Skip to content

Release consumed itertools.tee buffers #9322

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
youknowone merged 1 commit into from
Oct 23, 2026
Merged

Release consumed itertools.tee buffers #9322

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
102 changes: 69 additions & 33 deletions crates/vm/src/stdlib/itertools.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1025,55 +1025,38 @@ mod decl {
struct PyItertoolsTeeData {
iterable: PyIter,
values: PyMutex<Vec<PyObjectRef>>,
next: PyMutex<Option<PyRef<Self>>>,
#[pytraverse(skip)]
running: AtomicBool,
}

#[pyclass(flags(DISALLOW_INSTANTIATION))]
impl PyItertoolsTeeData {
const BLOCK_SIZE: usize = 64;

fn new(iterable: PyIter, vm: &VirtualMachine) -> PyRef<Self> {
Self {
iterable,
values: PyMutex::new(vec![]),
values: PyMutex::new(Vec::with_capacity(Self::BLOCK_SIZE)),
next: PyMutex::new(None),
running: AtomicBool::new(false),
}
.into_ref(&vm.ctx)
}

fn get_item(&self, vm: &VirtualMachine, index: usize) -> PyResult<PyIterReturn> {
// Return cached value if available
{
let Some(values) = self.values.try_lock() else {
return Err(vm.new_runtime_error("cannot re-enter the tee iterator"));
};
if index < values.len() {
return Ok(PyIterReturn::Return(values[index].clone()));
}
}
// Prevent concurrent/reentrant calls to iterable.next(). The claim
// covers caching the value as well: released any earlier, a second
// tee at the same index fetches a value of its own and one of the
// two is dropped without ever reaching a caller.
if self.running.swap(true, Ordering::Acquire) {
return Err(vm.new_runtime_error("cannot re-enter the tee iterator"));
}
scopeguard::defer! { self.running.store(false, Ordering::Release) }
let obj = raise_if_stop!(self.iterable.next(vm)?);
let Some(mut values) = self.values.try_lock() else {
return Err(vm.new_runtime_error("cannot re-enter the tee iterator"));
};
if values.len() == index {
values.push(obj);
}
Ok(PyIterReturn::Return(values[index].clone()))
fn next_block(&self, vm: &VirtualMachine) -> PyRef<Self> {
self.next
.lock()
.get_or_insert_with(|| Self::new(self.iterable.clone(), vm))
.clone()
}
}

#[pyattr]
#[pyclass(name = "_tee", traverse)]
#[derive(Debug, PyPayload)]
struct PyItertoolsTee {
tee_data: PyRef<PyItertoolsTeeData>,
tee_data: PyMutex<PyRef<PyItertoolsTeeData>>,
#[pytraverse(skip)]
index: AtomicCell<usize>,
#[pytraverse(skip)]
Expand All @@ -1091,7 +1074,7 @@ mod decl {
return Ok(tee.__copy__());
}
Ok(Self {
tee_data: PyItertoolsTeeData::new(iterator, vm),
tee_data: PyMutex::new(PyItertoolsTeeData::new(iterator, vm)),
index: AtomicCell::new(0),
advancing: AtomicBool::new(false),
})
Expand All @@ -1105,7 +1088,7 @@ mod decl {
return vm.call_special_method(&iterator, identifier!(vm, __copy__), ());
}
Ok(Self {
tee_data: PyItertoolsTeeData::new(iterator, vm),
tee_data: PyMutex::new(PyItertoolsTeeData::new(iterator, vm)),
index: AtomicCell::new(0),
advancing: AtomicBool::new(false),
}
Expand All @@ -1118,8 +1101,11 @@ mod decl {
impl Py<PyItertoolsTee> {
#[pymethod]
fn __copy__(&self) -> PyItertoolsTee {
// A block transition resets the index under this lock, so the pair
// is a consistent snapshot even while another thread advances us.
let data = self.tee_data.lock();
PyItertoolsTee {
tee_data: self.tee_data.clone(),
tee_data: PyMutex::new(data.clone()),
index: AtomicCell::new(self.index.load()),
advancing: AtomicBool::new(false),
}
Expand Down Expand Up @@ -1162,15 +1148,65 @@ mod decl {
impl SelfIter for PyItertoolsTee {}
impl IterNext for PyItertoolsTee {
fn next(zelf: &Py<Self>, vm: &VirtualMachine) -> PyResult<PyIterReturn> {
let mut current = zelf.tee_data.lock();
// Reading the index and moving it on is one step: two callers that
// read the same index hand out the same value twice and leave the
// buffer to be filled out of order.
if zelf.advancing.swap(true, Ordering::Acquire) {
// The block lock serializes claims; an atomic exchange is not needed.
if zelf.advancing.load(Ordering::Acquire) {
drop(current);
return Err(vm.new_runtime_error("cannot re-enter the tee iterator"));
}
zelf.advancing.store(true, Ordering::Relaxed);
scopeguard::defer! { zelf.advancing.store(false, Ordering::Release) }
if zelf.index.load() == PyItertoolsTeeData::BLOCK_SIZE {
let data = current.clone();
drop(current);
let next = data.next_block(vm);
current = zelf.tee_data.lock();
let previous = core::mem::replace(&mut *current, next);
zelf.index.store(0);
drop(current);
// Releasing a consumed block may run item finalizers. No cursor
// or buffer lock may be held when the last owner lets it go.
drop(data);
drop(previous);
current = zelf.tee_data.lock();
}

// Cached reads neither call Python nor release the block, so they
// can update the cursor without cloning its data reference.
let index = zelf.index.load();
let value = raise_if_stop!(zelf.tee_data.get_item(vm, index)?);
let cached = {
let values = current.values.lock();
if let Some(value) = values.get(index) {
Some(value.clone())
} else {
// Claim the source while the cache is locked: another tee
// cannot fill this slot between the check and the claim.
if current.running.load(Ordering::Acquire) {
drop(values);
drop(current);
return Err(vm.new_runtime_error("cannot re-enter the tee iterator"));
}
current.running.store(true, Ordering::Relaxed);
None
}
};
if let Some(value) = cached {
zelf.index.store(index + 1);
return Ok(PyIterReturn::Return(value));
}

let data = current.clone();
drop(current);
// Prevent concurrent/reentrant calls to iterable.next(). The claim
// covers caching the value as well: released any earlier, a second
// tee at the same index fetches a value of its own and one of the
// two is dropped without ever reaching a caller.
scopeguard::defer! { data.running.store(false, Ordering::Release) }
let value = raise_if_stop!(data.iterable.next(vm)?);
data.values.lock().push(value.clone());
zelf.index.store(index + 1);
Ok(PyIterReturn::Return(value))
}
Expand Down
19 changes: 19 additions & 0 deletions extra_tests/snippets/stdlib_itertools.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -679,3 +679,22 @@ def __init__(self, iterable, *, label):
pairs = PairwiseWithLabel([1, 2], label="pairs")
assert pairs.label == "pairs"
assert next(pairs) == (1, 2)


# Consumed tee values must not remain alive for the lifetime of the iterators.
import weakref


class TeeItem:
pass


tee_left, tee_right = itertools.tee(TeeItem() for _ in range(200))
tee_first = next(tee_left)
tee_ref = weakref.ref(tee_first)
del tee_first
next(tee_right)
for _ in range(199):
next(tee_left)
next(tee_right)
assert tee_ref() is None
Loading