I'm worried about the change from Millis to Nanos. It seems not so good. I think we (dremio) need to better understand impact. Praveen, maybe you could take a look at impact?
On Wed, Nov 14, 2018, 4:27 PM Bryan Cutler <cutl...@gmail.com wrote: > Hi all, > > I'm picking this back up again and have WIP pr at > https://github.com/apache/arrow/pull/2966. Please take a look at the new > APIs and see if they impact you downstream. In addition to the API changes > mentioned before by Li, there is also > > (4) IntervalDayVector now uses java.time.Duration, while IntervalYearVector > uses java.time.Period. I think Joda Period was basically a combination of > these two. > > Thanks, > Bryan > > On Thu, Jun 28, 2018 at 2:31 PM Li Jin <ice.xell...@gmail.com> wrote: > > > I did a grep of "joda" in the java vector codebase. The current reference > > to joda includes: > > > > (1) auto-generated reader class under arrow/vector/complex/impl and > > arrow/vector/complex/reader > > > > These class currently have API that returns joda classes, e.g.: > > > > NullableTimestampMicroHolderReaderImpl: > > > > // read friendly type > > @Override > > public LocalDateTime readLocalDateTime() { > > if (!isSet()) { > > return null; > > } > > > > LocalDateTime value = new LocalDateTime(this.holder.value); > > return value; > > } > > > > UnionReader: > > > > @Override > > public LocalDateTime readLocalDateTime() { > > return getReaderForIndex(idx()).readLocalDateTime(); > > } > > > > > > (2) Timestamp vector classes: > > These classes have API that returns joda classes, e.g.: > > > > TimestampMicroVector: > > > > /** > > * Same as {@link #get(int)}. > > * > > * @param index position of element > > * @return element at given index > > */ > > public LocalDateTime getObject(int index) { > > if (isSet(index) == 0) { > > return null; > > } else { > > /* value is truncated when converting microseconds to > > milliseconds in order to use DateTime type */ > > final long micros = valueBuffer.getLong(index * TYPE_WIDTH); > > final long millis = > > java.util.concurrent.TimeUnit.MICROSECONDS.toMillis(micros); > > final org.joda.time.LocalDateTime localDateTime = new > > org.joda.time.LocalDateTime(millis, > > org.joda.time.DateTimeZone.UTC); > > return localDateTime; > > } > > } > > > > > > (3) arrow/vector/util/DateUtility.java has some functions that uses joda > > time > > > > > > On Thu, Jun 28, 2018 at 2:19 PM, Jacques Nadeau <jacq...@apache.org> > > wrote: > > > > > Can we get a proposed listing of apis that would change? Definitely > will > > be > > > effort for people like us to rewrite all code to use different items. > > > > > > On Wed, Jun 27, 2018 at 10:54 AM, Li Jin <ice.xell...@gmail.com> > wrote: > > > > > > > Hi, > > > > > > > > There has been a recent pr 2171 for ARROW-2015 to replace Joda time > > with > > > > Java8 time. > > > > > > > > I think this change is good as we move toward newer Java version and > > > wonder > > > > if we should include this in the 0.10 release. > > > > > > > > The biggest concern is that this is a breaking change and could > impact > > > > downstream projects like Dremio. What do people think? > > > > > > > > Li > > > > > > > > > >
