Throw if truncation happens but no indicators was provided about soci HOT 3 CLOSED

So I've finally started working on this. The bad news is that actually no backend even sets the indicator correctly currently, here is the result of the tests:

1. Firebird: indicator is i_ok (the value is, of course, still truncated).
2. MySQL: indicator is i_ok.
3. ODBC using PostgreSQL backend: exception thrown.
4. PostgreSQL: exception thrown.
5. SQLite: indicator is i_ok.

I can't test Oracle which is a pity because, looking at the code, this is the only backend which has some code for returning i_truncated (ODBC backend also has it but clearly this doesn't work there).

Anyhow, I still think that the current situation is wrong, but it's even wronger than I thought and my changes will break the code for all backends (possibly excluding Oracle) as not only will the indicator be set now if available, but also the exception currently thrown by PostgreSQL backend will not be thrown any more. If anybody objects against changing this and sees some other way of fixing this in a more backward compatible manner, please comment.

from soci.

Update: I have now tested with Oracle too and it behaves as PostgreSQL, i.e. an exception is thrown.

So I actually have a new plan: just make all backends throw an exception on truncation except, perhaps, SQLite, which doesn't lose data (even though it doesn't truncate neither).

The main problem I have with this is that I don't know how to make MySQL generate an exception in non-strict mode, would anybody know how to do it? If not, I'll probably just leave MySQL alone.

from soci.

The new plan sounds good 👍

I don't know. @pfedor is our MySQL guru.

from soci.