pepijnve commented on code in PR #18152:
URL: https://github.com/apache/datafusion/pull/18152#discussion_r2455898813
##########
datafusion/physical-expr/src/expressions/case.rs:
##########
@@ -122,6 +123,181 @@ fn is_cheap_and_infallible(expr: &Arc<dyn PhysicalExpr>)
-> bool {
expr.as_any().is::<Column>()
}
+/// Creates a [FilterPredicate] from a boolean array.
+fn create_filter(predicate: &BooleanArray, optimize: bool) -> FilterPredicate {
+ let mut filter_builder = FilterBuilder::new(predicate);
+ if optimize {
+ filter_builder = filter_builder.optimize();
+ }
+ filter_builder.build()
+}
+
+fn filter_record_batch(
+ record_batch: &RecordBatch,
+ filter: &FilterPredicate,
+) -> std::result::Result<RecordBatch, ArrowError> {
+ let filtered_columns = record_batch
+ .columns()
+ .iter()
+ .map(|a| filter_array(a, filter))
+ .collect::<std::result::Result<Vec<_>, _>>()?;
+ unsafe {
+ Ok(RecordBatch::new_unchecked(
+ record_batch.schema(),
+ filtered_columns,
+ filter.count(),
+ ))
+ }
+}
+
+#[inline(always)]
+fn filter_array(
+ array: &dyn Array,
+ filter: &FilterPredicate,
+) -> std::result::Result<ArrayRef, ArrowError> {
+ filter.filter(array)
+}
+
+struct ResultBuilder {
+ data_type: DataType,
+ // A Vec of partial results that should be merged.
`partial_result_indices` contains
+ // indexes into this vec.
+ partial_results: Vec<ArrayData>,
+ // Indicates per result row from which array in `partial_results` a value
should be taken.
+ // The indexes in this array are offset by +1. The special value 0
indicates null values.
+ partial_result_indices: Vec<usize>,
+ // An optional result that is the covering result for all rows.
+ // This is used as an optimisation to avoid the cost of merging when all
rows
+ // evaluate to the same case branch.
+ covering_result: Option<ColumnarValue>,
Review Comment:
I couldn't come up with a better word for this. It's in contrast with
partial and as in 'covering index'.
This gets filled in when one branch of the case expression matches all the
input rows from the record batch. In that case there's no need to do the
complex merge logic.
Open to suggestions for a better name.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
