Struct DataChunk 

pub struct DataChunk {
    pub fields: Vec<Arc<dyn Array>>,
    pub field_types: Vec<PhysicalTypeID>,
    pub size: usize,
    pub field_names: Vec<String>,
    pub sel_vector: Option<SelectionVector>,
}

A contiguous batch of columnar data — the fundamental unit of data exchange between physical operators.

Each DataChunk holds zero or more Arrow ArrayRef columns of equal length. Operators consume input chunks and produce output chunks, forming a pipeline.

Selection vectors

When sel_vector is set, only the rows indicated by the selection are “active”. Call active_rows() or iter_rows() to respect the selection; do not iterate 0..size directly.

Example

let chunk = DataChunk::new(fields, field_types).with_names(names);
for row in chunk.iter_rows() {
    let val = chunk.get_value(0, row); // column 0, current row
}

Fields

fields: Vec<Arc<dyn Array>>

field_types: Vec<PhysicalTypeID>

size: usize

field_names: Vec<String>

sel_vector: Option<SelectionVector>

Optional selection vector for zero-copy filtering. When set, only the rows at sel_vector.indices[0..sel_vector.size] are considered active. Operators should iterate using the sel_vector rather than size for correct results.

Implementations

impl DataChunk

pub fn new( fields: Vec<Arc<dyn Array>>, field_types: Vec<PhysicalTypeID>, ) -> DataChunk

Create a new DataChunk from Arrow arrays and their physical types.

The chunk size is set to the length of the first field. All fields must have the same length.

pub fn from_legacy(legacy_fields: Vec<LegacyValueVector>) -> DataChunk

Construct a DataChunk from legacy vectors (for backward-compatibility in benchmarks)

pub fn with_names(self, names: Vec<String>) -> DataChunk

Attach column names to this chunk (builder pattern).

pub fn with_sel(self, sel: SelectionVector) -> DataChunk

Attach a selection vector to this chunk.

pub fn field(&self, idx: usize) -> &Arc<dyn Array>

Return a reference to the Arrow array at the given column index.

pub fn num_fields(&self) -> usize

Return the number of columns in this chunk.

pub fn resize(&mut self, new_size: usize)

Resize the chunk (and all its fields) to the given number of rows.

pub fn active_rows(&self) -> usize

Return the number of active rows, taking the selection vector into account.

pub fn iter_rows(&self) -> RowIter<'_>

Iterate over active row indices. When a selection vector is present, yields the selected indices. Otherwise, yields 0..size.

pub fn materialize(&mut self)

Materialize the selection vector into physical data. After calling this, sel_vector is None and only the selected rows remain in the data buffers. This is a no-op if sel_vector is already None.

pub fn is_null(&self, field_idx: usize, row_idx: usize) -> bool

Return true if the value at (field_idx, row_idx) is null.

pub fn get_i64(&self, field_idx: usize, row_idx: usize) -> Option<i64>

Read an i64 value from column field_idx at row_idx, or None if null.

Returns None if the column is not physically an Int64 array (e.g. a UInt64 column) — callers reading UInt64 columns must use get_u64.

pub fn get_u64(&self, field_idx: usize, row_idx: usize) -> Option<u64>

Read a u64 value from column field_idx at row_idx, or None if null.

pub fn get_i32(&self, field_idx: usize, row_idx: usize) -> Option<i32>

Read an i32 value from column field_idx at row_idx, or None if null.

pub fn get_f64(&self, field_idx: usize, row_idx: usize) -> Option<f64>

Read an f64 value from column field_idx at row_idx, or None if null.

pub fn get_f32(&self, field_idx: usize, row_idx: usize) -> Option<f32>

Read an f32 value from column field_idx at row_idx, or None if null.

pub fn get_bool(&self, field_idx: usize, row_idx: usize) -> Option<bool>

Read a bool value from column field_idx at row_idx, or None if null.

pub fn get_string(&self, field_idx: usize, row_idx: usize) -> Option<&str>

Read a string slice from column field_idx at row_idx, or None if null.

pub fn get_value(&self, field_idx: usize, row_idx: usize) -> Option<Value>

Read a value from column field_idx at row_idx, returning a boxed Value.

Returns None if the cell is null. The physical type of the column determines which Value variant is produced.

Trait Implementations

impl Clone for DataChunk

fn clone(&self) -> DataChunk

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for DataChunk

fn fmt(&self, f: &mut Formatter<'_>) -> Result<(), Error>

Formats the value using the given formatter.