Shouldn't statement->execute() set used to true?

[mqudsi]

Hello,

Thanks for this library, I just migrated a project over from CppSqlite (which I'd been maintaining) over to this.

I have a question on prepared statements - is there a reason manually calling execute on a statement (and not statement++) does not set used to true? From the documentation and going by logic, I assumed execute() would execute the statement and set it to used. reset would unbind parameters and set used to false. But that's not the case since it remains "unused" even after explicitly executing and will run once more before destructing unless used(false) is manually called.

Thanks.

[Killili]

Hi, and thanks.

As far as i remember, reset will reset the bindings and if you want to reuse them, as is the case with prepared selects you might not want to use op++.

[mqudsi]

Thanks for the quick reply. I guess my question wasn't too clear.

ErrorCode DBWrapper::InsertBlock(int blockID, uint32_t overlapOffset /*= -1*/)
{
    auto statement2 = *_masterDB << "INSERT INTO 'Overlaps'\
                    (BlockID, OverlapOffset) VALUES \
                    (?, ?);";

    return SafeSql(__FUNCTION__, [&] {
            statement2->reset();
            statement2 << blockID;
            statement2 << (int)overlapOffset;
            statement2->execute();
                    statement2->used(true);
    });
}

In the above code, if statement2->used(true); is not commented out, when SafeSql returns (which just executes the lambda in a try..catch) and statement2's destructor is called, it will execute once again (and fail because the bindings are invalid).

The destructor for database_binder checks if execution_started is false in order to determine if it should be run once more. But calling execute() does not set execution_started to true in the above case.

[aminroosta]

@mqudsi This looks like a bug to me.
@Killili Will #35 fix this?

[Killili]

Ok after thinking a bit about it (has been a few weeks since i wrote that).
And i had to read a lot of old Delphi code today so have mercy.
I think the intentional separation between execute and reset is the thing that i was thinking about and this "used" issue is an oversight.

So go ahead pull the fix.

[aminroosta]

@Killili Cool!
Please take another look at diff in #35, If it looks good to you please merge it :-)
I also set ps.used(false) on ps.reset(), that should be fine. right?

[mqudsi]

@Killili no worries, thank you kindly.