Skip to content

Binary-accepting encoders (Fsst/VarBinView/Zstd) assume String[] internally #352

Description

@dfa1

`FsstEncodingEncoder`, `VarBinViewEncodingEncoder`, and `ZstdEncodingEncoder` all declare `accepts(DType.Binary) == true` but their `encode()`/`encodeCascade()` unconditionally cast `data` to `String[]` — a `DType.Binary` column (raw bytes, e.g. an embedded audio/image blob) throws `ClassCastException` if the cascade competition picks one of them.

Found while adding nested Parquet import (Raincloud's `waxal-dagbani-asr-test` `audio.bytes` field). Fixed for now by narrowing each encoder's `accepts()` to `Utf8` only, so `Binary` columns fall back to plain `vortex.varbin` storage (already byte-safe) — no compression competition for Binary yet.

Follow-up: give each of these three encoders a real `byte[][]` path (mirroring `VarBinEncodingEncoder`'s fix) so Binary columns get the same compression competition Utf8 does.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions