Hive on Spark list all partitions for specific hive table and adding a partition

5
votes

I am using spark 2.0 and I was wondering ,Is it possible to list all files for specific hive table? If so, I can incrementally update those files directly using spark sc.textFile("file.orc") . How can I add a new partition to hive table? is there any api on the hive metastore that I can use from spark?

Is there any way to get the internal hive function that map dataframe row => partition_path

my main reasoning is incremental updates for a table. right now the only way I have figured out is FULL OUTER JOIN SQL +SaveMode.Overwrite, which is not so efficient because he will overwrite all the table while my main interest is incremental updates for some specific partitions/adding new partition

EDIT from what I have saw on the HDFS, when SaveMode.Overwrite spark will emit the table definition i.e CREATE TABLE my_table .... PARTITION BY (month,..). spark is putting all files under the $HIVE/my_table and not under $HIVE/my_table/month/... which means he is not partitioning the data. when I wrote df.write.partitionBy(...).mode(Overwrite).saveAsTable("my_table") I have saw on hdfs that it is correct. I have used SaveMode.Overwrite because I am updating records and not appending data.

I load data using spark.table("my_table") which means spark lazily load the table which is a problem since I don't want to load all the table just part of if.

for the question:

1.Does spark going to shuffle the data because I have used partitionBy() ,or he compares current partition and if its the same he will not shuffle the data.

2.Does spark smart enough to use partition pruning when mutating part from the data i.e just for specific month/year, and apply that change instead of loading all the data? (FULL OUTER JOIN is basically operation that scan all the table)

2
apache-sparkhive

2 Answers

8
votes

Adding partitions:

Adding partition from spark can be done with partitionBy provided in DataFrameWriter for non-streamed or with DataStreamWriter for streamed data.

public DataFrameWriter<T> partitionBy(scala.collection.Seq<String> colNames)

so if you want to partition data by year and month spark will save the data to folder like:

year=2016/month=01/
year=2016/month=02/

You have mentioned orc - you can use saving as a orc format with:

df.write.partitionBy('year', 'month').format("orc").save(path)

but you can easily insert into hive table like:

df.write.partitionBy('year', 'month').insertInto(String tableName)

Getting all partitions:

Spark sql is based on hive query language so you can use SHOW PARTITIONS to get list of partitions in the specific table.

sparkSession.sql("SHOW PARTITIONS partitionedHiveTable")

Just make sure you have .enableHiveSupport() when you are creating session with SparkSessionBuilder and also make sure whether you have hive-conf.xml etc. configured properly

1
votes

In spark using scala, we can use the catalog to get partitions: 

spark.catalog.listColumns(<databasename>, <tablename>)
.filter($"isPartition" === true)
.select($"name").collect()