simplicity/

value.rs

// SPDX-License-Identifier: CC0-1.0

//! # Simplicity values
//!
//! Simplicity processes data in terms of [`Value`]s,
//! i.e., inputs, intermediate results and outputs.

use crate::dag::{Dag, DagLike};
use crate::types::{CompleteBound, Final};
use crate::{BitIter, Tmr};

use crate::{BitCollector, EarlyEndOfStreamError};
use core::{cmp, fmt, iter};
use std::collections::VecDeque;
use std::hash::Hash;
use std::sync::Arc;

/// A Simplicity value.
#[derive(Clone)]
pub struct Value {
    /// The underlying data, in "padded" bit-encoded form, with the leftmost
    /// bits encoded in the most-significant bits. So e.g. right(unit) is 0x80.
    ///
    /// We use the padded form, even though it is space-inefficient (sometimes
    /// significantly so), for a couple reasons: first, this is the representation
    /// used by the bit machine, so we want encoding/decoding to be as fast as
    /// possible. Secondly, when destructuring a value (which we need to do during
    /// scribing and pruning, at least) it is quite difficult to destructure a
    /// compact-encoded product, since we would need to determine the compact
    /// bitlengths of the children, and these quantities are not readily available.
    inner: Arc<[u8]>,
    /// An offset, in bits, at which the actual data starts. This is useful
    /// because it allows constructing sub-values of a value without needing
    /// to construct a new `inner` with all the bits offset.
    bit_offset: usize,
    /// The Simplicity type of the value.
    ty: Arc<Final>,
}

/// Reference to a value, or to a sub-value of a value.
#[derive(Debug, Clone, Copy)]
pub struct ValueRef<'v> {
    inner: &'v Arc<[u8]>,
    bit_offset: usize,
    ty: &'v Arc<Final>,
}

// Because two equal values may have different bit offsets, we must manually
// implement the comparison traits. We do so by first comparing types, which
// is constant overhead (this just compares TMRs). If those match, we know
// the lengths and structures match, so we then compare the underlying byte
// iterators.
impl PartialEq for Value {
    fn eq(&self, other: &Self) -> bool {
        self.ty == other.ty && self.raw_byte_iter().eq(other.raw_byte_iter())
    }
}
impl Eq for Value {}

impl PartialOrd for Value {
    fn partial_cmp(&self, other: &Self) -> Option<core::cmp::Ordering> {
        Some(self.cmp(other))
    }
}
impl Ord for Value {
    fn cmp(&self, other: &Self) -> core::cmp::Ordering {
        self.ty
            .cmp(&other.ty)
            .then_with(|| self.raw_byte_iter().cmp(other.raw_byte_iter()))
    }
}

impl core::hash::Hash for Value {
    fn hash<H: core::hash::Hasher>(&self, h: &mut H) {
        b"Simplicity\x1fValue".hash(h);
        self.ty.hash(h);
        for val in self.raw_byte_iter() {
            val.hash(h);
        }
    }
}

impl DagLike for ValueRef<'_> {
    type Node = Self;

    fn data(&self) -> &Self {
        self
    }

    fn as_dag_node(&self) -> Dag<Self> {
        if let Some((left, right)) = self.as_product() {
            Dag::Binary(left, right)
        } else if let Some(left) = self.as_left() {
            Dag::Unary(left)
        } else if let Some(right) = self.as_right() {
            Dag::Unary(right)
        } else {
            assert!(self.ty.is_unit());
            Dag::Nullary
        }
    }
}

impl<'v> ValueRef<'v> {
    /// Check if the value is a unit.
    pub fn is_unit(&self) -> bool {
        self.ty.is_unit()
    }

    /// Helper function to read the first bit of a value
    ///
    /// If the first bit is not available (e.g. if the value has zero size)
    /// then returns None.
    fn first_bit(&self) -> Option<bool> {
        let mask = if self.bit_offset % 8 == 0 {
            0x80
        } else {
            1 << (7 - self.bit_offset % 8)
        };
        let res = self
            .inner
            .get(self.bit_offset / 8)
            .map(|x| x & mask == mask);
        res
    }

    /// Access the inner value of a left sum value.
    pub fn as_left(&self) -> Option<Self> {
        if self.first_bit() == Some(false) {
            if let Some((lty, rty)) = self.ty.as_sum() {
                let sum_width = 1 + cmp::max(lty.bit_width(), rty.bit_width());
                Some(Self {
                    inner: self.inner,
                    bit_offset: self.bit_offset + sum_width - lty.bit_width(),
                    ty: lty,
                })
            } else {
                None
            }
        } else {
            None
        }
    }

    /// Access the inner value of a right sum value.
    pub fn as_right(&self) -> Option<Self> {
        if self.first_bit() == Some(true) {
            if let Some((lty, rty)) = self.ty.as_sum() {
                let sum_width = 1 + cmp::max(lty.bit_width(), rty.bit_width());
                Some(Self {
                    inner: self.inner,
                    bit_offset: self.bit_offset + sum_width - rty.bit_width(),
                    ty: rty,
                })
            } else {
                None
            }
        } else {
            None
        }
    }

    /// Access the inner values of a product value.
    pub fn as_product(&self) -> Option<(Self, Self)> {
        if let Some((lty, rty)) = self.ty.as_product() {
            Some((
                Self {
                    inner: self.inner,
                    bit_offset: self.bit_offset,
                    ty: lty,
                },
                Self {
                    inner: self.inner,
                    bit_offset: self.bit_offset + lty.bit_width(),
                    ty: rty,
                },
            ))
        } else {
            None
        }
    }

    /// Convert the reference back to a value.
    pub fn to_value(&self) -> Value {
        Value {
            inner: Arc::clone(self.inner),
            bit_offset: self.bit_offset,
            ty: Arc::clone(self.ty),
        }
    }

    /// Try to convert the value into a word.
    pub fn to_word(&self) -> Option<Word> {
        self.ty.as_word().map(|n| Word {
            value: self.to_value(),
            n,
        })
    }

    /// Yields an iterator over the "raw bytes" of the value.
    ///
    /// The returned bytes match the padded bit-encoding of the value. You
    /// may wish to call [`Self::iter_padded`] instead to obtain the bits,
    /// but this method is more efficient in some contexts.
    pub fn raw_byte_iter(&self) -> RawByteIter<'v> {
        RawByteIter {
            value: *self,
            yielded_bytes: 0,
        }
    }

    /// Return an iterator over the compact bit encoding of the value.
    ///
    /// This encoding is used for writing witness data and for computing IHRs.
    pub fn iter_compact(&self) -> CompactBitsIter<'v> {
        CompactBitsIter::new(*self)
    }

