Skip to content

DictEncoder::reset does not clear the encoder state #9692

Description

@mnshdw

What happened?

reset is documented as "Clear the encoder state to make it ready for a new round of decoding", but the implementations do not clear it properly.

  • BytesDictBuilder::reset keeps lookup and null_code
  • PrimitiveDictBuilder::reset keeps lookup, null_code and values

Steps to reproduce

BytesDictBuilder:

let session = vortex_array::array_session();
let mut ctx = session.create_execution_ctx();

let first = VarBinViewArray::from_iter_str(["one", "two"]).into_array();
let mut encoder = dict_encoder(&first, &UNCONSTRAINED);
encoder.encode(&first, &mut ctx).unwrap();

encoder.reset();

let second = VarBinViewArray::from_iter_str(["one", "three"]).into_array();
encoder.encode(&second, &mut ctx).unwrap(); // index out of bounds: the len is 0 but the index is 0

PrimitiveDictBuilder

let session = vortex_array::array_session();
let mut ctx = session.create_execution_ctx();

let first = PrimitiveArray::from_option_iter([Some(1i32), None, Some(3)]).into_array();
let mut encoder = dict_encoder(&first, &UNCONSTRAINED);
encoder.encode(&first, &mut ctx).unwrap();

encoder.reset();

let second = PrimitiveArray::from_option_iter([Some(4i32), Some(5)]).into_array();
let codes = encoder.encode(&second, &mut ctx).unwrap();
assert_eq!(codes.as_slice::<u32>(), &[0, 1]); // assertion `left == right` failed
                                              //   left: [3, 4]
                                              //  right: [0, 1]

Environment

  • Vortex develop
  • Rust 1.98.0

Additional context

No response

Metadata

Metadata

Assignees

No one assigned

    Labels

    bugA bug issue

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions