So, if I understand correctly, then D-1 contains only yesterdays data and D-2 the data from the day before yesterday.
If that's true, one simple and efficient approach would be to perform a UNION of both tables, mapping the quantity information to a separate column each (with a constant 0 in the "missing" component).
Once this is done, an aggregation melts the records together and a calculated column can deliver the difference/delta value.
Let me know how you go with that.