On Mar 29, 2010, at 10:40 AM, Jay A. Kreibich wrote: > On Mon, Mar 29, 2010 at 12:16:57PM +0200, Rickard Utgren scratched > on the wall: >> On Mon, Mar 29, 2010 at 12:07, Dan Kennedy <danielk1...@gmail.com> >> wrote: >> >>> This database was created with a different version of the collation >>> sequence than found in the script. I think the -nocase switch was >>> probably passed to [string compare]. > >> I didn't realize this would matter, but at least now I don't have to >> worry about my code breaking it again, thanks. > > If you change a collation that is used in an index, you must rebuild > the index. That's what the VACUUM command is doing, and why it is > "fixing" the problem. You can do this more quickly with the REINDEX > command.
Yes. In fact, the REINDEX command was originally created for the specific purpose of rebuilding indices when the definition of a collating function changed. That's why REINDEX will accept the name of a collating function as its argument and only rebuild those indices that use the named collating function. D. Richard Hipp d...@hwaci.com _______________________________________________ sqlite-users mailing list sqlite-users@sqlite.org http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
