This has the disadvantage that one
1) Cannot set alternatives location for those jars (e.g. enterprise artifact repositories/https links)
2) One cannot add other additional libraries (e.g. external VTL functions)
I propose here to change simply the order
Firstly set the default location of the Trevas jars and then load the configuration files to allow then the user to define any custom spark.jars.
The kernel loads the default Spark configuration and then adds a default location of the Trevas library for Spark: https://github.com/InseeFrLab/Trevas-Jupyter/blob/main/src/main/java/fr/insee/trevas/jupyter/SparkUtils.java#L32
This has the disadvantage that one 1) Cannot set alternatives location for those jars (e.g. enterprise artifact repositories/https links) 2) One cannot add other additional libraries (e.g. external VTL functions)
I propose here to change simply the order Firstly set the default location of the Trevas jars and then load the configuration files to allow then the user to define any custom spark.jars.