Closed soininen closed 2 years ago
The above commit removes most duplicates from the mapped_data
data structure in import_mappings
reducing the memory footprint by over 20% when importing 10000 objects with single parameter value into empty database without noticeable impact on import time.
I did some profiling on import_mappings
and managed to bring down the time spent in get_mapped_data()
by more than 50% in my test case.
It is possible to run out of memory when importing large datasets (see the discussion in Spine-project/Spine-Toolbox#1742). We should try to reduce the memory footprint of
import_mappings
.