Rollup merge of #72414 - KodrAus:feat/stdlazy, r=Mark-Simulacrum

 Add lazy initialization primitives to std

Follow-up to #68198

Current RFC: rust-lang/rfcs#2788

Rebased and fixed up a few of the dangling comments. Some notes carried over from the previous PR:

- [ ] Naming. I'm ok to just roll with the `Sync` prefix like `SyncLazy` for now, but [have a personal preference for `Atomic`](rust-lang/rfcs#2788 (comment)) like `AtomicLazy`.
- [x] [Poisoning](rust-lang/rfcs#2788 (comment)). It seems like there's [some regret around poisoning in other `std::sync` types that we might want to just avoid upfront for `std::lazy`, especially if that would align with a future `std::mutex` that doesn't poison](https://rust-lang.zulipchat.com/#narrow/stream/219381-t-libs/topic/parking_lot.3A.3AMutex.20in.20std/near/190331199). Personally, if we're adding these types to `std::lazy` instead of `std::sync`, I'd be on-board with not worrying about poisoning in `std::lazy`, and potentially deprecating `std::sync::Once` and `lazy_static` in favour of `std::lazy` down the track if it's possible, rather than attempting to replicate their behavior. cc @Amanieu @sfackler.
- [ ] [Consider making`SyncOnceCell::get` blocking](matklad/once_cell#92). There doesn't seem to be consensus in the linked PR on whether or not that's strictly better than the non-blocking variant.

In general, none of these seem to be really blocking an initial unstable merge, so we could possibly kick off a FCP if y'all are happy?

cc @matklad @pitdicker have I missed anything, or were there any other considerations that have come up since we last looked at this?

