voonhous commented on code in PR #20513:
URL: https://github.com/apache/hudi/pull/20513#discussion_r4470686703
##########
hudi-spark-datasource/hudi-spark-common/src/main/scala/org/apache/spark/sql/hudi/dataskipping/rules/CastRule.scala:
##########
@@ -69,12 +63,31 @@ object HoodieSparkTypeUtils {
// This is conservative (for ex, double to float rounds
monotonically but is
// rejected), which only costs pruning opportunity, never
correctness
case (fromNum: NumericType, toNum: NumericType) =>
Cast.canUpCast(fromNum, toNum)
+ // NOTE: false < true maps to 0 < 1. Decimal targets are excluded since,
for ex,
+ // DECIMAL(1, 1) cannot hold 1 and overflows into NULL
+ case (BooleanType, ByteType | ShortType | IntegerType | LongType |
FloatType | DoubleType) => true
+
+ // Date and time
// NOTE: Timestamps cast to numbers through their epoch seconds, which
only fit bigint and
// floating-point targets; narrower integral targets wrap around
and decimals overflow
// into NULL in non-ANSI mode
case (TimestampType, LongType | FloatType | DoubleType) => true
- case (TimestampType, _: NumericType) => false
+ // NOTE: A date maps to the start of that day, which increases with the
date
+ case (DateType, TimestampType | TimestampNTZType) => true
+ // NOTE: Truncating to the (local) date is non-decreasing, also across a
DST fall-back since
+ // the date does not repeat
+ case (TimestampType | TimestampNTZType, DateType) => true
+ // NOTE: Not order-preserving in general (for ex, "2022-1-5" >
"2022-01-10" as strings, but parses
+ // to an earlier date), but kept so that expression indexes on
to_date/to_timestamp, which the
+ // optimizer rewrites into this cast, keep pruning. TODO #19481
+ case (StringType, DateType | TimestampType | TimestampNTZType) => true
Review Comment:
Confirmed: `extractQueryAndLiterals` strips the cast (and `ParseToDate` /
`ParseToTimestamp`), and `filterQueryWithRecordKey` rebuilds `=` / `IN` over
the bare column, so expression indexes compare their stored bounds and never
reach this arm. Dropped it in 4da99a212b5f: string to date/timestamp is now
rejected, the NOTE and the test comment are gone, and the test asserts `None`
for date, timestamp and timestamp_ntz. That also takes this item off
[#19481](https://github.com/apache/hudi/issues/19481).
--
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]
