Struct SortBuilder 

pub struct SortBuilder { /* private fields */ }

Streaming sort index builder.

Create with SortBuilder::new, feed batches via SortBuilder::push, then call SortBuilder::finish to produce the sorted PermutationIndex.

Example

let mut builder = SortBuilder::new(vec![
    SortField::new(DataType::Int64),
    SortField::new_with_options(DataType::Utf8, SortOptions { descending: true, .. }),
])?;

for batch in record_batch_stream {
    builder.push(&batch, &[0, 2])?;  // sort by columns 0 and 2
}

let index = builder.finish()?;

Implementations

impl SortBuilder

pub fn new(fields: Vec<SortField>) -> Result<Self, IndexError>

Create a new builder for a sort with the given fields.

fields defines the sort key schema and direction. One SortField per sort column, in priority order (first = primary, second = tiebreaker, …).

Errors

• IndexError::EmptyColumns if fields is empty.
• IndexError::RowEncodingFailed if the RowConverter cannot be created.

pub fn push( &mut self, batch: &RecordBatch, sort_columns: &[usize], ) -> Result<(), IndexError>

Ingest one RecordBatch.

sort_columns are the column indices within batch that correspond to the SortFields passed to SortBuilder::new. Must have the same length.

The batch’s sort columns are encoded into fixed-width row keys; the batch itself is NOT retained.

Errors

• IndexError::LengthMismatch if sort_columns.len() != fields.len().
• IndexError::TooManyRows if cumulative rows exceed u32::MAX.
• IndexError::RowEncodingFailed on encoding failure.

pub fn rows_ingested(&self) -> u64

Total rows ingested so far.

pub fn finish(self) -> Result<PermutationIndex, IndexError>

Consume the builder, run parallel argsort, and produce a PermutationIndex.

Errors

• IndexError::EmptyColumns if no rows were pushed.
• [IndexError::MmapError] if mmap storage creation fails (when mmap feature is enabled).

Trait Implementations

impl Debug for SortBuilder

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

Formats the value using the given formatter.