AbsaOSS
Add ability to configure how Spark handles dates in parquet files.
Background
With Spark 3 new option were added how to work with dates pre 1900 in parquet files
The settings are:
spark.sql.parquet.datetimeRebaseModeInRead
spark.sql.parquet.datetimeRebaseModeInWrite
spark.sql.parquet.int96RebaseModeInRead
spark.sql.parquet.int96RebaseModeInWrite
Feature
Allow setting of the options for Enceladus jobs
### Tasks
- [ ] ~Add command line options to be able to set the **read** options. Set a default behavior either to `EXCEPTION` or `LEGACY`.~
- [ ] ~Modify the helper scripts to recognize these settings~
- [ ] ~Add an `reference.conf`/`application.conf` setting to be applied to write options. The default should be `LEGACY`~
- [ ] Modify the helper scripts to be able to easily send the Spark settings into the `spark submit` - the defaults remain the same as described above
To discuss
- The command line option names
- The command line defaults
- The write configuration names
This behaviour can be reached by adding:
--conf spark.sql.parquet.datetimeRebaseModeInRead=LEGACY --conf spark.sql.parquet.datetimeRebaseModeInWrite=LEGACY
into spark job json file call "spark-submit": "spark-submit --num-executors 2 --executor-memory 2G --deploy-mode client --conf spark.sql.parquet.datetimeRebaseModeInRead=LEGACY --conf spark.sql.parquet.datetimeRebaseModeInWrite=LEGACY",
No code changes in Enceladus are needed. See example usage in json file.
This behaviour can be reached by adding:
--conf spark.sql.parquet.datetimeRebaseModeInRead=LEGACY --conf spark.sql.parquet.datetimeRebaseModeInWrite=LEGACYinto spark job json file call"spark-submit": "spark-submit --num-executors 2 --executor-memory 2G --deploy-mode client --conf spark.sql.parquet.datetimeRebaseModeInRead=LEGACY --conf spark.sql.parquet.datetimeRebaseModeInWrite=LEGACY",No code changes in Enceladus are needed. See example usage in json file.
Great finding and solution. So only the Helper scripts needs to be enhanced.