using with text/csv import metadata.csv failed. quickLoader.groovy seems involved in the above problem with error message on Java heap.
The corpus is just one file 5,7 Mo.
Without the metadata.csv the import works fine.
This problem is due to the usage of DOM to inject the metadata in the XML version of the sources.
The solution is to use Stax
The problem will be solved in the next version of TXM (0.6)
Fixed in TXM 0.6