On Sep 20, 2010, at 3:51 PM, Randall Nortman wrote:
> On Sep 19, 1:52 pm, Michael Bayer wrote:
>> On Sep 19, 2010, at 9:54 AM, Randall Nortman wrote:
> [...]
>> I think this is a bug in Pysqlite2.I can reproduce it with sqlite3
>> directly.
> [...]
>>> (I don't see where the spec says
>>>
you got it, re97f096e41aa
On Sep 21, 2010, at 10:23 AM, Peter Hansen wrote:
> On 2010-09-20 5:27 PM, Randall Nortman wrote:
>>> I've always thought BEGIN was clearest but I'm
> not going to fight hard on that, so just suggest something it could say.
>
>> "BEGIN (implicit)" would be enough to c
On 2010-09-20 5:27 PM, Randall Nortman wrote:
I've always thought BEGIN was clearest but I'm
not going to fight hard on that, so just suggest something it could say.
"BEGIN (implicit)" would be enough to clue me in that something may
not be as it seems, and shouldn't confuse too many people (f
On Sep 20, 4:22 pm, Michael Bayer wrote:
> On Sep 20, 2010, at 3:51 PM, Randall Nortman wrote:
>
>
>
> > On Sep 19, 1:52 pm, Michael Bayer wrote:
> >> On Sep 19, 2010, at 9:54 AM, Randall Nortman wrote:
> > The DB-SIG list agrees with you:
>
> >http://mail.python.org/pipermail/db-sig/2010-Septemb
On Sep 20, 2010, at 3:51 PM, Randall Nortman wrote:
> On Sep 19, 1:52 pm, Michael Bayer wrote:
>> On Sep 19, 2010, at 9:54 AM, Randall Nortman wrote:
> [...]
>> I think this is a bug in Pysqlite2.I can reproduce it with sqlite3
>> directly.
> [...]
>>> (I don't see where the spec says
>>>
On Sep 19, 1:52 pm, Michael Bayer wrote:
> On Sep 19, 2010, at 9:54 AM, Randall Nortman wrote:
[...]
> I think this is a bug in Pysqlite2. I can reproduce it with sqlite3
> directly.
[...]
> > (I don't see where the spec says
> > that a DBAPI connection is always in a transaction.)
>
> Wel
On Sep 19, 2010, at 9:54 AM, Randall Nortman wrote:
>
> Unfortunately, the isolation_level parameter to pysqlite2 only
> controls the type of BEGIN issued, not when it is issued (I just
> tested to be sure). It still waits until there is an INSERT/UPDATE/
> DELETE to begin the transaction. I e
On Sep 18, 11:59 pm, Michael Bayer wrote:
> On Sep 18, 2010, at 6:52 PM, Randall Nortman wrote:
>
> > In testing my code for concurrency, I discovered that transactions are
> > not properly isolated on sqlite, even with
> > isolation_level='SERIALIZABLE'. It turns out that on the sqlite
> > diale
On Sep 18, 6:52 pm, Randall Nortman wrote:
[...]
> isolated as expected. Dirty writes are easy to construct; using the
[...]
Sorry, I meant "lost update" rather than "dirty write". In my head I
think of dirty read and dirty write, but I should translate to
standard terminology when the concept