Module filter 

Precise single-column predicates and record batch filtering.

Modules

predicate 🔒

Predicate recognition, execution, and precompiled regular expressions.

timestamp 🔒

Timestamp-unit adaptation for predicates evaluated against historical SSTs.

Structs

MetadataFilter

Prunes candidates using known metadata columns. Missing columns and unsupported predicates cannot exclude a candidate. Complete SQL predicates must still run after pruning.

SimpleFilterEvaluator

Evaluates supported single-column predicates with precompiled regular expressions. Compound predicates preserve SQL null semantics until conversion to a scan mask.

Enums

TimestampUnitCast

The result of casting a SimpleFilterEvaluator to a different timestamp unit, preserving the predicate’s semantics on rows stored in that unit. See SimpleFilterEvaluator::cast_timestamp_unit.

Functions

batch_filter

Evaluate the predicate on the input RecordBatch, and return a new RecordBatch. Copy from datafusion::physical_plan::src::filter.rs

boolean_array_to_scan_mask 🔒

Converts nullable SQL predicate values to a scan mask, where NULL is false.

regexp_is_match_dictionary

Similar to regexp_is_match_scalar but for StringDictionaryArray. Iterates through dictionary keys to get string values and applies regex matching.

regexp_is_match_scalar

The same as arrow regexp_is_match_scalar() with pre-compiled regex. See https://github.com/apache/arrow-rs/blob/54.2.0/arrow-string/src/regexp.rs#L204-L246 for the implementation details.

timestamp_scalar_value

Builds a tz-naive timestamp literal for value in unit.