Struct RowSampler 

pub struct RowSampler { /* private fields */ }

Applies a SamplingStrategy to a stream of rows, holding all the state the strategy needs across rows.

A fresh state per row — which is what calling a stateless helper amounts to — silently disables every stateful strategy, so engines must create one sampler per scan and keep it for the whole scan.

Implementations

impl RowSampler

pub fn new(strategy: &SamplingStrategy) -> Result<Self, DataProfilerError>

Build a sampler for strategy, validating that it can be applied.

Rejects here rather than mid-scan: a caller learns that a strategy is unusable before the source is read, not after a partial profile exists.

pub fn is_buffered(&self) -> bool

Whether rows must be handed to offer instead of being folded in as accept approves them.

pub fn is_noop(&self) -> bool

Whether this sampler can ever exclude a row.

pub fn accept(&mut self, row: RowView<'_>) -> bool

Decide whether a row passes the streaming filters.

Always call this, including on the buffered path, so the filters of a multi-stage strategy run before the fixed-size stage sees a row.

pub fn offer(&mut self, values: Vec<String>)

Hand a row that passed accept to the fixed-size stage.

Only meaningful when is_buffered is true.

pub fn take_sample(&mut self) -> Vec<Vec<String>>

The final sample from the fixed-size stage, ready to fold into the statistics. Empty for a purely streaming strategy.

pub fn iterated_rows(&self) -> usize

Rows the sampler has seen, sampled or not.

pub fn sampled_rows(&self) -> usize

Rows that ended up in the sample.

Trait Implementations

impl Debug for RowSampler

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

Formats the value using the given formatter.