    /// Return an iterator over the padded bit encoding of the value.
    ///
    /// This encoding is used to represent the value in the Bit Machine.
    pub fn iter_padded(&self) -> PreOrderIter<'v> {
        PreOrderIter {
            inner: BitIter::new(self.raw_byte_iter()).take(self.ty.bit_width()),
        }
    }
}

pub struct RawByteIter<'v> {
    value: ValueRef<'v>,
    yielded_bytes: usize,
}

impl Iterator for RawByteIter<'_> {
    type Item = u8;
    fn next(&mut self) -> Option<Self::Item> {
        if 8 * self.yielded_bytes >= self.value.ty.bit_width() {
            None
        } else if self.value.bit_offset % 8 == 0 {
            self.yielded_bytes += 1;

            Some(self.value.inner[self.value.bit_offset / 8 + self.yielded_bytes - 1])
        } else {
            self.yielded_bytes += 1;

            let ret1 = self.value.inner[self.value.bit_offset / 8 + self.yielded_bytes - 1];
            let ret2 = self
                .value
                .inner
                .get(self.value.bit_offset / 8 + self.yielded_bytes)
                .copied()
                .unwrap_or(0);
            let bit_offset = self.value.bit_offset % 8;
            Some((ret1 << bit_offset) | (ret2 >> (8 - bit_offset)))
        }
    }
}

pub struct PreOrderIter<'v> {
    inner: iter::Take<BitIter<RawByteIter<'v>>>,
}

impl Iterator for PreOrderIter<'_> {
    type Item = bool;
    fn next(&mut self) -> Option<Self::Item> {
        self.inner.next()
    }
}

/// Helper function to right-shift a value by one bit.
///
/// This is used for putting a value into a sum type. It takes the value's
/// bit encoding and current bit offset, and returns a new bit-encoding with
/// a new bit inserted upfront and a new bit offset.
fn right_shift_1(inner: &Arc<[u8]>, bit_offset: usize, new_bit: bool) -> (Arc<[u8]>, usize) {
    // If the current bit offset is nonzero this is super easy: we just
    // lower the bit offset and call that a fix.
    if bit_offset > 0 {
        let new_bit_offset = bit_offset - 1;
        let mask = 1u8 << (7 - new_bit_offset % 8);
        let current_bit = inner[new_bit_offset / 8] & mask != 0;
        if current_bit == new_bit {
            (Arc::clone(inner), new_bit_offset)
        } else if new_bit {
            let mut bx: Box<[u8]> = inner.as_ref().into();
            bx[new_bit_offset / 8] |= mask;
            (bx.into(), new_bit_offset)
        } else {
            let mut bx: Box<[u8]> = inner.as_ref().into();
            bx[new_bit_offset / 8] &= !mask;
            (bx.into(), new_bit_offset)
        }
    } else {
        // If the current bit offset is 0, we just shift everything right by 8
        // and then do pretty-much the same thing as above. This sometimes will
        // waste 7 bits, but it avoids needing to iterate through all of `inner`.
        let mut new = Vec::with_capacity(inner.len() + 1);
        new.push(u8::from(new_bit));
        new.extend(inner.iter().copied());
        (new.into(), 7)
    }
}

/// Helper function to copy `nbits` bits from `src`, starting at bit-offset `src_offset`,
/// into `dst`, starting at bit-offset `dst_offset`.
///
/// Thanks ChatGPT for suggesting we extract this function.
fn copy_bits(src: &[u8], src_offset: usize, dst: &mut [u8], dst_offset: usize, nbits: usize) {
    // For each bit i in 0..nbits, extract the bit from `src`
    // and insert it into `dst`.
    for i in 0..nbits {
        let bit = (src[(src_offset + i) / 8] >> (7 - (src_offset + i) % 8)) & 1;
        dst[(dst_offset + i) / 8] |= bit << (7 - (dst_offset + i) % 8);
    }
}

/// Helper function to take the product of two values (i.e. their concatenation).
///
/// If either `left_inner` or `right_inner` is not provided, it is assumed to be
/// padding and will be stored as all zeros.
///
/// Returns the new bit data and the offset (NOT the length) of the data.
fn product(
    left: Option<(&Arc<[u8]>, usize)>,
    left_bit_length: usize,
    right: Option<(&Arc<[u8]>, usize)>,
    right_bit_length: usize,
) -> (Arc<[u8]>, usize) {
    if left_bit_length == 0 {
        if let Some((right, right_bit_offset)) = right {
            (Arc::clone(right), right_bit_offset)
        } else if right_bit_length == 0 {
            (Arc::new([]), 0)
        } else {
            (Arc::from(vec![0; right_bit_length.div_ceil(8)]), 0)
        }
    } else if right_bit_length == 0 {
        if let Some((lt, left_bit_offset)) = left {
            (Arc::clone(lt), left_bit_offset)
        } else {
            (Arc::from(vec![0; left_bit_length.div_ceil(8)]), 0)
        }
    } else {
        // Both left and right have nonzero lengths. This is the only "real" case
        // in which we have to do something beyond cloning Arcs or allocating
        // zeroed vectors. In this case we left-shift both as much as possible.
        let mut bx = Box::<[u8]>::from(vec![0; (left_bit_length + right_bit_length).div_ceil(8)]);
        if let Some((left, left_bit_offset)) = left {
            copy_bits(left, left_bit_offset, &mut bx, 0, left_bit_length);
        }
        if let Some((right, right_bit_offset)) = right {
            copy_bits(
                right,
                right_bit_offset,
                &mut bx,
                left_bit_length,
                right_bit_length,
            );
        }

        (bx.into(), 0)
    }
}

impl Value {
    /// Make a cheap copy of the value.
    pub fn shallow_clone(&self) -> Self {
        Self {
            inner: Arc::clone(&self.inner),
            bit_offset: self.bit_offset,
            ty: Arc::clone(&self.ty),
        }
    }

    /// Access the type of the value.
    pub fn ty(&self) -> &Final {
        &self.ty
    }

    /// Create the unit value.
    pub fn unit() -> Self {
        Self {
            inner: Arc::new([]),
            bit_offset: 0,
            ty: Final::unit(),
        }
    }

    /// Create a left value that wraps the given `inner` value.
    pub fn left(inner: Self, right: Arc<Final>) -> Self {
        let total_width = cmp::max(inner.ty.bit_width(), right.bit_width());

        let (concat, concat_offset) = product(
            None,
            total_width - inner.ty.bit_width(),
            Some((&inner.inner, inner.bit_offset)),
            inner.ty.bit_width(),
        );
        let (new_inner, new_bit_offset) = right_shift_1(&concat, concat_offset, false);
        Self {
            inner: new_inner,
            bit_offset: new_bit_offset,
            ty: Final::sum(Arc::clone(&inner.ty), right),
        }
    }