Author: Manishearth

src/libcore/lazy.rs

@@ -0,0 +1,379 @@
+//! Lazy values and one-time initialization of static data.
+
+use crate::cell::{Cell, UnsafeCell};
+use crate::fmt;
+use crate::mem;
+use crate::ops::Deref;
+
+/// A cell which can be written to only once.
+///
+/// Unlike `RefCell`, a `OnceCell` only provides shared `&T` references to its value.
+/// Unlike `Cell`, a `OnceCell` doesn't require copying or replacing the value to access it.
+///
+/// # Examples
+///
+/// ```
+/// #![feature(once_cell)]
+///
+/// use std::lazy::OnceCell;
+///
+/// let cell = OnceCell::new();
+/// assert!(cell.get().is_none());
+///
+/// let value: &String = cell.get_or_init(|| {
+///     "Hello, World!".to_string()
+/// });
+/// assert_eq!(value, "Hello, World!");
+/// assert!(cell.get().is_some());
+/// ```
+#[unstable(feature = "once_cell", issue = "74465")]
+pub struct OnceCell<T> {
+    // Invariant: written to at most once.
+    inner: UnsafeCell<Option<T>>,
+}
+
+#[unstable(feature = "once_cell", issue = "74465")]
+impl<T> Default for OnceCell<T> {
+    fn default() -> Self {
+        Self::new()
+    }
+}
+
+#[unstable(feature = "once_cell", issue = "74465")]
+impl<T: fmt::Debug> fmt::Debug for OnceCell<T> {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+        match self.get() {
+            Some(v) => f.debug_tuple("OnceCell").field(v).finish(),
+            None => f.write_str("OnceCell(Uninit)"),
+        }
+    }
+}
+
+#[unstable(feature = "once_cell", issue = "74465")]
+impl<T: Clone> Clone for OnceCell<T> {
+    fn clone(&self) -> OnceCell<T> {
+        let res = OnceCell::new();
+        if let Some(value) = self.get() {
+            match res.set(value.clone()) {
+                Ok(()) => (),
+                Err(_) => unreachable!(),
+            }
+        }
+        res
+    }
+}
+
+#[unstable(feature = "once_cell", issue = "74465")]
+impl<T: PartialEq> PartialEq for OnceCell<T> {
+    fn eq(&self, other: &Self) -> bool {
+        self.get() == other.get()
+    }
+}
+
+#[unstable(feature = "once_cell", issue = "74465")]
+impl<T: Eq> Eq for OnceCell<T> {}
+
+#[unstable(feature = "once_cell", issue = "74465")]
+impl<T> From<T> for OnceCell<T> {
+    fn from(value: T) -> Self {
+        OnceCell { inner: UnsafeCell::new(Some(value)) }
+    }
+}
+
+impl<T> OnceCell<T> {
+    /// Creates a new empty cell.
+    #[unstable(feature = "once_cell", issue = "74465")]
+    pub const fn new() -> OnceCell<T> {
+        OnceCell { inner: UnsafeCell::new(None) }
+    }
+
+    /// Gets the reference to the underlying value.
+    ///
+    /// Returns `None` if the cell is empty.
+    #[unstable(feature = "once_cell", issue = "74465")]
+    pub fn get(&self) -> Option<&T> {
+        // Safety: Safe due to `inner`'s invariant
+        unsafe { &*self.inner.get() }.as_ref()
+    }
+
+    /// Gets the mutable reference to the underlying value.
+    ///
+    /// Returns `None` if the cell is empty.
+    #[unstable(feature = "once_cell", issue = "74465")]
+    pub fn get_mut(&mut self) -> Option<&mut T> {
+        // Safety: Safe because we have unique access
+        unsafe { &mut *self.inner.get() }.as_mut()
+    }
+
+    /// Sets the contents of the cell to `value`.
+    ///
+    /// # Errors
+    ///
+    /// This method returns `Ok(())` if the cell was empty and `Err(value)` if
+    /// it was full.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// #![feature(once_cell)]
+    ///
+    /// use std::lazy::OnceCell;
+    ///
+    /// let cell = OnceCell::new();
+    /// assert!(cell.get().is_none());
+    ///
+    /// assert_eq!(cell.set(92), Ok(()));
+    /// assert_eq!(cell.set(62), Err(62));
+    ///
+    /// assert!(cell.get().is_some());
+    /// ```
+    #[unstable(feature = "once_cell", issue = "74465")]
+    pub fn set(&self, value: T) -> Result<(), T> {
+        // Safety: Safe because we cannot have overlapping mutable borrows
+        let slot = unsafe { &*self.inner.get() };
+        if slot.is_some() {
+            return Err(value);
+        }
+
+        // Safety: This is the only place where we set the slot, no races
+        // due to reentrancy/concurrency are possible, and we've
+        // checked that slot is currently `None`, so this write
+        // maintains the `inner`'s invariant.
+        let slot = unsafe { &mut *self.inner.get() };
+        *slot = Some(value);
+        Ok(())
+    }
+
+    /// Gets the contents of the cell, initializing it with `f`
+    /// if the cell was empty.
+    ///
+    /// # Panics
+    ///
+    /// If `f` panics, the panic is propagated to the caller, and the cell
+    /// remains uninitialized.
+    ///
+    /// It is an error to reentrantly initialize the cell from `f`. Doing
+    /// so results in a panic.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// #![feature(once_cell)]
+    ///
+    /// use std::lazy::OnceCell;
+    ///
+    /// let cell = OnceCell::new();
+    /// let value = cell.get_or_init(|| 92);
+    /// assert_eq!(value, &92);
+    /// let value = cell.get_or_init(|| unreachable!());
+    /// assert_eq!(value, &92);
+    /// ```
+    #[unstable(feature = "once_cell", issue = "74465")]
+    pub fn get_or_init<F>(&self, f: F) -> &T
+    where
+        F: FnOnce() -> T,
+    {
+        match self.get_or_try_init(|| Ok::<T, !>(f())) {
+            Ok(val) => val,
+        }
+    }
+
+    /// Gets the contents of the cell, initializing it with `f` if
+    /// the cell was empty. If the cell was empty and `f` failed, an
+    /// error is returned.
+    ///
+    /// # Panics
+    ///
+    /// If `f` panics, the panic is propagated to the caller, and the cell
