Assuming a complex entity with many fields or properties, it is possible that there is a majority of use cases in which only a subset of these properties are actually used; there could be benefits in loading the remaining properties only on effective usage, and particularly storage benefits in not storing at all the tuple representing the property value.

Two aspects:

• avoid loading (null entries) and (unlikely needed entries)
• avoid storing null entries

Avoiding to load null entries is beneficial only in case there is no co-location (OGM-304) enabled, and would require some tracking anyway to store the information about properties which are defined vs. those which are not, so this aspect might not be too compelling.