    /// Create a right value that wraps the given `inner` value.
    pub fn right(left: Arc<Final>, inner: Self) -> Self {
        let total_width = cmp::max(left.bit_width(), inner.ty.bit_width());
        let (concat, concat_offset) = product(
            None,
            total_width - inner.ty.bit_width(),
            Some((&inner.inner, inner.bit_offset)),
            inner.ty.bit_width(),
        );
        let (new_inner, new_bit_offset) = right_shift_1(&concat, concat_offset, true);
        Self {
            inner: new_inner,
            bit_offset: new_bit_offset,
            ty: Final::sum(left, Arc::clone(&inner.ty)),
        }
    }

    /// Create a product value that wraps the given `left` and `right` values.
    pub fn product(left: Self, right: Self) -> Self {
        let (new_inner, new_bit_offset) = product(
            Some((&left.inner, left.bit_offset)),
            left.ty.bit_width(),
            Some((&right.inner, right.bit_offset)),
            right.ty.bit_width(),
        );
        Self {
            inner: new_inner,
            bit_offset: new_bit_offset,
            ty: Final::product(Arc::clone(&left.ty), Arc::clone(&right.ty)),
        }
    }

    /// Create a none value.
    pub fn none(right: Arc<Final>) -> Self {
        Self::left(Value::unit(), right)
    }

    /// Create a some value.
    pub fn some(inner: Self) -> Self {
        Self::right(Final::unit(), inner)
    }

    /// Return the bit length of the value in compact encoding.
    pub fn compact_len(&self) -> usize {
        self.iter_compact().count()
    }

    /// Return the bit length of the value in padded encoding.
    pub fn padded_len(&self) -> usize {
        self.ty.bit_width()
    }

    /// Check if the value is a nested product of units.
    /// In this case, the value contains no information.
    pub fn is_empty(&self) -> bool {
        self.ty.bit_width() == 0
    }

    /// Check if the value is a unit.
    pub fn is_unit(&self) -> bool {
        self.ty.is_unit()
    }

    /// A reference to this value, which can be recursed over.
    pub fn as_ref(&self) -> ValueRef<'_> {
        ValueRef {
            inner: &self.inner,
            bit_offset: self.bit_offset,
            ty: &self.ty,
        }
    }

    /// Access the inner value of a left sum value.
    pub fn as_left(&self) -> Option<ValueRef<'_>> {
        self.as_ref().as_left()
    }

    /// Access the inner value of a right sum value.
    pub fn as_right(&self) -> Option<ValueRef<'_>> {
        self.as_ref().as_right()
    }

    /// Access the inner values of a product value.
    pub fn as_product(&self) -> Option<(ValueRef<'_>, ValueRef<'_>)> {
        self.as_ref().as_product()
    }

    /// Create a 1-bit integer.
    ///
    /// ## Panics
    ///
    /// The value is out of range.
    pub fn u1(value: u8) -> Self {
        assert!(value <= 1, "{} out of range for Value::u1", value);
        Self {
            inner: Arc::new([value]),
            bit_offset: 7,
            ty: Final::two_two_n(0),
        }
    }

    /// Create a 2-bit integer.
    ///
    /// ## Panics
    ///
    /// The value is out of range.
    pub fn u2(value: u8) -> Self {
        assert!(value <= 3, "{} out of range for Value::u2", value);
        Self {
            inner: Arc::new([value]),
            bit_offset: 6,
            ty: Final::two_two_n(1),
        }
    }

    /// Create a 4-bit integer.
    ///
    /// ## Panics
    ///
    /// The value is ouf of range.
    pub fn u4(value: u8) -> Self {
        assert!(value <= 15, "{} out of range for Value::u2", value);
        Self {
            inner: Arc::new([value]),
            bit_offset: 4,
            ty: Final::two_two_n(2),
        }
    }

    /// Create an 8-bit integer.
    pub fn u8(value: u8) -> Self {
        Self {
            inner: Arc::new([value]),
            bit_offset: 0,
            ty: Final::two_two_n(3),
        }
    }

    /// Create a 16-bit integer.
    pub fn u16(bytes: u16) -> Self {
        Self {
            inner: Arc::new(bytes.to_be_bytes()),
            bit_offset: 0,
            ty: Final::two_two_n(4),
        }
    }

    /// Create a 32-bit integer.
    pub fn u32(bytes: u32) -> Self {
        Self {
            inner: Arc::new(bytes.to_be_bytes()),
            bit_offset: 0,
            ty: Final::two_two_n(5),
        }
    }

    /// Create a 64-bit integer.
    pub fn u64(bytes: u64) -> Self {
        Self {
            inner: Arc::new(bytes.to_be_bytes()),
            bit_offset: 0,
            ty: Final::two_two_n(6),
        }
    }

    /// Create a 128-bit integer.
    pub fn u128(bytes: u128) -> Self {
        Self {
            inner: Arc::new(bytes.to_be_bytes()),
            bit_offset: 0,
            ty: Final::two_two_n(7),
        }
    }

    /// Create a 256-bit integer.
    pub fn u256(bytes: [u8; 32]) -> Self {
        Self {
            inner: Arc::new(bytes),
            bit_offset: 0,
            ty: Final::two_two_n(8),
        }
    }

    /// Create a 512-bit integer.
    pub fn u512(bytes: [u8; 64]) -> Self {
        Self {
            inner: Arc::new(bytes),
            bit_offset: 0,
            ty: Final::two_two_n(9),
        }
    }

    /// Create a value from a byte array.
    ///
    /// ## Panics
    ///
    /// The array length is not a power of two.
    pub fn from_byte_array<const N: usize>(bytes: [u8; N]) -> Self {
        assert!(N.is_power_of_two(), "Array length must be a power of two");
        let mut values: VecDeque<_> = bytes.into_iter().map(Value::u8).collect();

        while values.len() > 1 {
            let mut alt_values = VecDeque::with_capacity(values.len() / 2);

            while let (Some(left), Some(right)) = (values.pop_front(), values.pop_front()) {
                alt_values.push_back(Value::product(left, right));
            }

            values = alt_values;
        }

        values.into_iter().next().unwrap()
    }

    /// Yields an iterator over the "raw bytes" of the value.
    ///
    /// The returned bytes match the padded bit-encoding of the value. You
    /// may wish to call [`Self::iter_padded`] instead to obtain the bits,
    /// but this method is more efficient in some contexts.
    pub fn raw_byte_iter(&self) -> RawByteIter<'_> {
        self.as_ref().raw_byte_iter()
    }

    /// Return an iterator over the compact bit encoding of the value.
    ///
    /// This encoding is used for writing witness data and for computing IHRs.
    pub fn iter_compact(&self) -> CompactBitsIter<'_> {
        self.as_ref().iter_compact()
    }

    /// Return an iterator over the padded bit encoding of the value.
    ///
    /// This encoding is used to represent the value in the Bit Machine.
    pub fn iter_padded(&self) -> PreOrderIter<'_> {
        self.as_ref().iter_padded()
    }

