Recently upgraded, and we ran into failures because references to vtables (via opcode VOpen's p4 column) changed from vtab:module_ptr:vtable_ptr to just vtab:vtable_ptr.
What was the intent behind this change? How can code one determine the module the vtable is coming from now? This is not a showstopper for us, because we have a single module and we were just asserting it matched our "singleton" module, but I definitely missed that in the release notes :) I understand that https://www.sqlite.org/opcode.html doesn't explicitly say what VOpen's p4 column will contain, so I guess one could argue this is undocumented behavior we should not rely on, but the alternatives of query parsing, w/o an SQLite-provided AST of its queries would be even worse IMHO. Just trying to get clarification on this. Thanks, --DD /*! * \brief Introspects the pcode (compiled query) of a view. * * Uses an EXPLAIN SQL statement to list all SQLite VDBE instructions a simple * query accessing the view compiles into, and looks for \em TableLock and * \em VOpen opcodes, gathering their p4 arguments. In the case of VOpen, uses * a virtual table specific introspection function. * * \param view_name a SQL view name. * \return the list of table or virtual tables accessed by that view. * * \sa VirtualSchema::find_by_pcode() * * \note this type of introspection should be much more reliable than trying * to parse the SQL definition of the view, given SQL's complex grammar. */
