Skip to content

fix RLE encoding positive/negative float zeros #6510

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
a10y merged 1 commit into from
Feb 13, 2026
Merged

fix RLE encoding positive/negative float zeros #6510

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
29 changes: 25 additions & 4 deletions encodings/fastlanes/src/rle/array/rle_compress.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -6,6 +6,7 @@ use arrayref::array_ref;
use fastlanes::RLE;
use vortex_array::IntoArray;
use vortex_array::ToCanonical;
use vortex_array::arrays::NativeValue;
use vortex_array::arrays::PrimitiveArray;
use vortex_array::validity::Validity;
use vortex_array::vtable::ValidityHelper;
Expand All @@ -31,13 +32,14 @@ impl RLEArray {
fn rle_encode_typed<T>(array: &PrimitiveArray) -> VortexResult<RLEArray>
where
T: NativePType + RLE,
NativeValue<T>: RLE,
{
let values = array.as_slice::<T>();
let len = values.len();
let padded_len = len.next_multiple_of(FL_CHUNK_SIZE);

// Allocate capacity up to the next multiple of chunk size.
let mut values_buf = BufferMut::<T>::with_capacity(padded_len);
let mut values_buf = BufferMut::<NativeValue<T>>::with_capacity(padded_len);
let mut indices_buf = BufferMut::<u16>::with_capacity(padded_len);

// Pre-allocate for one offset per chunk.
Expand All @@ -50,8 +52,11 @@ where
let mut chunks = values.chunks_exact(FL_CHUNK_SIZE);

let mut process_chunk = |chunk_start_idx: usize, input: &[T; FL_CHUNK_SIZE]| {
// SAFETY: `MaybeUninit<T>` and `T` have the same layout.
let rle_vals: &mut [T] =
// SAFETY: NativeValue is repr(transparent)
let input: &[NativeValue<T>; FL_CHUNK_SIZE] = unsafe { std::mem::transmute(input) };

// SAFETY: `MaybeUninit<NativeValue<T>>` and `NativeValue<T>` have the same layout.
let rle_vals: &mut [NativeValue<T>] =
unsafe { std::mem::transmute(&mut values_uninit[value_count_acc..][..FL_CHUNK_SIZE]) };

// SAFETY: `MaybeUninit<u16>` and `u16` have the same layout.
Expand All @@ -62,7 +67,7 @@ where
// returned from `T::encode` are relative to the chunk.
values_idx_offsets.push(value_count_acc as u64);

let value_count = T::encode(
let value_count = NativeValue::<T>::encode(
input,
array_mut_ref![rle_vals, 0, FL_CHUNK_SIZE],
array_mut_ref![rle_idxs, 0, FL_CHUNK_SIZE],
Expand Down Expand Up @@ -92,6 +97,9 @@ where
indices_buf.set_len(padded_len);
}

// SAFETY: NativeValue<T> is repr(transparent) to T.
let values_buf = unsafe { values_buf.transmute::<T>().freeze() };

RLEArray::try_new(
values_buf.into_array(),
PrimitiveArray::new(indices_buf.freeze(), padded_validity(array)).into_array(),
Expand Down Expand Up @@ -254,4 +262,17 @@ mod tests {
let expected = PrimitiveArray::new(values, primitive.validity().clone());
assert_arrays_eq!(decoded, expected);
}

// Regression test: RLE compression properly supports decoding pos/neg zeros
// See <https://github.com/vortex-data/vortex/issues/6491>
#[rstest]
#[case(vec![f16::ZERO, f16::NEG_ZERO])]
#[case(vec![0f32, -0f32])]
#[case(vec![0f64, -0f64])]
fn test_float_zeros<T: NativePType + RLE>(#[case] values: Vec<T>) {
let primitive = PrimitiveArray::from_iter(values);
let rle = RLEArray::encode(&primitive).unwrap();
let decoded = rle.to_primitive();
assert_arrays_eq!(primitive, decoded);
}
}
Loading