    /// Check if the value is of the given type.
    pub fn is_of_type(&self, ty: &Final) -> bool {
        self.ty.as_ref() == ty
    }

    /// Get the zero value for the given type.
    ///
    /// The zero value serializes to a string of zeroes.
    ///
    /// ## Construction
    ///
    /// - `zero( 1 )` = `()`
    /// - `zero( A + B )` = `zero(A)`
    /// - `zero( A × B )` = `zero(A) × zero(B)`
    pub fn zero(ty: &Final) -> Self {
        Self {
            inner: Arc::from(vec![0; ty.bit_width().div_ceil(8)]),
            bit_offset: 0,
            ty: Arc::new(ty.clone()),
        }
    }

    /// Try to convert the value into a word.
    ///
    /// The value is cheaply cloned.
    pub fn to_word(&self) -> Option<Word> {
        self.ty.as_word().map(|n| Word {
            value: self.shallow_clone(),
            n,
        })
    }

    /// Prune the value down to the given type.
    ///
    /// The pruned type must be _smaller than or equal to_ the current type of the value.
    /// Otherwise, this method returns `None`.
    ///
    /// ## Smallness
    ///
    /// - `T` ≤ `T` for all types `T`
    /// - `1` ≤ `T` for all types `T`
    /// - `A1 + B1` ≤ `A2 + B2` if `A1` ≤ `A2` and `B1` ≤ `B2`
    /// - `A1 × B1` ≤ `A2 × B2` if `A1` ≤ `A2` and `B1` ≤ `B2`
    ///
    /// ## Pruning
    ///
    /// - `prune( v: T, 1 )` = `(): 1`
    /// - `prune( L(l): A1 + B1, A2 + B2 )` = `prune(l: A1, A2) : A2 + B2`
    /// - `prune( R(r): A1 + B1, A2 + B2 )` = `prune(r: B1, B2) : A2 + B2`
    /// - `prune( (l, r): A1 × B1, A2 × B2 )` = `( prune(l: A1, A2), prune(r: B1, B2): A2 × B2`
    pub fn prune(&self, pruned_ty: &Final) -> Option<Self> {
        enum Task<'v, 'ty> {
            Prune(ValueRef<'v>, &'ty Final),
            MakeLeft(Arc<Final>),
            MakeRight(Arc<Final>),
            MakeProduct,
        }

        let mut stack = vec![Task::Prune(self.as_ref(), pruned_ty)];
        let mut output = vec![];

        while let Some(task) = stack.pop() {
            match task {
                Task::Prune(value, pruned_ty) if value.ty.as_ref() == pruned_ty => {
                    output.push(value.to_value())
                }
                Task::Prune(value, pruned_ty) => match pruned_ty.bound() {
                    CompleteBound::Unit => output.push(Value::unit()),
                    CompleteBound::Sum(l_ty, r_ty) => {
                        if let Some(l_value) = value.as_left() {
                            stack.push(Task::MakeLeft(Arc::clone(r_ty)));
                            stack.push(Task::Prune(l_value, l_ty));
                        } else {
                            let r_value = value.as_right()?;
                            stack.push(Task::MakeRight(Arc::clone(l_ty)));
                            stack.push(Task::Prune(r_value, r_ty));
                        }
                    }
                    CompleteBound::Product(l_ty, r_ty) => {
                        let (l_value, r_value) = value.as_product()?;
                        stack.push(Task::MakeProduct);
                        stack.push(Task::Prune(r_value, r_ty));
                        stack.push(Task::Prune(l_value, l_ty));
                    }
                },
                Task::MakeLeft(r_ty) => {
                    let l_value = output.pop().unwrap();
                    output.push(Value::left(l_value, r_ty));
                }
                Task::MakeRight(l_ty) => {
                    let r_value = output.pop().unwrap();
                    output.push(Value::right(l_ty, r_value));
                }
                Task::MakeProduct => {
                    let r_value = output.pop().unwrap();
                    let l_value = output.pop().unwrap();
                    output.push(Value::product(l_value, r_value));
                }
            }
        }

        debug_assert_eq!(output.len(), 1);
        output.pop()
    }
}

impl fmt::Debug for Value {
    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
        f.debug_struct("Value")
            .field("value", &format_args!("{}", self))
            .field("ty", &self.ty)
            .field("raw_value", &self.inner)
            .field("raw_bit_offset", &self.bit_offset)
            .finish()
    }
}

impl fmt::Display for Value {
    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
        // This is a copy of the logic inside `CompactBitsIter` except
        // that we handle products more explicitly.
        enum S<'v> {
            Disp(ValueRef<'v>),
            DispCh(char),
        }

        let mut stack = Vec::with_capacity(1024);
        // Next node to visit.
        stack.push(S::Disp(self.as_ref()));

        'main_loop: while let Some(next) = stack.pop() {
            let value = match next {
                S::Disp(ref value) => value,
                S::DispCh(ch) => {
                    write!(f, "{}", ch)?;
                    continue;
                }
            };

            if value.is_unit() {
                f.write_str("ε")?;
            } else {
                // First, write any bitstrings out
                for tmr in &Tmr::TWO_TWO_N {
                    if value.ty.tmr() == *tmr {
                        if value.ty.bit_width() < 4 {
                            f.write_str("0b")?;
                            for bit in value.iter_padded() {
                                f.write_str(if bit { "1" } else { "0" })?;
                            }
                        } else {
                            f.write_str("0x")?;
                            // Annoyingly `array_chunks` is unstable so we have to do it manually
                            // https://github.com/rust-lang/rust/issues/100450
                            let mut iter = value.iter_padded();
                            while let (Some(a), Some(b), Some(c), Some(d)) =
                                (iter.next(), iter.next(), iter.next(), iter.next())
                            {
                                let n = (u8::from(a) << 3)
                                    + (u8::from(b) << 2)
                                    + (u8::from(c) << 1)
                                    + u8::from(d);
                                write!(f, "{:x}", n)?;
                            }
                        }
                        continue 'main_loop;
                    }
                }

                // If we don't have a bitstring, then write out the explicit value.
                if let Some(l_value) = value.as_left() {
                    f.write_str("L(")?;
                    stack.push(S::DispCh(')'));
                    stack.push(S::Disp(l_value));
                } else if let Some(r_value) = value.as_right() {
                    f.write_str("R(")?;
                    stack.push(S::DispCh(')'));
                    stack.push(S::Disp(r_value));
                } else if let Some((l_value, r_value)) = value.as_product() {
                    stack.push(S::DispCh(')'));
                    stack.push(S::Disp(r_value));
                    stack.push(S::DispCh(','));
                    stack.push(S::Disp(l_value));
                    stack.push(S::DispCh('('));
                } else {
                    unreachable!()
                }
            }
        }
        Ok(())
    }
}

