Re: performance problems with bulk inserts/updates on tsrange with gist-based exclude constrains

Hey Tom,

I appreciate you're busy, but did you ever get a chance to look at this?

On 19/09/2016 08:40, Chris Withers wrote:
> On 16/09/2016 15:29, Tom Lane wrote:
>> Chris Withers <chris(at)simplistix(dot)co(dot)uk> writes:
>>> On 16/09/2016 14:54, Igor Neyman wrote:
>>>> So, what is the value for "end ts", when the record is inserted (the
>>>> range just started)?
>>
>>> It's open ended, so the period is [start_ts, )
>>
>> I've not looked at the GiST range opclass, but I would not be
>> surprised if
>> having lots of those is pretty destructive to the index's ability to be
>> selective about && searches.
>
> If that's so, that's a little disappointing...
> (I'd have thought the special case end value (open ended) and the ending
> type (inclusive/exclusive) would just be sentinel values)
>
> How would I verify your suspicions?

cheers,

Chris