Ryan Johnson <ryan.johnson@...> writes: > So if I understand correctly, you start out with only C2, use C2 to > drive an expensive/slow computation that derives all other fields in the > row, and set C1 != NULL when the derivation completes? And you don't > want a trigger because that would pin the slow computation on the > inserting (application) thread rather than the batch-updating > (background) thread?
Yes. > In that case, your AUTOINCREMENT approach is probably the best you can do... There is also a dumb approach of copying entire DB to a new file, inserting/deleting/updating rows in this file, then renaming this file back. This requires an custom lock, otherwise is simple and fast. Alas, none of the above answered my original question "How to interrupt a long running update without roll back". But, guess, I will not get an answer. Thank you for the discussion. _______________________________________________ sqlite-users mailing list sqlite-users@sqlite.org http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
