On Thu, 2005-04-21 at 14:10 -0700, Vladimir Vukicevic wrote: > Is there any reason that step() can't recreate the > bits of the VDBE that need to be updated itself in case of a > SQLITE_SCHEMA error, while preserving current parameter bindings? >
You make it sound as if only a few parameters in a VM change when the schema changes. Sometimes that is true. But other times, the entire VM structure changes. For example, if the schema change added a new index, the parser may decide to use that new index to speed a query rather than do a full table scan. Or if an index was being used by a query but that index was dropped, clearly the entire query would have to be completely restructured in order to avoid using the now missing index. So when the schema changes, it is not a matter of changes a few bits and pieces of the VM. The VM must be completely reconstructed. Its a complete do-over. -- D. Richard Hipp <[EMAIL PROTECTED]>