+    /// remains uninitialized.
+    ///
+    /// It is an error to reentrantly initialize the cell from `f`. Doing
+    /// so results in a panic.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// #![feature(once_cell)]
+    ///
+    /// use std::lazy::OnceCell;
+    ///
+    /// let cell = OnceCell::new();
+    /// assert_eq!(cell.get_or_try_init(|| Err(())), Err(()));
+    /// assert!(cell.get().is_none());
+    /// let value = cell.get_or_try_init(|| -> Result<i32, ()> {
+    ///     Ok(92)
+    /// });
+    /// assert_eq!(value, Ok(&92));
+    /// assert_eq!(cell.get(), Some(&92))
+    /// ```
+    #[unstable(feature = "once_cell", issue = "74465")]
+    pub fn get_or_try_init<F, E>(&self, f: F) -> Result<&T, E>
+    where
+        F: FnOnce() -> Result<T, E>,
+    {
+        if let Some(val) = self.get() {
+            return Ok(val);
+        }
+        let val = f()?;
+        // Note that *some* forms of reentrant initialization might lead to
+        // UB (see `reentrant_init` test). I believe that just removing this
+        // `assert`, while keeping `set/get` would be sound, but it seems
+        // better to panic, rather than to silently use an old value.
+        assert!(self.set(val).is_ok(), "reentrant init");
+        Ok(self.get().unwrap())
+    }
+
+    /// Consumes the cell, returning the wrapped value.
+    ///
+    /// Returns `None` if the cell was empty.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// #![feature(once_cell)]
+    ///
+    /// use std::lazy::OnceCell;
+    ///
+    /// let cell: OnceCell<String> = OnceCell::new();
+    /// assert_eq!(cell.into_inner(), None);
+    ///
+    /// let cell = OnceCell::new();
+    /// cell.set("hello".to_string()).unwrap();
+    /// assert_eq!(cell.into_inner(), Some("hello".to_string()));
+    /// ```
+    #[unstable(feature = "once_cell", issue = "74465")]
+    pub fn into_inner(self) -> Option<T> {
+        // Because `into_inner` takes `self` by value, the compiler statically verifies
+        // that it is not currently borrowed. So it is safe to move out `Option<T>`.
+        self.inner.into_inner()
+    }
+
+    /// Takes the value out of this `OnceCell`, moving it back to an uninitialized state.
+    ///
+    /// Has no effect and returns `None` if the `OnceCell` hasn't been initialized.
+    ///
+    /// Safety is guaranteed by requiring a mutable reference.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// #![feature(once_cell)]
+    ///
+    /// use std::lazy::OnceCell;
+    ///
+    /// let mut cell: OnceCell<String> = OnceCell::new();
+    /// assert_eq!(cell.take(), None);
+    ///
+    /// let mut cell = OnceCell::new();
+    /// cell.set("hello".to_string()).unwrap();
+    /// assert_eq!(cell.take(), Some("hello".to_string()));
+    /// assert_eq!(cell.get(), None);
+    /// ```
+    #[unstable(feature = "once_cell", issue = "74465")]
+    pub fn take(&mut self) -> Option<T> {
+        mem::take(self).into_inner()
+    }
+}
+
+/// A value which is initialized on the first access.
+///
+/// # Examples
+///
+/// ```
+/// #![feature(once_cell)]
+///
+/// use std::lazy::Lazy;
+///
+/// let lazy: Lazy<i32> = Lazy::new(|| {
+///     println!("initializing");
+///     92
+/// });
+/// println!("ready");
+/// println!("{}", *lazy);
+/// println!("{}", *lazy);
+///
+/// // Prints:
+/// //   ready
+/// //   initializing
+/// //   92
+/// //   92
+/// ```
+#[unstable(feature = "once_cell", issue = "74465")]
+pub struct Lazy<T, F = fn() -> T> {
+    cell: OnceCell<T>,
+    init: Cell<Option<F>>,
+}
+
+#[unstable(feature = "once_cell", issue = "74465")]
+impl<T: fmt::Debug, F> fmt::Debug for Lazy<T, F> {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+        f.debug_struct("Lazy").field("cell", &self.cell).field("init", &"..").finish()
+    }
+}
+
+impl<T, F> Lazy<T, F> {
+    /// Creates a new lazy value with the given initializing function.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// #![feature(once_cell)]
+    ///
+    /// # fn main() {
+    /// use std::lazy::Lazy;
+    ///
+    /// let hello = "Hello, World!".to_string();
+    ///
+    /// let lazy = Lazy::new(|| hello.to_uppercase());
+    ///
+    /// assert_eq!(&*lazy, "HELLO, WORLD!");
+    /// # }
+    /// ```
+    #[unstable(feature = "once_cell", issue = "74465")]
+    pub const fn new(init: F) -> Lazy<T, F> {
+        Lazy { cell: OnceCell::new(), init: Cell::new(Some(init)) }
+    }
+}
+
+impl<T, F: FnOnce() -> T> Lazy<T, F> {
+    /// Forces the evaluation of this lazy value and returns a reference to
+    /// the result.
+    ///
+    /// This is equivalent to the `Deref` impl, but is explicit.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// #![feature(once_cell)]
+    ///
+    /// use std::lazy::Lazy;
+    ///
+    /// let lazy = Lazy::new(|| 92);
+    ///
+    /// assert_eq!(Lazy::force(&lazy), &92);
+    /// assert_eq!(&*lazy, &92);
+    /// ```
+    #[unstable(feature = "once_cell", issue = "74465")]
+    pub fn force(this: &Lazy<T, F>) -> &T {
+        this.cell.get_or_init(|| match this.init.take() {
+            Some(f) => f(),
+            None => panic!("`Lazy` instance has previously been poisoned"),
+        })
+    }
+}
+
+#[unstable(feature = "once_cell", issue = "74465")]
+impl<T, F: FnOnce() -> T> Deref for Lazy<T, F> {
+    type Target = T;
+    fn deref(&self) -> &T {
+        Lazy::force(self)
+    }
+}
+
+#[unstable(feature = "once_cell", issue = "74465")]
+impl<T: Default> Default for Lazy<T> {
+    /// Creates a new lazy value using `Default` as the initializing function.
+    fn default() -> Lazy<T> {
+        Lazy::new(T::default)
+    }
+}

