[
https://issues.apache.org/jira/browse/PHOENIX-6823?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17640716#comment-17640716
]
ASF GitHub Bot commented on PHOENIX-6823:
-----------------------------------------
stoty commented on code in PR #1530:
URL: https://github.com/apache/phoenix/pull/1530#discussion_r1034771915
##########
phoenix-core/src/test/java/org/apache/phoenix/expression/RoundFloorCeilExpressionsTest.java:
##########
@@ -688,4 +720,1207 @@ public void testFloorDateExpressionForWeek() throws
Exception {
Date resultDate = (Date)result;
assertEquals(DateUtil.parseDate("2016-01-04 00:00:00"), resultDate);
}
+
+ private RoundDateExpression getRoundMsExpression(String s, TimeUnit u, int
m) throws SQLException {
+ return
(RoundDateExpression)RoundDateExpression.create(LiteralExpression.newConstant(s),
u, m );
+ }
+
+ // The three tests below are backported from PHOENIX-5066.
+ // When PHOENIX-5066 lands, these can be removed as redundant.
+
+ @Test
+ public void testRoundingGMT() throws SQLException {
+ // We operate on Instants for time units up to Days, simply counting
millis
+
+ RoundDateExpression oddWholeSecondExp =
+ getRoundMsExpression("2022-11-11 11:11:11", TimeUnit.SECOND,
1);
+ java.sql.Timestamp oddWholeSecond =
+ new java.sql.Timestamp(DateUtil.parseDate("2022-11-11
11:11:11").getTime());
+ long lowerBoundaryOddWholeSecond = oddWholeSecond.getTime() - HALF_SEC;
+ long upperBoundaryOddWholeSecond = oddWholeSecond.getTime() + HALF_SEC
- 1;
+ assertEquals(lowerBoundaryOddWholeSecond,
+ oddWholeSecondExp.rangeLower(oddWholeSecond.getTime()));
+ assertEquals(upperBoundaryOddWholeSecond,
+ oddWholeSecondExp.rangeUpper(oddWholeSecond.getTime()));
+ assertEquals(oddWholeSecond,
+ new
java.sql.Timestamp(oddWholeSecondExp.roundTime(lowerBoundaryOddWholeSecond)));
+ assertNotEquals(oddWholeSecond,
+ new
java.sql.Timestamp(oddWholeSecondExp.roundTime(lowerBoundaryOddWholeSecond -
1)));
+ assertEquals(oddWholeSecond,
+ new
java.sql.Timestamp(oddWholeSecondExp.roundTime(upperBoundaryOddWholeSecond)));
+ assertNotEquals(oddWholeSecond,
+ new
java.sql.Timestamp(oddWholeSecondExp.roundTime(upperBoundaryOddWholeSecond +
1)));
+
Review Comment:
Yes, I was thinking about how save some copy-paste, but did not find a
solution I liked.
> calling Joda-based round() function on temporal PK field causes division by
> zero error
> --------------------------------------------------------------------------------------
>
> Key: PHOENIX-6823
> URL: https://issues.apache.org/jira/browse/PHOENIX-6823
> Project: Phoenix
> Issue Type: Bug
> Components: core
> Affects Versions: 5.2.0, 5.1.2
> Reporter: Istvan Toth
> Assignee: Istvan Toth
> Priority: Major
>
>
> {noformat}
> 0: jdbc:phoenix:> create table bubu (df date primary key);
> No rows affected (0.689 seconds)
> 0: jdbc:phoenix:> select * from bubu where round(df,'YEAR',1)=DATE
> '2022-01-01 00:00:00';
> java.lang.ArithmeticException: / by zero
> at
> org.apache.phoenix.expression.function.RoundDateExpression$1.getKeyRange(RoundDateExpression.java:272)
> ...{noformat}
> This is caused by getKeyRange not being updated when the Joda-based time
> units were added.
> Also, the keyrange pushdown logic is just wrong, even for the non-joda cases.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
