Struct iceberg::scan::TableScanBuilder

pub struct TableScanBuilder<'a> { /* private fields */ }

Builder to create table scan.

Implementations

impl<'a> TableScanBuilder<'a>

pub fn with_batch_size(self, batch_size: Option<usize>) -> Self

Sets the desired size of batches in the response to something other than the default

pub fn with_case_sensitive(self, case_sensitive: bool) -> Self

Sets the scan’s case sensitivity

pub fn with_filter(self, predicate: Predicate) -> Self

Specifies a predicate to use as a filter

pub fn select_all(self) -> Self

Select all columns.

pub fn select_empty(self) -> Self

Select empty columns.

pub fn select( self, column_names: impl IntoIterator<Item = impl ToString>, ) -> Self

Select some columns of the table.

pub fn snapshot_id(self, snapshot_id: i64) -> Self

Set the snapshot to scan. When not set, it uses current snapshot.

pub fn with_concurrency_limit(self, limit: usize) -> Self

Sets the concurrency limit for both manifest files and manifest entries for this scan

pub fn with_data_file_concurrency_limit(self, limit: usize) -> Self

Sets the data file concurrency limit for this scan

pub fn with_manifest_entry_concurrency_limit(self, limit: usize) -> Self

Sets the manifest entry concurrency limit for this scan

pub fn with_row_group_filtering_enabled( self, row_group_filtering_enabled: bool, ) -> Self

Determines whether to enable row group filtering. When enabled, if a read is performed with a filter predicate, then the metadata for each row group in the parquet file is evaluated against the filter predicate and row groups that cant contain matching rows will be skipped entirely.

Defaults to enabled, as it generally improves performance or keeps it the same, with performance degradation unlikely.

pub fn with_row_selection_enabled(self, row_selection_enabled: bool) -> Self

Determines whether to enable row selection. When enabled, if a read is performed with a filter predicate, then (for row groups that have not been skipped) the page index for each row group in a parquet file is parsed and evaluated against the filter predicate to determine if ranges of rows within a row group can be skipped, based upon the page-level statistics for each column.

Defaults to being disabled. Enabling requires parsing the parquet page index, which can be slow enough that parsing the page index outweighs any gains from the reduced number of rows that need scanning. It is recommended to experiment with partitioning, sorting, row group size, page size, and page row limit Iceberg settings on the table being scanned in order to get the best performance from using row selection.

pub fn build(self) -> Result<TableScan>

Build the table scan.