src/libcore/lib.rs

@@ -239,6 +239,8 @@ pub mod char;
 pub mod ffi;
 #[cfg(not(test))] // See #65860
 pub mod iter;
+#[unstable(feature = "once_cell", issue = "74465")]
+pub mod lazy;
 pub mod option;
 pub mod panic;
 pub mod panicking;

src/libcore/tests/lazy.rs

@@ -0,0 +1,124 @@
+use core::{
+    cell::Cell,
+    lazy::{Lazy, OnceCell},
+    sync::atomic::{AtomicUsize, Ordering::SeqCst},
+};
+
+#[test]
+fn once_cell() {
+    let c = OnceCell::new();
+    assert!(c.get().is_none());
+    c.get_or_init(|| 92);
+    assert_eq!(c.get(), Some(&92));
+
+    c.get_or_init(|| panic!("Kabom!"));
+    assert_eq!(c.get(), Some(&92));
+}
+
+#[test]
+fn once_cell_get_mut() {
+    let mut c = OnceCell::new();
+    assert!(c.get_mut().is_none());
+    c.set(90).unwrap();
+    *c.get_mut().unwrap() += 2;
+    assert_eq!(c.get_mut(), Some(&mut 92));
+}
+
+#[test]
+fn once_cell_drop() {
+    static DROP_CNT: AtomicUsize = AtomicUsize::new(0);
+    struct Dropper;
+    impl Drop for Dropper {
+        fn drop(&mut self) {
+            DROP_CNT.fetch_add(1, SeqCst);
+        }
+    }
+
+    let x = OnceCell::new();
+    x.get_or_init(|| Dropper);
+    assert_eq!(DROP_CNT.load(SeqCst), 0);
+    drop(x);
+    assert_eq!(DROP_CNT.load(SeqCst), 1);
+}
+
+#[test]
+fn unsync_once_cell_drop_empty() {
+    let x = OnceCell::<&'static str>::new();
+    drop(x);
+}
+
+#[test]
+fn clone() {
+    let s = OnceCell::new();
+    let c = s.clone();
+    assert!(c.get().is_none());
+
+    s.set("hello").unwrap();
+    let c = s.clone();
+    assert_eq!(c.get().map(|c| *c), Some("hello"));
+}
+
+#[test]
+fn from_impl() {
+    assert_eq!(OnceCell::from("value").get(), Some(&"value"));
+    assert_ne!(OnceCell::from("foo").get(), Some(&"bar"));
+}
+
+#[test]
+fn partialeq_impl() {
+    assert!(OnceCell::from("value") == OnceCell::from("value"));
+    assert!(OnceCell::from("foo") != OnceCell::from("bar"));
+
+    assert!(OnceCell::<&'static str>::new() == OnceCell::new());
+    assert!(OnceCell::<&'static str>::new() != OnceCell::from("value"));
+}
+
+#[test]
+fn into_inner() {
+    let cell: OnceCell<&'static str> = OnceCell::new();
+    assert_eq!(cell.into_inner(), None);
+    let cell = OnceCell::new();
+    cell.set("hello").unwrap();
+    assert_eq!(cell.into_inner(), Some("hello"));
+}
+
+#[test]
+fn lazy_new() {
+    let called = Cell::new(0);
+    let x = Lazy::new(|| {
+        called.set(called.get() + 1);
+        92
+    });
+
+    assert_eq!(called.get(), 0);
+
+    let y = *x - 30;
+    assert_eq!(y, 62);
+    assert_eq!(called.get(), 1);
+
+    let y = *x - 30;
+    assert_eq!(y, 62);
+    assert_eq!(called.get(), 1);
+}
+
+#[test]
+fn aliasing_in_get() {
+    let x = OnceCell::new();
+    x.set(42).unwrap();
+    let at_x = x.get().unwrap(); // --- (shared) borrow of inner `Option<T>` --+
+    let _ = x.set(27); // <-- temporary (unique) borrow of inner `Option<T>`   |
+    println!("{}", at_x); // <------- up until here ---------------------------+
+}
+
+#[test]
+#[should_panic(expected = "reentrant init")]
+fn reentrant_init() {
+    let x: OnceCell<Box<i32>> = OnceCell::new();
+    let dangling_ref: Cell<Option<&i32>> = Cell::new(None);
+    x.get_or_init(|| {
+        let r = x.get_or_init(|| Box::new(92));
+        dangling_ref.set(Some(r));
+        Box::new(62)
+    });
+    eprintln!("use after free: {:?}", dangling_ref.get().unwrap());
+}

