[
https://issues.apache.org/jira/browse/IGNITE-23772?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18050836#comment-18050836
]
Vladimir Steshin edited comment on IGNITE-23772 at 1/9/26 11:51 AM:
--------------------------------------------------------------------
*Calcite* does:
{code:java}
DateStringLiteral -> RexLiteral#getValueAs(Integer) ->
DateString#getDaysSinceEpoch() -> DateTimeUtils#ymdToUnixDate(year, month, day)
-> return ymdToJulian(year, month, day) - EPOCH_JULIAN{code}
{code:java}
Result#getDate() -> return DateTimeUtils#unixDateToSqlDate(v.intValue(),
calendar);
{code}
{code:java}
LocalDate localDate = LocalDate.ofEpochDay(-141441);
java.sql.Date sqlDate = java.sql.Date.valueOf(localDate); {code}
*We* do:
{code:java}
DateStringLiteral -> RexLiteral#getValueAs(Integer) ->
DateString#getDaysSinceEpoch() -> DateTimeUtils#ymdToUnixDate(year, month, day)
-> return ymdToJulian(year, month, day) - EPOCH_JULIAN{code}
{code:java}
TypeUtils#fromInternal() -> return new java.sql.Date(fromLocalTs(ctx, -141441 *
DateTimeUtils.MILLIS_PER_DAY));
{code}
Also we have `DateValueUtils` with another dates convertation. Probably, we
should revise our `date-to-int/long` convertation and use the new java dates
system and/or Calcite's convertation. And keep it in a single place (without
{_}DateValueUtils{_})
was (Author: vladsz83):
*Calcite* does:
{code:java}
DateStringLiteral -> RexLiteral#getValueAs(Integer) ->
DateString#getDaysSinceEpoch() -> DateTimeUtils#ymdToUnixDate(year, month, day)
-> `return ymdToJulian(year, month, day) - EPOCH_JULIAN`{code}
{code:java}
Result#getDate() -> return DateTimeUtils#unixDateToSqlDate(v.intValue(),
calendar);
{code}
{code:java}
LocalDate localDate = LocalDate.ofEpochDay(-141441);
java.sql.Date sqlDate = java.sql.Date.valueOf(localDate); {code}
*We* do:
{code:java}
DateStringLiteral -> RexLiteral#getValueAs(Integer) ->
DateString#getDaysSinceEpoch() -> DateTimeUtils#ymdToUnixDate(year, month, day)
-> `ymdToJulian(year, month, day) - EPOCH_JULIAN`{code}
{code:java}
TypeUtils#fromInternal() -> `return new java.sql.Date(fromLocalTs(ctx, -141441
* DateTimeUtils.MILLIS_PER_DAY));`
{code}
Also we have `DateValueUtils` with another dates convertation. Probably, we
should revise our `date-to-int/long` convertation and use the new java dates
system and/or Calcite's convertation. And keep it in a single place (without
`{_}DateValueUtils`{_})
> Calcite. Incorrect processing of old date literals.
> ---------------------------------------------------
>
> Key: IGNITE-23772
> URL: https://issues.apache.org/jira/browse/IGNITE-23772
> Project: Ignite
> Issue Type: Bug
> Affects Versions: 2.14
> Reporter: Vladimir Steshin
> Assignee: Vladimir Steshin
> Priority: Minor
> Labels: calcite, ignite-2, ise
>
> Consider:
> *1)* {code:java}
> assertQuery("SELECT DATE '1582-10-05' + INTERVAL 1
> DAYS").returns(Date.valueOf("1582-10-06")).check();
> ---
> assertQuery("SELECT TIMESTAMP '1582-10-05 00:00:00' + INTERVAL 1
> DAYS").returns(Timestamp.valueOf("1582-10-06 00:00:00")).check();
> {code}
> Result:
> {code:java}
> class org.apache.ignite.internal.processors.query.IgniteSQLException: Failed
> to validate query. From line 1, column 25 to line 1, column 41: Illegal DATE
> literal '1582-10-05': not in format 'yyyy-MM-dd'
> ---
> class org.apache.ignite.internal.processors.query.IgniteSQLException: Failed
> to validate query. From line 1, column 25 to line 1, column 55: Illegal
> TIMESTAMP literal '1582-10-05 00:00:00': not in format 'yyyy-MM-dd HH:mm:ss'
> {code}
> *2)* {code:java}
> assertQuery("SELECT DATE '1582-10-01' + INTERVAL 1
> DAYS").returns(Date.valueOf("1582-10-02")).check();
> ---
> assertQuery("SELECT TIMESTAMP '1582-10-01 00:00:00' + INTERVAL 1
> DAYS").returns(Timestamp.valueOf("1582-10-02 00:00:00")).check();
> {code}
> Result:
> Doesn't fail, but but the result is incorrect.
> {code:java}
> java.lang.AssertionError: Collections are not equal (position 0):
> Expected: [[1582-10-02]]
> Actual: [[1582-09-22]]
> ---
> java.lang.AssertionError: Collections are not equal (position 0):
> Expected: [[1582-10-02 00:00:00.0]]
> Actual: [[1582-09-22 00:00:00.0]]
> {code}
> *3)* OK
> {code:java}
> assertQuery("SELECT DATE '1583-10-05' + INTERVAL 1
> DAYS").returns(Date.valueOf("1583-10-06")).check();
> ---
> assertQuery("SELECT TIMESTAMP '1583-10-01 00:00:00' + INTERVAL 1
> DAYS").returns(Timestamp.valueOf("1583-10-02 00:00:00")).check();
> {code}
> The problem is in variation of date conversion in java.util.Date (used in
> _TypeUtils#fromInterna()_). There is a threshold year 1582 (october).
> https://github.com/AdoptOpenJDK/openjdk-jdk11/blob/19fb8f93c59dfd791f62d41f332db9e306bc1422/src/java.base/share/classes/java/util/Date.java#L1238
> https://github.com/AdoptOpenJDK/openjdk-jdk11/blob/19fb8f93c59dfd791f62d41f332db9e306bc1422/src/java.base/share/classes/java/util/Date.java#L1279
> When Avatica always does trivial calculation using days, milliseconds related
> to 1970-01-01.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
