Comment 1 for bug 856345

We can avoid instantiating a lot of java objects through jpype if we just reparse source model files every time we need to modify their values. The overhead is still manageable: parsing california_fault_model.xml (1478 sources, 1662739 bytes) takes 280 ms, parsing SEAsiaGSHAP.xml (398 sources, 571688 bytes) takes 185 ms and Myanmar.trial1.xml (44549 sources, 96477898 bytes) takes 8837 msec. Heavy files with many point sources are slow to parse, however, they are not likely to be used with complex logic trees.