[CALCITE-6252] BigQuery FORMAT_DATE uses the wrong calendar for Julian dates #3680
Conversation
| // The Calendar and SimpleDateFormatter do not seem to give correct results | ||
| // for the day of the week prior to the Julian to Gregorian date change. | ||
| // So we resort to using a LocalDate representation. | ||
| LocalDate ld = |
There was a problem hiding this comment.
Can use java.sql.Date directly?
((java.sql.Date) date).toLocalDate()
There was a problem hiding this comment.
I didn't know this trick, I will try it.
There was a problem hiding this comment.
it doesn't work, the subclass relationship goes the other way. sql.date extends util.date.
The calendar-based conversion seems to do the right thing and it's not too complicated.
| final DateFormat eeeFormat = new SimpleDateFormat(DY.javaFmt, Locale.ROOT); | ||
| final DateFormat mmmFormat = new SimpleDateFormat(MON.javaFmt, Locale.ROOT); | ||
| final DateFormat mmmmFormat = new SimpleDateFormat(MONTH.javaFmt, Locale.ROOT); | ||
| final DateFormat eeeFormat = new SimpleDateFormat(DY.javaFmt, Locale.US); |
There was a problem hiding this comment.
SQL requires the output to be in English. Using Locale.ROOT gives the wrong result for the eeeeFormat at least (Mon instead of Monday).
| @@ -281,9 +301,9 @@ static Work get() { | |||
|
|
|||
| /** Uses Locale.US instead of Locale.ROOT to fix formatting in Java 11 */ | |||
There was a problem hiding this comment.
Hello, Should we change or add comments here?
There was a problem hiding this comment.
There are only Locale.ROOT comments here
mihaibudiu
force-pushed
the
issue6252
branch
from
b337021
to
9eb8508
Compare
…n dates Signed-off-by: Mihai Budiu <mbudiu@feldera.com>
mihaibudiu
force-pushed
the
issue6252
branch
2 times, most recently
from
b202d5a
to
c3eeec9
Compare
|
This was much trickier to fix than I expected.
I think that it's because of some bugs in the Java Calendar/SimpleDateFormatter implementation.
Using the LocalDate class provides correct results.
I have also changed the return type of FORMAT_DATE from VARCHAR(2000) to VARCHAR, because it can really return results longer than 2000 characters, if the format string is long.