Home Original page

Object oriented databae for Python

achrist at easystreet.com achrist at easystreet.com
Fri Mar 21 16:55:51 EST 2003 
Gerhard Haering wrote:
> 
> * Konstantin Knizhnik <knizhnik at garret.ru> [2003-03-21 21:31 +0300]:
> > Hello achrist,
> >
> > This is problem with definition DL_EXPORT macro in Python 2.2.2 [...]
> 
> There is no problem. The problem might be in the Makefiles you use. So
> don't use Makefiles. Use distutils through the setup.py I created :-)
> 
> The only thing you'll likely have to change is the library to link
> against, cos MSVC doesn't have any libstdc++. It's called otherwise,
> but I don't recall how.
> 

Thanks. You got me going.

Well, I just commented out the ", libraries = ["stdc++"]" from the
setup.py file that came with DyBase, and setup.py worked ok.  Then
I copied the resulting pythonapi.pyd from
Dybase\Python\Build\Lib.Win32-2.2 up to Dybase\Python.  From there
the samples run fine.

On my 200 MHz machine, the index test showed about 850 Inserts per
second, 1150 retrievals per second, and 650 deletes per second.  That's
very good performance compared to my vague recollections of other
things I have timed, but I haven't timed much recently.

I watched the memory in the Win NT task mangler as this was running.
It ran up to 38 Mb used by Python, but it didn't get any higher than 
that.  I wonder why it stopped there.   At that point, my total 
memory used was about 140% of real, about 70% of total virtual
memory available.   This is not a problem, and it looked like a very
nice limit of memory consumption for my machine, but it makes me wonder
if this was determined by some graceful algorithm that takes into
consideration the total available, or if memory could be a problem on a
machine with 64 Mb or so that happens to be running some memory hogs
from MS or a big java application at the same time.

IAC, for now it looks great and kudos of praise and thanks again to
Konstantin Knizhnik. I have reported here only the little problems with
which I consternated myself not as complaints, but so that the right
approaches are exposed in public to help others that are operating at
a level of ignorance comparable to mine.  


Al
More information about the Python-list mailing list