Is this a consequence of the hierarchical order of operators in Python?
not(True)* True
#False
True* not(True)
#SyntaxError: invalid syntax
CodePudding user response:
Is this a consequence of the hierarchical order of operators in Python?
Yes (although the usual term is operator precedence). Summing up and simplifying:
not
isn't a function; it's an operator. Therefore, we don't need to write parentheses fornot (True)
, and in fact they don't do anything here. All that happens is that the parentheses are treated as ordinary grouping parentheses -(True)
is evaluated before anything else, becomingTrue
. So, let's consider the examples without the parentheses.not True * True
meansnot (True * True)
. It does not mean(not True) * True
, due to operator precedence. This is by design:>>> not 1 * 0 True >>> not (1 * 0) True >>> (not 1) * 0 0
It would, the developers figured, be unexpected to write something like
not 1 * 0
and get an integer result, and unexpected to writenot
in front of a mathematical operation and have thenot
only apply to the first thing in that expression.Because of that same operator precedence,
True * not True
is a syntax error. Python parses thenot
by itself as the right-hand side of the*
, because it hasn't had a chance to applynot True
yet.True * not
is obviously nonsense. Or, another way of looking at it: "not
followed by an expression" isn't in the list of "things that can be an operand for*
".This is perhaps surprising because the other commonly used unary operator,
-
(i.e., unary negation), doesn't have this issue. But that's because the precedence is the other way around: unary negation is processed before multiplication, not after.The same is true for
and
andor
combinations:>>> 3 * 5 and 1 # 3 * 5 is evaluated first 1 >>> 3 * (5 and 1) 3 >>> 3 or 1 * 5 # 1 * 5 is evaluated first, even though it comes later 3 >>> (3 or 1) * 5 15
CodePudding user response:
Anthony Sotille's answer says Python is trying to parse * not
as an "is not" or "not in" operator, but there's nothing in the parser or grammar that would try to do that. This is really a matter of precedence, and how precedence is implemented.
*
has a higher precedence than not
, and the way that works in the Python grammar is that there's a hierarchy of expression types, structured so that higher-precedence operators can be the "root" operator of an argument to a lower-precedence operator, but not the other way around.
For example, the grammar rule for multiplicative expressions is
term[expr_ty]:
| a=term '*' b=factor { _PyAST_BinOp(a, Mult, b, EXTRA) }
| a=term '/' b=factor { _PyAST_BinOp(a, Div, b, EXTRA) }
| a=term '//' b=factor { _PyAST_BinOp(a, FloorDiv, b, EXTRA) }
| a=term '%' b=factor { _PyAST_BinOp(a, Mod, b, EXTRA) }
| a=term '@' b=factor { CHECK_VERSION(expr_ty, 5, "The '@' operator is", _PyAST_BinOp(a, MatMult, b, EXTRA)) }
| factor
term
is the grammar rule for multiplicative expressions. The first 5 options in this rule consist of a multiplicative-precedence operator in the middle, another term
on the left of the operator, and a factor
on the right, where factor
is the rule for the next higher-precedence operator class. The 6th option is just a factor
.
Structuring the grammar like this makes sure the parsed syntax tree always matches the structure given by operator precedence, but it also means that lower-precedence operators cannot be the "root" operator of an argument to a higher-precedence operator, even when the expression would seem unambiguous. There's just no grammar rule that would allow a not
expression as an argument to a *
expression.
(The grammar rules for most expressions follow the above structure, but there are exceptions. For example, the grammar rules for parentheses don't follow the "no lower-precedence operators within higher-precedence expressions" structure, which is why you can write things like 3 * (4 5)
. Exponentiation is another exception - **
binds tighter than a unary
/-
/~
on the left, but not on the right, so the rules for **
and unary
/-
/~
don't follow a clear precedence hierarchy.)