/// An iterator over the bits of the compact encoding of a [`Value`].
#[derive(Debug, Clone)]
pub struct CompactBitsIter<'v> {
    stack: Vec<ValueRef<'v>>,
}

impl<'v> CompactBitsIter<'v> {
    /// Create an iterator over the bits of the compact encoding of the `value`.
    fn new(value: ValueRef<'v>) -> Self {
        Self { stack: vec![value] }
    }
}

impl Iterator for CompactBitsIter<'_> {
    type Item = bool;

    fn next(&mut self) -> Option<Self::Item> {
        while let Some(value) = self.stack.pop() {
            if value.ty.bit_width() == 0 {
                // NOP
            } else if let Some(l_value) = value.as_left() {
                self.stack.push(l_value);
                return Some(false);
            } else if let Some(r_value) = value.as_right() {
                self.stack.push(r_value);
                return Some(true);
            } else if let Some((l_value, r_value)) = value.as_product() {
                self.stack.push(r_value);
                self.stack.push(l_value);
            }
        }

        None
    }
}

impl Value {
    /// Decode a value of the given type from its compact bit encoding.
    pub fn from_compact_bits<I: Iterator<Item = u8>>(
        bits: &mut BitIter<I>,
        ty: &Final,
    ) -> Result<Self, EarlyEndOfStreamError> {
        enum State<'a> {
            ProcessType(&'a Final),
            DoSumL(Arc<Final>),
            DoSumR(Arc<Final>),
            DoProduct,
        }

        let mut stack = vec![State::ProcessType(ty)];
        let mut result_stack = vec![];
        while let Some(state) = stack.pop() {
            match state {
                State::ProcessType(ty) if ty.has_padding() => match ty.bound() {
                    CompleteBound::Unit => result_stack.push(Value::unit()),
                    CompleteBound::Sum(ref l, ref r) => {
                        if !bits.next().ok_or(EarlyEndOfStreamError)? {
                            stack.push(State::DoSumL(Arc::clone(r)));
                            stack.push(State::ProcessType(l));
                        } else {
                            stack.push(State::DoSumR(Arc::clone(l)));
                            stack.push(State::ProcessType(r));
                        }
                    }
                    CompleteBound::Product(ref l, ref r) => {
                        stack.push(State::DoProduct);
                        stack.push(State::ProcessType(r));
                        stack.push(State::ProcessType(l));
                    }
                },
                State::ProcessType(ty) => {
                    // no padding no problem
                    result_stack.push(Value::from_padded_bits(bits, ty)?);
                }
                State::DoSumL(r) => {
                    let val = result_stack.pop().unwrap();
                    result_stack.push(Value::left(val, r));
                }
                State::DoSumR(l) => {
                    let val = result_stack.pop().unwrap();
                    result_stack.push(Value::right(l, val));
                }
                State::DoProduct => {
                    let val_r = result_stack.pop().unwrap();
                    let val_l = result_stack.pop().unwrap();
                    result_stack.push(Value::product(val_l, val_r));
                }
            }
        }
        debug_assert_eq!(result_stack.len(), 1);
        Ok(result_stack.pop().unwrap())
    }

    /// Decode a value of the given type from its padded bit encoding.
    pub fn from_padded_bits<I: Iterator<Item = u8>>(
        bits: &mut BitIter<I>,
        ty: &Final,
    ) -> Result<Self, EarlyEndOfStreamError> {
        const MAX_INITIAL_ALLOC: usize = 32 * 1024 * 1024; // 4 megabytes

        let cap = cmp::min(MAX_INITIAL_ALLOC, ty.bit_width().div_ceil(8));
        let mut blob = Vec::with_capacity(cap);
        for _ in 0..ty.bit_width() / 8 {
            blob.push(bits.read_u8()?);
        }
        let mut last = 0u8;
        for i in 0..ty.bit_width() % 8 {
            if bits.read_bit()? {
                last |= 1 << (7 - i);
            }
        }
        blob.push(last);

        Ok(Value {
            inner: blob.into(),
            bit_offset: 0,
            ty: Arc::new(ty.clone()),
        })
    }
}

/// A Simplicity word. A value of type `TWO^(2^n)` for some `0 ≤ n < 32`.
#[derive(Clone, Eq, PartialEq, PartialOrd, Ord, Hash)]
pub struct Word {
    /// Value of type `TWO^(2^n)`.
    value: Value,
    /// 0 ≤ n < 32.
    n: u32,
}

macro_rules! construct_word_fallible {
    ($name: ident, $n: expr, $text: expr) => {
        #[doc = "Create"]
        #[doc = $text]
        #[doc = "word.\n\n"]
        #[doc = "## Panics\n"]
        #[doc = "The value is ouf of range."]
        pub fn $name(bit: u8) -> Self {
            Self {
                value: Value::$name(bit),
                n: $n,
            }
        }
    };
}

macro_rules! construct_word {
    ($name: ident, $ty: ty, $n: expr, $text: expr) => {
        #[doc = "Create"]
        #[doc = $text]
        #[doc = "word."]
        pub fn $name(bit: $ty) -> Self {
            Self {
                value: Value::$name(bit),
                n: $n,
            }
        }
    };
}

impl Word {
    /// Concatenate two words into a larger word.
    ///
    /// Both words have to have the same length, which is 2^n bits.
    /// The resulting word will be 2^(n + 1) bits long.
    ///
    /// Returns `None` if the words differ in length.
    ///
    /// Returns `None` if the words are already 2^31 bits long
    /// _(the resulting word would be longer than 2^31 bits, which is not supported)_.
    pub fn product(self, right: Self) -> Option<Self> {
        if self.n == right.n && self.n < 30 {
            Some(Self {
                value: Value::product(self.value, right.value),
                n: self.n + 1,
            })
        } else {
            None
        }
    }

    construct_word_fallible!(u1, 0, "a 1-bit");
    construct_word_fallible!(u2, 1, "a 2-bit");
    construct_word_fallible!(u4, 2, "a 4-bit");
    construct_word!(u8, u8, 3, "an 8-bit");
    construct_word!(u16, u16, 4, "a 16-bit");
    construct_word!(u32, u32, 5, "a 32-bit");
    construct_word!(u64, u64, 6, "a 64-bit");
    construct_word!(u128, u128, 7, "a 128-bit");
    construct_word!(u256, [u8; 32], 8, "a 256-bit");
    construct_word!(u512, [u8; 64], 9, "a 512-bit");

    /// Make a cheap copy of the word.
    pub fn shallow_clone(&self) -> Self {
        Self {
            value: self.value.shallow_clone(),
            n: self.n,
        }
    }

    /// Access the value of the word.
    pub fn as_value(&self) -> &Value {
        &self.value
    }

    /// The word is of type `TWO^(2^n)`. Return `n`.
    pub fn n(&self) -> u32 {
        self.n
    }

