Module prefilter 

Prefilter framework for parquet reader.

Prefilter optimization reduces I/O by reading only a subset of columns first (the prefilter phase), applying filters to compute a refined row selection, then reading the remaining columns with the refined selection.

Structs

BulkFilterPlan 🔒

How the bulk-memtable read should apply each predicate.

CachedPrimaryKeyFilter 🔒

PrefilterContext 🔒

Context for prefiltering a row group.

PrefilterContextBuilder 🔒

Pre-built state for constructing PrefilterContext per row group.

PrefilterResult 🔒

Result of prefiltering a row group.

ReaderFilterPlan 🔒

How the parquet reader should apply each predicate.

Constants

PREFILTER_COLUMN_RATIO_THRESHOLD 🔒

PREFILTER_MIN_REMAINING_COLUMNS 🔒

Functions

apply_filters_to_batch 🔒

build_bulk_filter_plan 🔒

build_reader_filter_plan 🔒

Splits a query [Predicate] into a ReaderFilterPlan: predicates that can run during the prefilter pass (on a reduced projection, to compute a refined row selection) versus predicates that must run on the normal read path (alongside the full projection).

compute_projection_mask 🔒

Executes prefiltering on a row group.

execute_prefilter 🔒

matching_row_ranges_by_primary_key 🔒

prefilter_flat_batch_by_primary_key 🔒

Filters a flat-format record batch by primary key, returning only rows whose primary key matches the filter. Returns None if all rows are filtered out.

should_use_prefilter 🔒