[red-knot] - Flow-control for boolean operations (#13940)

## Summary

As python uses short-circuiting boolean operations in runtime, we should
mimic that logic in redknot as well.
For example, we should detect that in the following code `x` might be
undefined inside the block:

```py
if flag or (x := 1):
    print(x) 
```

## Test Plan

Added mdtest suit for boolean expressions.

---------

Co-authored-by: Carl Meyer <carl@astral.sh>
This commit is contained in:
TomerBin 2024-10-27 05:33:01 +02:00 committed by GitHub
parent b6ffa51c16
commit 66c3aaa307
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
3 changed files with 125 additions and 1 deletions

View file

@ -0,0 +1,78 @@
# Short-Circuit Evaluation
## Not all boolean expressions must be evaluated
In `or` expressions, if the left-hand side is truthy, the right-hand side is not evaluated.
Similarly, in `and` expressions, if the left-hand side is falsy, the right-hand side is not
evaluated.
```py
def bool_instance() -> bool:
return True
if bool_instance() or (x := 1):
# error: [possibly-unresolved-reference]
reveal_type(x) # revealed: Unbound | Literal[1]
if bool_instance() and (x := 1):
# error: [possibly-unresolved-reference]
reveal_type(x) # revealed: Unbound | Literal[1]
```
## First expression is always evaluated
```py
def bool_instance() -> bool:
return True
if (x := 1) or bool_instance():
reveal_type(x) # revealed: Literal[1]
if (x := 1) and bool_instance():
reveal_type(x) # revealed: Literal[1]
```
## Statically known truthiness
```py
if True or (x := 1):
# TODO: infer that the second arm is never executed so type should be just "Unbound".
# error: [possibly-unresolved-reference]
reveal_type(x) # revealed: Unbound | Literal[1]
if True and (x := 1):
# TODO: infer that the second arm is always executed so type should be just "Literal[1]".
# error: [possibly-unresolved-reference]
reveal_type(x) # revealed: Unbound | Literal[1]
```
## Later expressions can always use variables from earlier expressions
```py
def bool_instance() -> bool:
return True
bool_instance() or (x := 1) or reveal_type(x) # revealed: Literal[1]
# error: [unresolved-reference]
bool_instance() or reveal_type(y) or (y := 1) # revealed: Unbound
```
## Nested expressions
```py
def bool_instance() -> bool:
return True
if bool_instance() or ((x := 1) and bool_instance()):
# error: "Name `x` used when possibly not defined"
reveal_type(x) # revealed: Unbound | Literal[1]
if ((y := 1) and bool_instance()) or bool_instance():
reveal_type(y) # revealed: Literal[1]
# error: [possibly-unresolved-reference]
if (bool_instance() and (z := 1)) or reveal_type(z): # revealed: Unbound | Literal[1]
# error: [possibly-unresolved-reference]
reveal_type(z) # revealed: Unbound | Literal[1]
```

View file

@ -1106,6 +1106,27 @@ where
}, },
); );
} }
ast::Expr::BoolOp(ast::ExprBoolOp {
values,
range: _,
op: _,
}) => {
// TODO detect statically known truthy or falsy values (via type inference, not naive
// AST inspection, so we can't simplify here, need to record test expression for
// later checking)
let mut snapshots = vec![];
for (index, value) in values.iter().enumerate() {
// The first item of BoolOp is always evaluated
if index > 0 {
snapshots.push(self.flow_snapshot());
}
self.visit_expr(value);
}
for snapshot in snapshots {
self.flow_merge(snapshot);
}
}
_ => { _ => {
walk_expr(self, expr); walk_expr(self, expr);
} }

View file

@ -148,7 +148,7 @@ static HEADER_RE: LazyLock<Regex> =
/// Matches a code block fenced by triple backticks, possibly with language and `key=val` /// Matches a code block fenced by triple backticks, possibly with language and `key=val`
/// configuration items following the opening backticks (in the "tag string" of the code block). /// configuration items following the opening backticks (in the "tag string" of the code block).
static CODE_RE: LazyLock<Regex> = LazyLock::new(|| { static CODE_RE: LazyLock<Regex> = LazyLock::new(|| {
Regex::new(r"^```(?<lang>(?-u:\w)+)?(?<config>(?: +\S+)*)\s*\n(?<code>(?:.|\n)*?)\n?```\s*\n") Regex::new(r"^```(?<lang>(?-u:\w)+)?(?<config>(?: +\S+)*)\s*\n(?<code>(?:.|\n)*?)\n?```\s*\n?")
.unwrap() .unwrap()
}); });
@ -421,6 +421,31 @@ mod tests {
assert_eq!(file.code, "x = 1"); assert_eq!(file.code, "x = 1");
} }
#[test]
fn no_new_line_at_eof() {
let source = dedent(
"
```py
x = 1
```",
);
let mf = super::parse("file.md", &source).unwrap();
let [test] = &mf.tests().collect::<Vec<_>>()[..] else {
panic!("expected one test");
};
assert_eq!(test.name(), "file.md");
let [file] = test.files().collect::<Vec<_>>()[..] else {
panic!("expected one file");
};
assert_eq!(file.path, "test.py");
assert_eq!(file.lang, "py");
assert_eq!(file.code, "x = 1");
}
#[test] #[test]
fn multiple_tests() { fn multiple_tests() {
let source = dedent( let source = dedent(