    /// Return the bit length of the word.
    ///
    /// The word is of type `TWO^(2^n)`. Return `2^n`.
    #[allow(clippy::len_without_is_empty)]
    pub fn len(&self) -> usize {
        2usize.pow(self.n)
    }

    /// Return an iterator over the bit encoding of the word.
    ///
    /// Words have no padding, so their compact encoding is the same as the padded encoding.
    /// The universal encoding can be used in all situations.
    pub fn iter(&self) -> impl Iterator<Item = bool> + '_ {
        self.value.iter_compact()
    }

    /// Decode a word of type `TWO^(2^n)` from bits.
    ///
    /// ## Panics
    ///
    /// n is greater than 31.
    pub fn from_bits<I: Iterator<Item = u8>>(
        bits: &mut BitIter<I>,
        n: u32,
    ) -> Result<Self, EarlyEndOfStreamError> {
        assert!(n < 32, "TWO^(2^{n}) is not supported as a word type");
        let ty = Final::two_two_n(n as usize); // cast safety: 32-bit machine or higher
        let value = Value::from_compact_bits(bits, &ty)?;
        Ok(Self { value, n })
    }
}

impl fmt::Debug for Word {
    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
        fmt::Display::fmt(self, f)
    }
}

impl fmt::Display for Word {
    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
        use hex::DisplayHex;

        if let Ok(hex) = self.iter().try_collect_bytes() {
            write!(f, "0x{}", hex.as_hex())
        } else {
            f.write_str("0b")?;
            for bit in self.iter() {
                match bit {
                    false => f.write_str("0")?,
                    true => f.write_str("1")?,
                }
            }
            Ok(())
        }
    }
}

#[cfg(test)]
mod tests {
    use super::*;
    use crate::bit_encoding::{BitCollector as _, BitIter};
    use crate::jet::type_name::TypeName;

    #[test]
    fn value_len() {
        let v = Value::u4(6);
        let s_v = Value::some(v.shallow_clone());
        let n_v = Value::none(Final::two_two_n(2));

        assert_eq!(v.compact_len(), 4);
        assert_eq!(v.padded_len(), 4);
        assert_eq!(s_v.compact_len(), 5);
        assert_eq!(s_v.padded_len(), 5);
        assert_eq!(n_v.compact_len(), 1);
        assert_eq!(n_v.padded_len(), 5);
    }

    #[test]
    fn value_display() {
        // Only test a couple values becasue we probably want to change this
        // at some point and will have to redo this test.
        assert_eq!(Value::u1(0).to_string(), "0b0",);
        assert_eq!(Value::u1(1).to_string(), "0b1",);
        assert_eq!(Value::u4(6).to_string(), "0x6",);
    }

    #[test]
    fn is_of_type() {
        let value_typename = [
            (Value::unit(), TypeName(b"1")),
            (Value::left(Value::unit(), Final::unit()), TypeName(b"+11")),
            (Value::right(Final::unit(), Value::unit()), TypeName(b"+11")),
            (
                Value::left(Value::unit(), Final::two_two_n(8)),
                TypeName(b"+1h"),
            ),
            (
                Value::right(Final::two_two_n(8), Value::unit()),
                TypeName(b"+h1"),
            ),
            (
                Value::product(Value::unit(), Value::unit()),
                TypeName(b"*11"),
            ),
            (Value::u8(u8::MAX), TypeName(b"c")),
            (Value::u64(u64::MAX), TypeName(b"l")),
        ];

        for (value, typename) in value_typename {
            let ty = typename.to_final();
            assert!(value.is_of_type(ty.as_ref()));
        }
    }

    #[test]
    fn prune_regression_1() {
        // Found this when fuzzing Elements; unsure how to reduce it further.
        let nontrivial_sum = Value::product(
            Value::right(Final::two_two_n(4), Value::u16(0)),
            Value::u8(0),
        );
        // Formatting should succeed and have no effect.
        let _ = format!("{nontrivial_sum}");
        // Pruning should succeed and have no effect.
        assert_eq!(
            nontrivial_sum.prune(nontrivial_sum.ty()),
            Some(nontrivial_sum)
        );
    }

    #[test]
    fn prune() {
        let test_vectors = [
            (Value::unit(), Value::unit()),
            (Value::u64(42), Value::unit()),
            (
                Value::left(Value::u64(42), Final::u64()),
                Value::left(Value::u64(42), Final::unit()),
            ),
            (
                Value::right(Final::u64(), Value::u64(1337)),
                Value::right(Final::unit(), Value::u64(1337)),
            ),
            (
                Value::product(Value::u64(42), Value::u64(1337)),
                Value::product(Value::u64(42), Value::unit()),
            ),
            (
                Value::product(Value::u64(42), Value::u64(1337)),
                Value::product(Value::unit(), Value::u64(1337)),
            ),
        ];

        for (value, expected_pruned_value) in test_vectors {
            assert_eq!(
                value.prune(expected_pruned_value.ty()),
                Some(expected_pruned_value)
            );
        }

        let bad_test_vectors = [
            (Value::unit(), Final::u1()),
            (
                Value::product(Value::unit(), Value::unit()),
                Final::sum(Final::unit(), Final::unit()),
            ),
            (
                Value::left(Value::unit(), Final::unit()),
                Final::product(Final::unit(), Final::unit()),
            ),
            (
                Value::right(Final::unit(), Value::unit()),
                Final::product(Final::unit(), Final::unit()),
            ),
        ];

        for (value, pruned_ty) in bad_test_vectors {
            assert_eq!(value.prune(&pruned_ty), None);
        }
    }

    #[test]
    fn zero_value() {
        let test_vectors = [
            (Final::unit(), Value::unit()),
            (Final::u8(), Value::u8(0)),
            (Final::u64(), Value::u64(0)),
            (
                Final::product(Final::u16(), Final::u32()),
                Value::product(Value::u16(0), Value::u32(0)),
            ),
            (
                Final::sum(Final::unit(), Final::u64()),
                Value::left(Value::unit(), Final::u64()),
            ),
            (
                Final::product(Final::unit(), Final::unit()),
                Value::product(Value::unit(), Value::unit()),
            ),
        ];

        for (ty, expected_default_value) in test_vectors {
            assert_eq!(Value::zero(&ty), expected_default_value);
        }
    }

    #[test]
    fn compact_round_trip() {
        let v = Value::u64(0xff00_00ff_ff00_00ff);
        let (bits, _) = v.iter_compact().collect_bits();
        let mut iter = BitIter::new(bits.into_iter());
        let new_v = Value::from_compact_bits(&mut iter, &v.ty).unwrap();
        assert_eq!(v, new_v);
    }

    #[test]
    fn padded_round_trip() {
        let v = Value::u64(0xff00_00ff_ff00_00ff);
        let (bits, _) = v.iter_padded().collect_bits();
        let mut iter = BitIter::new(bits.into_iter());
        let new_v = Value::from_padded_bits(&mut iter, &v.ty).unwrap();
        assert_eq!(v, new_v);
    }

    // --- Bug regression tests for right_shift_1 when new_bit=false ---
    //
    // right_shift_1(inner, bit_offset > 0, new_bit=false) is supposed to insert
    // a 0 (Left tag) at position bit_offset-1 by decrementing bit_offset.  The bug
    // is that it clones the Arc without clearing that position, so any pre-existing
    // 1-bit at bit_offset-1 corrupts the Left tag, causing the value to read as
    // Right instead of Left.
    //
    // The dirty bit originates from product sub-value extraction: as_product()
    // returns ValueRef children that share the parent buffer.  The right child has
    // bit_offset = parent.bit_offset + left_bit_width, so every 1-bit encoded by
    // the left component sits "before" the right child's logical start.  When that
    // right child is later wrapped in Left, right_shift_1 may see a stale 1-bit.

