You might be able to create shim and build in profile Like https://github.com/graphframes/graphframes/tree/branch-0.5/src/main
On Thu, Nov 12, 2020 at 3:23 PM Jia Yu <ji...@apache.org> wrote: > In order to support Spark 2.4, Sedona needs to use different logic for SQL > aggregation functions. I am not sure if this could be achieved by using > different profiles. > > Sedona for Spark 3.0: > https://github.com/apache/incubator-sedona/blob/master/sql/src/main/scala/org/apache/spark/sql/sedona_sql/expressions/AggregateFunctions.scala > Sedona for Spark 2.4: > https://github.com/apache/incubator-sedona/blob/spark-2.3-2.4/sql/src/main/scala/org/apache/spark/sql/geosparksql/expressions/AggregateFunctions.scala > > On Thu, Nov 12, 2020 at 9:42 AM Mo Sarwat <themosar...@gmail.com> wrote: > >> Yes, I believe we still need to support Spark 2.4 since many Sedona users >> are still using it >> >> On Wed, Nov 11, 2020 at 11:21 PM Netanel Malka <netan...@sela.co.il> >> wrote: >> >> > What do you mean by compile target? >> > >> > ? >> > >> > Iv'e found Apache Zeppelin handle multiple Spark versions here using >> > profile for each Spark version: >> > >> > https://github.com/apache/zeppelin/blob/master/spark/pom.xml#L185 >> > >> > >> > Do you think this method is better? >> > >> > >> > >> > Netanel Malka, >> > Big Data Consultant >> > [Description: Description: Description: Description: >> > cid:image001.jpg@01C85203.36A2AF30] >> > ________________________________ >> > From: Felix Cheung <felixche...@apache.org> >> > Sent: Thursday, November 12, 2020 04:05 >> > To: dev@sedona.apache.org >> > Cc: Jinxuan Wu; Mohamed Sarwat; Netanel Malka; Paweł Kociński; Zongsi >> > Zhang; d...@sedona.incubator.apache.org >> > Subject: Re: Support Spark 2.4 in Sedona 1.0 >> > >> > I am not sure it should be a branch? It is common to deal with this as a >> > compile target, not as a separate branch. A separate branch might have >> > difficulty to release? >> > >> > There are a few example in projects where they handle multiple Spark >> > target version like this. >> > >> > >> > On Wed, Nov 11, 2020 at 12:56 PM Jia Yu <ji...@apache.org<mailto: >> > ji...@apache.org>> wrote: >> > OK. I agree. I am gonna create a branch for spark-2.3/2.4. Regarding the >> > compiler used in each branch, >> > >> > For Sedona on Spark 3.0, I will compile it using Scala 2.12 >> > For Sedona on Spark 2.4, I will compile it using Scala 2.11. >> > >> > For the Java code in both branches, I will compile them using Java 1.8 >> > >> > Am I missing anything here? >> > >> > >> > On Wed, Nov 11, 2020 at 7:31 AM Netanel Malka <netan...@sela.co.il >> <mailto: >> > netan...@sela.co.il>> wrote: >> > >> > > Hi, >> > > >> > > I also think that we need to support 2.4. >> > > >> > > I saw that even Apache Spark still releases 2.4.x artifacts. (2.4.7 >> Sep >> > > 12, 2020) >> > > >> > > I also asked about it on *us...@spark.apache.org<mailto: >> > us...@spark.apache.org> <us...@spark.apache.org<mailto: >> > us...@spark.apache.org>>* >> > > : >> > > >> > > >> > > *Sean Owen (answered the question): * >> > > >> > > "I don't think there's an official EOL for Spark 2.4.x but would >> expect >> > > another maintenance release in the first half of 2021 at least. I'd >> also >> > > guess it wouldn't be maintained by 2022." >> > > >> > > >> > > ?BR, >> > > >> > > >> > > >> > > Netanel Malka, >> > > Big Data Consultant >> > > [image: Description: Description: Description: Description: >> > > cid:image001.jpg@01C85203.36A2AF30] >> > > ------------------------------ >> > > *From:* Paweł Kociński <pawel93kocin...@gmail.com<mailto: >> > pawel93kocin...@gmail.com>> >> > > *Sent:* Wednesday, November 11, 2020 00:29 >> > > *To:* Jia Yu >> > > *Cc:* dev@sedona.apache.org<mailto:dev@sedona.apache.org>; >> > d...@sedona.incubator.apache.org<mailto:d...@sedona.incubator.apache.org >> >; >> > Jinxuan Wu; >> > > Mohamed Sarwat; Netanel Malka; Zongsi Zhang >> > > *Subject:* Re: Support Spark 2.4 in Sedona 1.0 >> > > >> > > Hi Jia, >> > > I think we should support spark 2.4, a lot of users still use it. More >> > > than that I think more users still have jobs written in spark 2.4 than >> > > 3.0. We will use an additional branch for that use case ? I mean >> Spark >> > 2.4 >> > > with scala 2.12 is important one. >> > > Regards, >> > > Paweł >> > > >> > > pon., 9 lis 2020 o 20:44 Jia Yu <ji...@apache.org<mailto: >> > ji...@apache.org>> napisał(a): >> > > >> > >> Dear all, >> > >> >> > >> In Sedona 1.0, we definitely will support Spark 3.0. But I wonder >> > whether >> > >> we should support Spark 2.4. >> > >> >> > >> In order to support Spark 2.4, we need to do the following >> > >> >> > >> 1. Compile the source using Scala 2.11. Sedona master branch >> currently >> > is >> > >> compiled by Scala 2.12 and Java 1.8 >> > >> 2. For the Scala code of Sedona-SQL and Viz-SQL, I need to change the >> > (1) >> > >> UDF registration hook (2) the SQL aggregation function format >> > >> 3. In the future releases of Sedona, use git cherry-pick to pick >> > >> important features back to the Spark 2.4 branch. This is what I did >> in >> > >> GeoSpark to support Spark 2.1, 2.2, 2.3 >> > >> >> > >> GeoSpark 1.2.0 - 1.3.1 support Spark 2.4 already. We can simply >> leave it >> > >> that way and just support Spark 3.0. >> > >> >> > >> Do you think we should support Spark 2.4 in the future release? >> > >> >> > >> Thanks, >> > >> Jia Yu >> > >> >> > > >> > >> >