On Thu, Jun 11, 2009 at 2:48 PM, Marko Kreen<markokr@gmail.com> wrote:
> On 6/11/09, Matt Amos <zerebubuth@gmail.com> wrote:
>> On Thu, Jun 11, 2009 at 1:13 PM, Brett Henderson<brett@bretth.com> wrote:
>> >> See pgq.batch_event_sql() function in Skytools [2] for how to
>> >> query txids between snapshots efficiently and without being affected
>> >> by long transactions.
>> >
>> > I'll take a look.
>>
>> it was looking at the skytools stuff which got me thinking about using
>> txids in the first place. someone on the osm-dev list had suggested
>> using PgQ, but we weren't keen on the schema changes that would have
>> been necessary.
>
> Except the trigger, PgQ does not need any schema changes?
i've been having a look and it seems to me that PgQ requires some
extra tables as well as the trigger. am i missing something?
PgQ might be a good solution, but i'm worried that after calling
pgq.finish_batch() the batch is released. this would mean it wouldn't
be possible to regenerate older files (e.g: a few days to a week) in
case something unexpected went wrong. it might not be a major problem,
though.
i think we could get the same functionality without the extra daemons,
by putting a trigger on those tables for insert and recorded the
object id, version and 64-bit txid in another table. but if we're
going to alter the schema we might as well put the txid column
directly into those tables...
cheers,
matt