src/libcore/tests/lib.rs

@@ -43,6 +43,7 @@
 #![feature(option_unwrap_none)]
 #![feature(peekable_next_if)]
 #![feature(partition_point)]
+#![feature(once_cell)]
 #![feature(unsafe_block_in_unsafe_fn)]
 #![deny(unsafe_op_in_unsafe_fn)]
 
@@ -62,6 +63,7 @@ mod fmt;
 mod hash;
 mod intrinsics;
 mod iter;
+mod lazy;
 mod manually_drop;
 mod mem;
 mod nonzero;

src/libstd/lazy.rs

@@ -287,13 +287,15 @@
 #![feature(linkage)]
 #![feature(llvm_asm)]
 #![feature(log_syntax)]
+#![feature(maybe_uninit_extra)]
 #![feature(maybe_uninit_ref)]
 #![feature(maybe_uninit_slice)]
 #![feature(min_specialization)]
 #![feature(needs_panic_runtime)]
 #![feature(negative_impls)]
 #![feature(never_type)]
 #![feature(nll)]
+#![feature(once_cell)]
 #![feature(optin_builtin_traits)]
 #![feature(or_patterns)]
 #![feature(panic_info_message)]
@@ -477,6 +479,9 @@ pub mod process;
 pub mod sync;
 pub mod time;
 
+#[unstable(feature = "once_cell", issue = "74465")]
+pub mod lazy;
+
 #[stable(feature = "futures_api", since = "1.36.0")]
 pub mod task {
     //! Types and Traits for working with asynchronous tasks.

src/libstd/lib.rs

@@ -132,6 +132,7 @@ unsafe impl Send for Once {}
 #[derive(Debug)]
 pub struct OnceState {
     poisoned: bool,
+    set_state_on_drop_to: Cell<usize>,
 }
 
 /// Initialization value for static [`Once`] values.
@@ -321,7 +322,7 @@ impl Once {
         }
 
         let mut f = Some(f);
-        self.call_inner(true, &mut |p| f.take().unwrap()(&OnceState { poisoned: p }));
+        self.call_inner(true, &mut |p| f.take().unwrap()(p));
     }
 
     /// Returns `true` if some `call_once` call has completed
@@ -385,7 +386,7 @@ impl Once {
     // currently no way to take an `FnOnce` and call it via virtual dispatch
     // without some allocation overhead.
     #[cold]
-    fn call_inner(&self, ignore_poisoning: bool, init: &mut dyn FnMut(bool)) {
+    fn call_inner(&self, ignore_poisoning: bool, init: &mut dyn FnMut(&OnceState)) {
         let mut state_and_queue = self.state_and_queue.load(Ordering::Acquire);
         loop {
             match state_and_queue {
@@ -413,8 +414,12 @@ impl Once {
                     };
                     // Run the initialization function, letting it know if we're
                     // poisoned or not.
-                    init(state_and_queue == POISONED);
-                    waiter_queue.set_state_on_drop_to = COMPLETE;
+                    let init_state = OnceState {
+                        poisoned: state_and_queue == POISONED,
+                        set_state_on_drop_to: Cell::new(COMPLETE),
+                    };
+                    init(&init_state);
+                    waiter_queue.set_state_on_drop_to = init_state.set_state_on_drop_to.get();
                     break;
                 }
                 _ => {
@@ -554,6 +559,14 @@ impl OnceState {
     pub fn poisoned(&self) -> bool {
         self.poisoned
     }
+
+    /// Poison the associated [`Once`] without explicitly panicking.
+    ///
+    /// [`Once`]: struct.Once.html
+    // NOTE: This is currently only exposed for the `lazy` module
+    pub(crate) fn poison(&self) {
+        self.set_state_on_drop_to.set(POISONED);
+    }
 }
 
 #[cfg(all(test, not(target_os = "emscripten")))]