    /// Simplest direct construction: extract the right sub-value of a product
    /// whose left component has 1-bits, then wrap the sub-value in Left.
    ///
    /// `product(Right(u1(1)), Right(u1(0)))` encodes as [0xE0] at bit_offset=0:
    ///   bits 0..=1  = Right(u1(1))  = 1, 1   ← left component
    ///   bits 2..=3  = Right(u1(0))  = 1, 0   ← right component (bit_offset=2)
    ///
    /// Wrapping the right sub-value (bit_offset=2 in [0xE0]) in Left calls
    /// right_shift_1(false) which should clear bit 1.  The bug leaves bit 1 = 1,
    /// so the result reads as Right instead of Left.
    #[test]
    fn left_tag_not_corrupted_by_dirty_buffer() {
        let product_val = Value::product(
            Value::right(Final::u1(), Value::u1(1)), // encodes bits 0,1 as 1,1
            Value::right(Final::u1(), Value::u1(0)), // starts at bit_offset=2
        );

        let (_, right_sub) = product_val.as_product().unwrap();
        // right_sub shares [0xE0]; bit at position 1 (before its start) is 1.
        let right_sub_owned = right_sub.to_value();

        let left_val = Value::left(right_sub_owned, Final::unit());

        assert!(
            left_val.as_left().is_some(),
            "BUG: Left tag corrupted — right_shift_1(false) did not clear dirty bit; \
             value reads as Right instead of Left"
        );
        assert!(
            left_val.as_right().is_none(),
            "BUG: Left tag corrupted — as_right() should return None for a Left value"
        );
    }

    /// Verify the dirty-buffer bug at the bit level: the first padded bit of a
    /// Left value must always be 0 (the Left discriminant).
    #[test]
    fn left_tag_first_padded_bit_is_zero() {
        let product_val = Value::product(
            Value::right(Final::u1(), Value::u1(1)),
            Value::right(Final::u1(), Value::u1(0)),
        );

        let (_, right_sub) = product_val.as_product().unwrap();
        let left_val = Value::left(right_sub.to_value(), Final::unit());

        let first_bit = left_val.iter_padded().next();
        assert_eq!(
            first_bit,
            Some(false),
            "BUG: first padded bit of Left value is {:?}, expected Some(false)",
            first_bit
        );
    }

    /// End-to-end via prune: pruning can extract a sub-value via to_value() that
    /// shares the parent buffer and has 1-bits before its logical start.  When
    /// prune then calls Value::left() on it (Task::MakeLeft), the bug corrupts
    /// the Left tag.
    ///
    /// Original:  Left(product(Right(u1(1)), Right(u1(0))))
    ///            type: ((u1+u1) * (u1+u1)) + unit
    ///
    /// Pruned to: (unit * (u1+u1)) + unit
    ///
    /// The right component of the inner product keeps its type (u1+u1) unchanged,
    /// so prune returns it via to_value() — sharing the same dirty buffer.
    /// MakeLeft then wraps it, triggering the bug.
    #[test]
    fn prune_does_not_corrupt_left_tag_via_shared_buffer() {
        let original = Value::left(
            Value::product(
                Value::right(Final::u1(), Value::u1(1)),
                Value::right(Final::u1(), Value::u1(0)),
            ),
            Final::unit(),
        );

        // Shrink the left component: (u1+u1)*(u1+u1)  →  unit*(u1+u1).
        // The right sub-value of the product keeps type u1+u1 (matches exactly),
        // so prune returns it via to_value() with the dirty shared buffer intact.
        let pruned_ty = Final::sum(
            Final::product(Final::unit(), Final::sum(Final::u1(), Final::u1())),
            Final::unit(),
        );

        let pruned = original
            .prune(&pruned_ty)
            .expect("prune returned None; the target type is compatible");

        assert!(
            pruned.as_left().is_some(),
            "BUG: prune corrupted the Left tag — right_shift_1(false) did not clear \
             a dirty bit from the shared product buffer; value reads as Right"
        );
    }
}

#[cfg(bench)]
mod benches {
    use super::*;

    use crate::bit_encoding::{BitCollector as _, BitIter};

    use test::{black_box, Bencher};

    // Create values directly
    #[bench]
    fn bench_value_create_u64(bh: &mut Bencher) {
        bh.iter(|| black_box(Value::u64(0xff00_00ff_ff00_00ff)))
    }

    #[bench]
    fn bench_value_create_u2048(bh: &mut Bencher) {
        bh.iter(|| black_box(Value::from_byte_array([0xcd; 2048])));
    }

    #[bench]
    fn bench_value_create_deep_some(bh: &mut Bencher) {
        bh.iter(|| {
            let mut kilo = Value::from_byte_array([0xab; 1024]);
            for _ in 0..1000 {
                kilo = Value::some(kilo.shallow_clone());
            }
            black_box(kilo)
        });
    }

    #[bench]
    fn bench_value_create_64k(bh: &mut Bencher) {
        bh.iter(|| {
            let mut kilo = Value::from_byte_array([0xab; 1024]);
            for _ in 0..5 {
                kilo = Value::product(kilo.shallow_clone(), kilo.shallow_clone());
            }
            black_box(kilo)
        });
    }

    #[bench]
    fn bench_value_create_512_256(bh: &mut Bencher) {
        bh.iter(|| {
            // This is a common "slow" pattern in my Elements fuzzer; it is a deeply
            // nested product of 2^512 x 2^256, which is not a "primitive numeric
            // type" and therefore cannot be quickly decoded from compact bits by
            // simply matching to pre-computed types.
            //
            // However, this type is a giant product of bits. Its compact encoding
            // is therefore equal to its padded encoding and we should be able to
            // quickly decode it.
            let mut s512_256 = Value::product(
                Value::from_byte_array([0x12; 64]),
                Value::from_byte_array([0x23; 32]),
            );
            for _ in 0..5 {
                s512_256 = Value::product(s512_256.shallow_clone(), s512_256.shallow_clone());
            }
            black_box(s512_256);
        });
    }

