It cannot be implemented this way because you can have multiple tables and CSV files in the same XML file. I am open to suggestions but I can't think of any that wouldn't be difficult to implement for little gain or wouldn't have undefined behaviour.

LinuxJedi

Benefit is clear.
You can load a table with custom defined columns without creating a job for every load.

Most of the cases xml files include only 1 table.

So, if only 1 table is included, so it can overwritten if
count(//Table) =1

Elsewhere -p parameter can include 2 optional parameter.
-p colxml <optional tablename> <optional schemaname>

If optional table schema name used, only the first match will be used, other will be ignored.

I would argue that in many cases it is used for loading multiple tables (and even multiple XML files). This either becomes complex to implement due to having to add an initial pre-parse stage or we have it do bad things in a common use case.

A "sed" command would do the same thing as this requested feature and could easily be wrapped up into a shell script along with cpimport.

Cannot be implemented as suggested. The -p and -l options cannot be used together.

I have updated https://mariadb.com/kb/en/library/columnstore-bulk-data-loading/ to note the behavior.