在分区 hive 表中插入 spark Dataframe 而不覆盖数据

Question

I have one dataframe created from a partition table.

I need to insert this data frame in an already created partitioned hive table without overwriting the previous data.

I use partitionBy("columnname"),insertInto("hivetable") but it give me issue of partitionBy and intsertInto cant use at same time.

Answer 1

You can't do partitionBy with the insertInto operator. PartitionBy partitions the existing data into multiple hive partitions. The insertInto is used to insert data into a predefined partition.

Therefore, You can do something like this

spark.range(10)
.withColumn("p1", 'id % 2)
.write
.mode("overwrite")
.partitionBy("p1")
.saveAsTable("partitioned_table")

val insertIntoQ = sql("INSERT INTO TABLE 
partitioned_table PARTITION (p1 = 4) VALUES 41, 42")

If you require partitions to be added dynamically then you would need to set the hive.exec.dynamic.partition .

hiveContext.setConf("hive.exec.dynamic.partition", "true")

hiveContext.setConf("hive.exec.dynamic.partition.mode", "nonstrict")

I faced similar problem during data ingestion, I did something like

df.write().mode(SaveMode.Append).partitionBy("colname").saveAsTable("Table")

Answer 2

When you use insertInto there is no need to add PartitionBy or BucketBy in the code. This should be defined in the table creation request.