We need to set num partitions just to cheat Exchange of Spark.
We need to set num partitions just to cheat Exchange of Spark. This partition is not used for actual scan operator which depends on the actual RDD. Spark ClusteredDistribution is pretty simplistic to consider numShufflePartitions for its partitioning scheme as Spark always uses shuffle. Ideally it should consider child Spark plans partitioner.
If the row is already present, it gets updated otherwise it gets inserted into the table represented by this relation
If the row is already present, it gets updated otherwise it gets inserted into the table represented by this relation
the rows to be upserted
number of rows upserted
If the row is already present, it gets updated otherwise it gets inserted into the table represented by this relation
If the row is already present, it gets updated otherwise it gets inserted into the table represented by this relation
the DataFrame to be upserted
number of rows upserted
A LogicalPlan implementation for an Snappy row table whose contents are retrieved using a JDBC URL or DataSource.