Struct ColumnChunk 

pub struct ColumnChunk {
    pub update_info: Option<UpdateInfo>,
    pub stats: ColumnChunkStats,
    /* private fields */
}

An in-memory buffer that accumulates values before flushing to a Column.

Fields

update_info: Option<UpdateInfo>

Optional MVCC update version chain for this chunk. Tracks versioned updates to support snapshot isolation.

stats: ColumnChunkStats

Min/max stats for zone map predicate pushdown.

Implementations

impl ColumnChunk

pub fn new() -> Self

Create a new empty chunk with the default capacity (NODE_GROUP_SIZE).

pub fn with_capacity(capacity: usize) -> Self

Create a new empty chunk with a custom capacity.

pub fn enable_update_info(&mut self)

Enable MVCC update tracking for this chunk.

pub fn append(&mut self, value: Value)

Append a single value into the buffer.

Does not automatically flush when full — the caller should check is_full() and call flush_to_column() at the appropriate time.

pub fn set_value( &mut self, idx: usize, value: Value, ) -> Result<(), StorageError>

Set a value at a specific index (for in-place writes like INSERT UPDATE and DELETE nulling).

This is the non-versioned write path: it overwrites the cell without creating an MVCC version node, so snapshot readers see the new value immediately. Versioned writes must use Self::set_value_with_version so snapshot reads resolve the version chain correctly (P52.21). Returns an error if the index is out of bounds.

pub fn set_value_with_version( &mut self, idx: usize, value: Value, version: u64, ) -> Result<(), StorageError>

Set a value with MVCC version tracking.

Records the replaced value in the update version chain at version (a transaction commit timestamp) and overwrites the base cell. Snapshot readers with snapshot_ts < version keep seeing the replaced value; readers at or after version see the new value (P52.21). Returns an error if the index is out of bounds.

pub fn get_value_with_snapshot( &self, idx: usize, snapshot_ts: Option<u64>, _commit_history: &HashMap<u64, u64>, ) -> Option<&Value>

Get a value considering MVCC visibility at a given snapshot timestamp. If snapshot_ts is None, returns the latest value.

When a snapshot timestamp is provided and UpdateInfo has a version node for this row whose update is NOT yet visible at snapshot_ts (i.e. version > snapshot_ts), the snapshot predates the update and must see the replaced (old) value. Since this variant returns &Value it cannot deserialize the chain’s owned bytes, so callers that need a versioned read must use Self::get_value_owned_with_snapshot.

pub fn get_value_owned_with_snapshot( &self, idx: usize, snapshot_ts: Option<u64>, _commit_history: &HashMap<u64, u64>, ) -> Option<Value>

Get a value with MVCC snapshot isolation, returning an owned Value.

This variant properly handles version chain traversal by deserializing replaced values from the UpdateInfo chain. Returns the value visible at snapshot_ts, or None if the index is out of bounds.

pub fn num_values(&self) -> usize

Number of buffered values.

pub fn is_full(&self) -> bool

Whether the chunk has reached its capacity and should be flushed.

pub fn is_empty(&self) -> bool

Whether the chunk is empty.

pub fn as_slice(&self) -> &[Value]

Borrow the buffered values as a slice.

pub fn drain(&mut self) -> Vec<Value>

Drain all buffered values (leaves the chunk empty).

pub fn scan(&self, start: usize, count: usize) -> Vec<Value>

Scan a range of buffered values (inclusive start, exclusive end).

Panics if the range is out of bounds.

pub fn flush_to_column(&mut self, column: &mut Column) -> Result<usize>

Flush all buffered values into a Column via append_value, then clear the buffer.

Returns the number of values flushed.

pub fn flush_copy_to_column(&self, column: &mut Column) -> Result<usize>

Flush all buffered values into a Column, but keep the data in the buffer afterwards (for cases where the caller still needs it).

pub fn clear(&mut self)

Clear the buffer without flushing.

pub fn remaining(&self) -> usize

Remaining capacity before the chunk is full.

pub fn get(&self, index: usize) -> Option<&Value>

Access a single buffered value by index.

pub fn capacity(&self) -> usize

Capacity of this chunk.

pub fn to_arrow_array(&self, phys_type: PhysicalTypeID) -> ArrayRef

Convert buffered values directly into an Arrow array, skipping intermediate Vec<Vec<Value>> materialization.

This is the key optimization for the scan path: instead of cloning every Value into a Vec<Vec<Value>> and then building Arrow arrays from that, we read directly from self.values into Arrow builders.

Trait Implementations

impl Clone for ColumnChunk

fn clone(&self) -> ColumnChunk

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for ColumnChunk

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

Formats the value using the given formatter.

impl Default for ColumnChunk

fn default() -> Self

Returns the “default value” for a type.

impl From<Vec<Value>> for ColumnChunk

fn from(values: Vec<Value>) -> Self

Converts to this type from the input type.