currently accepted.
All exceptions that are currently raised during invalid
assignments will continue to be raised when using the assignment
expression. For example, a
NameError will be raised when in
example 1 and 2 above if
name is not previously defined, or an
IndexError if index 0 was out of range.
Examples from the Standard Library
The following two examples were chosen after a brief search through the standard library, specifically both are from ast.py which happened to be open at the time of the search.
def walk(node): from collections import deque todo = deque([node]) while todo: node = todo.popleft() todo.extend(iter_child_nodes(node)) yield node
Using assignment expression:
def walk(node): from collections import deque todo = deque([node]) while todo: todo.extend(iter_child_nodes(todo.popleft() -> node)) yield node
def get_docstring(node, clean=True): if not isinstance(node, (FunctionDef, ClassDef, Module)): raise TypeError("%r can't have docstrings" % node.__class__.__name__) if node.body and isinstance(node.body, Expr) and \ isinstance(node.body.value, Str): if clean: import inspect return inspect.cleandoc(node.body.value.s) return node.body.value.s
Using assignment expression:
def get_docstring(node, clean=True): if not isinstance(node, (FunctionDef, ClassDef, Module)): raise TypeError("%r can't have docstrings" % node.__class__.__name__) if node.body -> body and isinstance(body -> elem, Expr) and \ isinstance(elem.value -> value, Str): if clean: import inspect return inspect.cleandoc(value.s) return value.s
The examples shown below highlight some of the desirable features of the assignment expression, and some of the possible corner cases.
- Assignment in an if statement for use later:
def expensive(): import time; time.sleep(1) return 'spam' if expensive() -> res in ('spam', 'eggs'): dosomething(res)
- Assignment in a while loop clause:
while len(expensive() -> res) == 4: dosomething(res)
- Keep the iterator object from the for loop:
for ch in expensive() -> res: sell_on_internet(res)
- Corner case:
for ch -> please_dont in expensive(): pass # who would want to do this? Not I.
This document has been placed in the public domain.
Last modified: 2017-11-11 19:28:55 GMT