asolimando commented on code in PR #22718:
URL: https://github.com/apache/datafusion/pull/22718#discussion_r3342681416
##########
datafusion/physical-plan/src/filter.rs:
##########
@@ -311,10 +311,7 @@ impl FilterExec {
}
/// Calculates `Statistics` for `FilterExec`, by applying selectivity
- /// (either default, or estimated) to input statistics.
- ///
- /// Equality predicates (`col = literal`) set NDV to `Exact(1)`, or
- /// `Exact(0)` when the predicate is contradictory (e.g. `a = 1 AND a =
2`).
+ /// (either default or estimated) to input statistics.
Review Comment:
The function is quite interesting and IMO the docstring could summarize a
little more what it does, especially as this is public
##########
datafusion/physical-plan/src/filter.rs:
##########
@@ -846,6 +850,35 @@ fn collect_equality_columns(predicate: &Arc<dyn
PhysicalExpr>) -> (HashSet<usize
(eq_values.into_keys().collect(), infeasible)
}
+/// Collects columns that cannot be NULL in any surviving row.
+///
+/// A filter keeps only rows where the predicate is TRUE. If a direct column
+/// operand appears in an AND conjunct whose operator returns NULL on NULL
+/// input, rows where that column is NULL cannot survive.
+///
+/// This analysis is conservative; for example, OR clauses are not considered
+/// null-rejecting; neither are indirect operands like `a + 1 < 10`.
Review Comment:
Nit: this is incomplete and it's fine, but we might add support at least for
`IS NOT NULL` as it should be both easy and what people would intuitively
expect. wdyt?
##########
datafusion/physical-plan/src/filter.rs:
##########
@@ -861,15 +894,56 @@ fn interval_bound_to_precision(
}
}
-/// This function ensures that all bounds in the `ExprBoundaries` vector are
-/// converted to closed bounds. If a lower/upper bound is initially open, it
-/// is adjusted by using the next/previous value for its data type to convert
-/// it into a closed bound.
+/// Scales a column's `byte_size` by the estimated filter `selectivity`. An
+/// exact zero is preserved: an empty column stays exactly empty after
+/// filtering.
+fn scale_byte_size(byte_size: Precision<usize>, selectivity: f64) ->
Precision<usize> {
+ match byte_size {
+ Precision::Exact(0) => Precision::Exact(0),
+ byte_size => byte_size.with_estimated_selectivity(selectivity),
+ }
+}
+
+/// Caps a row-bounded column statistic (a null count or distinct count) at the
+/// filtered row estimate, since a column cannot have more nulls or distinct
+/// values than it has rows. Known counts are demoted to inexact because the
+/// filtered row count is itself an estimate.
+fn cap_at_rows(
Review Comment:
Not a must-have, but you might be interested in checking what is implemented
in Apache Calcite for the same case:
https://github.com/apache/calcite/blob/main/core/src/main/java/org/apache/calcite/rel/metadata/RelMdUtil.java#L317
This is a more refined estimator than the proposed one, but we can postpone
to a follow-up PR (this was on my radar already, I will get to it at some
point).
##########
datafusion/physical-plan/src/filter.rs:
##########
@@ -861,15 +894,56 @@ fn interval_bound_to_precision(
}
}
-/// This function ensures that all bounds in the `ExprBoundaries` vector are
-/// converted to closed bounds. If a lower/upper bound is initially open, it
-/// is adjusted by using the next/previous value for its data type to convert
-/// it into a closed bound.
+/// Scales a column's `byte_size` by the estimated filter `selectivity`. An
+/// exact zero is preserved: an empty column stays exactly empty after
+/// filtering.
+fn scale_byte_size(byte_size: Precision<usize>, selectivity: f64) ->
Precision<usize> {
+ match byte_size {
+ Precision::Exact(0) => Precision::Exact(0),
+ byte_size => byte_size.with_estimated_selectivity(selectivity),
+ }
+}
+
+/// Caps a row-bounded column statistic (a null count or distinct count) at the
+/// filtered row estimate, since a column cannot have more nulls or distinct
+/// values than it has rows. Known counts are demoted to inexact because the
+/// filtered row count is itself an estimate.
+fn cap_at_rows(
+ value: Precision<usize>,
+ filtered_num_rows: Precision<usize>,
+) -> Precision<usize> {
+ match filtered_num_rows {
+ Precision::Absent => value.to_inexact(),
+ rows => value.to_inexact().min(&rows),
+ }
+}
+
+/// Returns the NDV for a column constrained to one non-null value, such as
+/// `column = literal` or a singleton interval. The constraint gives NDV at
+/// most one; a zero-row output has no distinct values.
+///
+/// The caller is responsible for proving the singleton domain.
+fn distinct_count_for_singleton_domain(
+ filtered_num_rows: Precision<usize>,
+) -> Precision<usize> {
+ match filtered_num_rows {
+ Precision::Exact(0) | Precision::Inexact(0) => filtered_num_rows,
+ _ => Precision::Exact(1),
Review Comment:
Nit: what about returning `Inexact(1)` when matching `Absent`?
--
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]
