On Tue, Feb 6, 2018 at 11:15 AM, Ulrich Telle <ulrich.te...@gmx.de> wrote:

> > An alternative is to expose a virtual table with a fixed set of rows, and
> > accepting updates on the values, which can also then be "typed" too.
> > But that's a lot more complicated though.
> > (and refusing inserts/deletes too, of course).
> >
> > That vtable could also expose version information for the extension, for
> > example, and those would be read-only. Just thinking aloud.
> > Avoids non-deterministic functions.
> A vtable with a fixed number of rows, one for each config parameter - this
> approach sounds interesting.
> I'll have to investigate how complicated it will be to implement such an
> approach.

This approach could IMHO be one of the contributed vtable impls in ext/misc
to be reused by other loadable extension authors, and could become the
"semi official"
way to solve that problem, lacking extension specific pragmas that is. My
$0.02c. --DD

PS: There's also always environment variables, especially for 1-time at
startup settings.
  My main beef against env.vars. though is that they are not discoverable
and often hidden.

[1] https://www.sqlite.org/src/tree?name=ext/misc
sqlite-users mailing list

Reply via email to