Your thread made me ponder what might be up, so I wrote a test using fts3: http://pastebin.com/AKP2yHuM and AFAICT, it works alright. I haven't specifically verified each of the flags to sqlite3_open_v2(), I just spammed what looked relevant in there.
Hmm, should have commented the #if's before posting. The ":memory:" case is to verify that it doesn't work right with plain in-memory database. The real-file case doesn't work if run twice, to verify that I'm not accidentally working right with a shared in-memory database. The third version should be a real shared in-memory database, as best I can tell. The other #if verifies that it works the same for regular table or virtual table. If I were debugging this, the first place I'd look is my virtual table's xConnect/xCreate implementation. -scott On Tue, Oct 6, 2015 at 7:53 AM, Dominique Devienne <ddevienne at gmail.com> wrote: > On Tue, Oct 6, 2015 at 4:39 PM, Simon Slavin <slavins at bigfraud.org> wrote: > > > On 6 Oct 2015, at 2:51pm, Dominique Devienne <ddevienne at gmail.com> > wrote: > > > Is this use case supported by SQLite? > > > > Whether this works correctly depends on how your Virtual Module is > > written. It should be possible to implement this correctly, but I can > > imagine a virtual system written quickly which does not take the care to > > get this right. > > > > It was not written "quickly" but it was definitely not written with that > use-case in mind. And writing a virtual module correctly is not easy. > > > > There are also things Virtual Tables can't do. For instance you cannot > > index a Virtual Table using SQL commands. > > > > That's a bit beside the point. Our vtable already provide indexing support. > Just not dynamically added indexing via SQL commands. > > From Gunter's answer, it seems at least that this use case is possible, > even though his is multi-process and DB-file based, I think I see the > parallel he's trying to make to our multi-threaded (1 connection / thread) > and in-memory DB use case. > > I just wish I could get a bit more hand-holding on how to proceed, if > possible :). As it is not quite crystal clear to me yet. Especially since > that vmodule code was written a few years back, so I don't recall the > details. Thanks, --DD > _______________________________________________ > sqlite-users mailing list > sqlite-users at mailinglists.sqlite.org > http://mailinglists.sqlite.org/cgi-bin/mailman/listinfo/sqlite-users >