pond-ts API Reference (core)
    Preparing search index...

    Type Alias ColumnStorage

    ColumnStorage: "packed" | "chunked"

    Secondary discriminator: how a column physically stores its data.

    • 'packed' — a single flat typed-array buffer (or dictionary + indices for StringColumn, or array of cells for ArrayColumn). Reducers and other hot-path callers can dereference the kind- specific field (e.g. Float64Column._values) directly after narrowing on kind and storage === 'packed'.
    • 'chunked' — a sequence of packed chunks concatenated logically via chunkOffsets. Reads/scans route through chunk lookup; the per-chunk hot-path fields are not surfaced at the top level. Callers that need a flat buffer call materialize first.

    The same kind covers both: e.g. Float64Column and ChunkedFloat64Column both have kind: 'number'. Adding the storage discriminator at sub-step 1g lets the Column union widen to include chunked variants without breaking kind-based narrowing in code that doesn't touch hot-path fields.