Re: wrong indexes sizes on index page

Le 13/10/2010 10:54, Thom Brown a écrit :
> On 13 October 2010 09:27, Thom Brown <thom(at)linux(dot)com> wrote:
>> On 13 October 2010 08:30, Guillaume Lelarge <guillaume(at)lelarge(dot)info> wrote:
>>> Le 13/10/2010 00:35, Thom Brown a écrit :
>>>> On 12 October 2010 16:19, Guillaume Lelarge <guillaume(at)lelarge(dot)info> wrote:
>>>>> Le 12/10/2010 14:37, Thom Brown a écrit :
>>>>>> On 12 October 2010 11:01, Timon <timosha(at)gmail(dot)com> wrote:
>>>>>>> 2010/10/12 Guillaume Lelarge <guillaume(at)lelarge(dot)info>:
>>>>>>>> Le 12/10/2010 11:47, Thom Brown a écrit :
>>>>>>>>> On 12 October 2010 10:29, Timon <timosha(at)gmail(dot)com> wrote:
>>>>>>>>>> see attachment.
>>>>>>>>>> Last column on table shows index size, but 227MB - is table size.
>>>>>>>>>>
>>>>>>>>>> postgresql 8.4.4
>>>>>>>>>>
>>>>>>>>>> pgadmin 1.13 last dev build from git
>>>>>>>>>>
>>>>>>>>>
>>>>>>>>> I can confirm this using pgAdmin ... weird special version Dave gave
>>>>>>>>> me last week ... against PostgreSQL 8.3.1, and PostgreSQL 9.0.0.
>>>>>>>>>
>>>>>>>>
>>>>>>>> I don't think we changed anything in this area. So you should have the
>>>>>>>> same issue with 1.12.
>>>>>>> yes, I have this issue with 1.12.
>>>>>>
>>>>>> Looks like it's getting the OID of the table the index is associated
>>>>>> with rather than just using the indexrelid column from
>>>>>> pg_stat_all_indexes.
>>>>>>
>>>>>
>>>>> You're right. Not sure what I was thinking when I applied this on 1.12.
>>>>> Quick patch (not yet checked):
>>>>>
>>>>> diff --git a/pgadmin/schema/pgIndex.cpp b/pgadmin/schema/pgIndex.cpp
>>>>> index e55daf3..4c2fa23 100644
>>>>> --- a/pgadmin/schema/pgIndex.cpp
>>>>> +++ b/pgadmin/schema/pgIndex.cpp
>>>>> @@ -374,7 +374,7 @@ void pgIndexBase::ShowStatistics(frmMain *form,
>>>>> ctlListView
>>>>> wxT(", idx_blks_hit AS ") + qtIdent(_("Index Blocks Hit"));
>>>>>
>>>>> if (GetConnection()->HasFeature(FEATURE_SIZE))
>>>>> - sql += wxT(", pg_size_pretty(pg_relation_size(") + GetOidStr()
>>>>> + wxT(")
>>>>> + sql += wxT(", pg_size_pretty(pg_relation_size(indexrelid)) AS
>>>>> ") + qtId
>>>>>
>>>>> if (showExtendedStatistics)
>>>>> {
>>>>>
>>>>> If you want to test, it's all yours. Will check tonight and apply if
>>>>> good enough.
>>>>
>>>> Okay, built with change and works fine :)
>>>>
>>>
>>> You mean with your changes? because this pseudo-patch doesn't work. I
>>> worked on it yesterday, and found I didn't modify the good method.
>>> Anyway, I have a patch ready to commit.
>>
>> Yes, with my changes, which on glancing at yours, looked roughly the
>> same, but must be a difference somewhere.
>
> I now see why your one is different. My change involved a
> similar-looking piece of code, but in
> pgIndexBaseCollection::ShowStatistics, not
> pgIndexBase::ShowStatistics.
>

Exactly :)

--
Guillaume
http://www.postgresql.fr
http://dalibo.com