    // Create values from padded bits
    fn padded_bits(v: &Value) -> (Vec<u8>, &Final) {
        let (bits, _) = v.iter_padded().collect_bits();
        (bits, v.ty.as_ref())
    }

    #[bench]
    fn bench_value_create_u64_padded(bh: &mut Bencher) {
        let v = Value::u64(0xff00_00ff_ff00_00ff);
        let (bits, ty) = padded_bits(&v);
        bh.iter(|| {
            black_box(Value::from_padded_bits(
                &mut BitIter::new(bits.iter().copied()),
                ty,
            ))
        })
    }

    #[bench]
    fn bench_value_create_u2048_padded(bh: &mut Bencher) {
        let v = Value::from_byte_array([0xcd; 2048]);
        let (bits, ty) = padded_bits(&v);
        bh.iter(|| {
            black_box(Value::from_padded_bits(
                &mut BitIter::new(bits.iter().copied()),
                ty,
            ))
        })
    }

    #[bench]
    fn bench_value_create_deep_some_padded(bh: &mut Bencher) {
        let mut kilo = Value::from_byte_array([0xab; 1024]);
        for _ in 0..1000 {
            kilo = Value::some(kilo.shallow_clone());
        }
        let (bits, ty) = padded_bits(&kilo);
        bh.iter(|| {
            black_box(Value::from_padded_bits(
                &mut BitIter::new(bits.iter().copied()),
                ty,
            ))
        })
    }

    #[bench]
    fn bench_value_create_64k_padded(bh: &mut Bencher) {
        let mut kilo = Value::from_byte_array([0xab; 1024]);
        for _ in 0..5 {
            kilo = Value::product(kilo.shallow_clone(), kilo.shallow_clone());
        }
        let (bits, ty) = padded_bits(&kilo);
        bh.iter(|| {
            black_box(Value::from_padded_bits(
                &mut BitIter::new(bits.iter().copied()),
                ty,
            ))
        })
    }

    #[bench]
    fn bench_value_create_512_256_padded(bh: &mut Bencher) {
        // See comment in bench_value_create_512_256 about this value
        let mut s512_256 = Value::product(
            Value::from_byte_array([0x12; 64]),
            Value::from_byte_array([0x23; 32]),
        );
        for _ in 0..5 {
            s512_256 = Value::product(s512_256.shallow_clone(), s512_256.shallow_clone());
        }
        let (bits, ty) = padded_bits(&s512_256);
        bh.iter(|| {
            black_box(Value::from_padded_bits(
                &mut BitIter::new(bits.iter().copied()),
                ty,
            ))
        })
    }

    // Create values from compact bits
    fn compact_bits(v: &Value) -> (Vec<u8>, &Final) {
        let (bits, _) = v.iter_compact().collect_bits();
        (bits, v.ty.as_ref())
    }

    #[bench]
    fn bench_value_create_u64_compact(bh: &mut Bencher) {
        let v = Value::u64(0xff00_00ff_ff00_00ff);
        let (bits, ty) = compact_bits(&v);
        bh.iter(|| {
            black_box(Value::from_compact_bits(
                &mut BitIter::new(bits.iter().copied()),
                ty,
            ))
        })
    }

    #[bench]
    fn bench_value_create_u2048_compact(bh: &mut Bencher) {
        let v = Value::from_byte_array([0xcd; 2048]);
        let (bits, ty) = compact_bits(&v);
        bh.iter(|| {
            black_box(Value::from_compact_bits(
                &mut BitIter::new(bits.iter().copied()),
                ty,
            ))
        })
    }

    #[bench]
    fn bench_value_create_deep_some_compact(bh: &mut Bencher) {
        let mut kilo = Value::from_byte_array([0xab; 1024]);
        for _ in 0..1000 {
            kilo = Value::some(kilo.shallow_clone());
        }
        let (bits, ty) = compact_bits(&kilo);
        bh.iter(|| {
            black_box(Value::from_compact_bits(
                &mut BitIter::new(bits.iter().copied()),
                ty,
            ))
        })
    }

    #[bench]
    fn bench_value_create_64k_compact(bh: &mut Bencher) {
        let mut kilo = Value::from_byte_array([0xab; 1024]);
        for _ in 0..5 {
            kilo = Value::product(kilo.shallow_clone(), kilo.shallow_clone());
        }
        let (bits, ty) = compact_bits(&kilo);
        bh.iter(|| {
            black_box(Value::from_compact_bits(
                &mut BitIter::new(bits.iter().copied()),
                ty,
            ))
        })
    }

    #[bench]
    fn bench_value_create_512_256_compact(bh: &mut Bencher) {
        // See comment in bench_value_create_512_256 about this value
        let mut s512_256 = Value::product(
            Value::from_byte_array([0x12; 64]),
            Value::from_byte_array([0x23; 32]),
        );
        for _ in 0..5 {
            s512_256 = Value::product(s512_256.shallow_clone(), s512_256.shallow_clone());
        }
        let (bits, ty) = compact_bits(&s512_256);
        bh.iter(|| {
            black_box(Value::from_compact_bits(
                &mut BitIter::new(bits.iter().copied()),
                ty,
            ))
        })
    }

    // Display values
    #[bench]
    fn bench_value_display_u64(bh: &mut Bencher) {
        let v = Value::u64(0xff00_00ff_ff00_00ff);
        bh.iter(|| black_box(format!("{}", v)))
    }

    #[bench]
    fn bench_value_display_u2024(bh: &mut Bencher) {
        let v = Value::from_byte_array([0xcd; 2048]);
        bh.iter(|| black_box(format!("{}", v)))
    }

    #[bench]
    fn bench_value_display_deep_some(bh: &mut Bencher) {
        let mut kilo = Value::from_byte_array([0xab; 1024]);
        for _ in 0..1000 {
            kilo = Value::some(kilo.shallow_clone());
        }
        bh.iter(|| black_box(format!("{}", kilo)))
    }

    #[bench]
    fn bench_value_display_64k(bh: &mut Bencher) {
        let mut kilo = Value::from_byte_array([0xab; 1024]);
        for _ in 0..5 {
            kilo = Value::product(kilo.shallow_clone(), kilo.shallow_clone());
        }
        bh.iter(|| black_box(format!("{}", kilo)))
    }

    #[bench]
    fn bench_value_display_512_256(bh: &mut Bencher) {
        // See comment in bench_value_create_512_256 about this value
        let mut s512_256 = Value::product(
            Value::from_byte_array([0x12; 64]),
            Value::from_byte_array([0x23; 32]),
        );
        for _ in 0..5 {
            s512_256 = Value::product(s512_256.shallow_clone(), s512_256.shallow_clone());
        }
        bh.iter(|| black_box(format!("{}", s512_256)))
    }

    // Display values
}