PySpark - 将数据帧写入 Hive 表

Question

I have an empty Hive table. I have 18 jobs that I am running, and each one could have a data frame that I would need to add into the Hive table with a parquet file.

What I have is something like this:

df2.write.parquet(SOME_HDFS_DIR/my_table_dir)

But this doesn't seem quite right. Do I have to add some .parquet file name and keep appending it each time? I have seen some syntax is Scala but not Python.

Answer 1

df.write.parquet will overwrite the parquet files of the location, but with the option,

df.write.mode('append').parquet('path')

then it will create a new parquet file to the path and so you can read the data from the table.