ozankabak commented on code in PR #6982:
URL: https://github.com/apache/arrow-datafusion/pull/6982#discussion_r1268383993
##########
datafusion/physical-expr/src/intervals/interval_aritmetic.rs:
##########
@@ -1172,4 +1321,70 @@ mod tests {
let upper = 1.5;
capture_mode_change_f32((lower, upper), true, true);
}
+
+ #[test]
+ fn test_cardinality_of_intervals() -> Result<()> {
+ // In IEEE 754 standard for floating-point arithmetic, if we keep the
sign and exponent fields same,
+ // we can represent 4503599627370496 different numbers by changing the
mantissa
+ // (4503599627370496 = 2^52, since there are 52 bits in mantissa, and
2^23 = 8388608 for f32).
+ let distinct_f64 = 4503599627370496;
+ let distinct_f32 = 8388608;
+ let intervals = [
+ Interval::new(
+ IntervalBound::new(ScalarValue::from(0.25), false),
+ IntervalBound::new(ScalarValue::from(0.50), true),
+ ),
+ Interval::new(
+ IntervalBound::new(ScalarValue::from(0.5), false),
+ IntervalBound::new(ScalarValue::from(1.0), true),
+ ),
+ Interval::new(
+ IntervalBound::new(ScalarValue::from(1.0), false),
+ IntervalBound::new(ScalarValue::from(2.0), true),
+ ),
+ Interval::new(
+ IntervalBound::new(ScalarValue::from(32.0), false),
+ IntervalBound::new(ScalarValue::from(64.0), true),
+ ),
+ Interval::new(
+ IntervalBound::new(ScalarValue::from(-0.50), false),
+ IntervalBound::new(ScalarValue::from(-0.25), true),
+ ),
+ Interval::new(
+ IntervalBound::new(ScalarValue::from(-32.0), false),
+ IntervalBound::new(ScalarValue::from(-16.0), true),
+ ),
+ ];
+ for interval in intervals {
+ assert_eq!(interval.cardinality()?, distinct_f64);
+ }
+
+ let intervals = [
+ Interval::new(
+ IntervalBound::new(ScalarValue::from(0.25_f32), false),
+ IntervalBound::new(ScalarValue::from(0.50_f32), true),
+ ),
+ Interval::new(
+ IntervalBound::new(ScalarValue::from(-1_f32), false),
+ IntervalBound::new(ScalarValue::from(-0.5_f32), true),
+ ),
+ ];
+ for interval in intervals {
+ assert_eq!(interval.cardinality()?, distinct_f32);
+ }
+
+ let interval = Interval::new(
+ IntervalBound::new(ScalarValue::from(-0.0625), false),
+ IntervalBound::new(ScalarValue::from(0.0625), true),
+ );
+ assert_eq!(interval.cardinality()?, distinct_f64 * 2_048);
Review Comment:
FYI, in the above examples, values in the column "number of floating values"
are just illustrative, not the actual number of floating values in those ranges.
What it is trying to convey is that we halve selectivity because it is an
approximate measure, but we don't halve distinct count because it is a
conservative/bounding measure and we don't know how these distinct values are
distributed.
--
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]
