Re: [QUESTIONS] MySQL benchmark page - Mailing list pgsql-hackers

On 15 Mar 98 at 22:31, Bruce Momjian wrote:

> Nice to see the MySQL crashme page has updated to its PostgreSQL title
> to PostgreSQL 6.3, but it seems nonee of the items haven't been updated
> for this release, http://www.tcx.se/crash-me.html.
>
> I see no mention of the new subselects, and it says we don't support
> comments as --.  And I see an item called "atomic updates" which I think
> we decided was a very bizarre test, but still no mention of
> transactions, which is a very basic database capability.  We also have
> large objects.  Is that the same as blobs?  We have "Function
> concatenation with ||".  We have "Function CURRENT_DATE".  We have
> "Function LEFT".  We have "Function IN on strings in SELECT".  We have
> "Function = ALL".  We have "Update with sub select".  We have "case
> insensitive compare".  We have "lock table".
>

I have looked into crash-me and screened the items you mentioned.
Here is the result:

about the large objects. I looked in the manual of postgres and
couldn't find that fast anything about large objects.
blobs is a type where you can put binary or other big things in.

first the concatenation with ||
the test is select 'abc' || 'def'
and that's not working in postgres. In other db's it's working.

function left:
the test is select left('abcd',2)
looked in the manual of postgres and couldn't find the function left.
this isn't working and I tried some other ways but couldn't find any
working query.

function current_date:
the test is select current_date()
that's postgres not supporting only select current_date (without ()).
that's also the case with current_time, current_timestamp. I will make
patch so it's tested both ways.

Function IN on strings in SELECT:
the test is: select 'monty' in ('david','monty','allan')
and that's not working. it gives: using an explicit cast

Function = ALL:
query: select a from crash_me where b =all (select b from crash_me)
yes
so it's supported

Update with sub select:
query: create table crash_q (a integer,b char(10))
query: insert into crash_q values(1,'c')
query: update crash_q set crash_q.b= (select b from crash_me
where crash_q.a = crash_me.a)
Got error from query: 'update crash_q
set crash_q.b= (select b from crash_me whe re crash_q.a = crash_me.a)'
ERROR:  parser: parse error at or near "=" query: drop table crash_q
no
I think this test shows that it isn't possible todo an update with a
sub select

case insensitive compare:
query: select b from crash_me where b = 'A'
query: select b from crash_me where b = 'A' didn't return any
result
no
(field b in table crash_me has a value of 'a'.)

the transaction are still not inserted in the crash-me test because
monty and I didn't have time to insert it. But what I mentioned
before if you would like to see some other tests in the crash-me test
you can add them by your self and send us the patch. We will add it
to the release of crash-me.

I am not sure the results of postgres in the crash-me page is from
the final release of postgres. Because I have so much problems with
postgres it's at this moment hard to test. Further it's taking a very
lot of time to test the postgres db because it's eating my memory.
When I test the limit of the number of joins I can do it's eating my
memory till there is no memory left anymore. And that's a lot (96 mb
real + 470mb swap).

Greetz...

Luuk
|-----------------------------------------------------------------------|
|  Luuk de Boer -> <luuk@pi.net>                                      |
|  When you want to go fast .....                                       |
|  Drive a cobra on the road and take a T3 line for internet ...... :-) |
|  http://www.luuk.com/  -> The AC Cobra site                           |
|-----------------------------------------------------------------------|