Practical limits on object counts

[Wilhelm Sanke]

On Sat Oct 3, 2009, Richard Gaskin ambassador at fourthworld.com wrote:

> Maybe it meant "Too many for the IDE to display", since it seems the
> engine is fine. :)
>
> I wonder exactly what their threshold is for "too many", and how they
> arrived at it when they wrote their IDE.
>
> The more I played around with my tests, the more I see a very dramatic
> difference in responsiveness beyond a certain threshold somewhere beyond
> 6,000 objects.  I haven't yet pinned it down exactly, but definitely
> 10,000 is beyond acceptable for most uses, so it's somewhere between 6k
> and 9k.
>
> The difference between 6k and 9k objects is so dramatic that I have to
> wonder if there's some aspect of how the engine's memory mapping or
> object hashes were written that may play a role in the sharp performance
> degradation.
>
> I dunno.  I just know not to try working with 9k objects. ;)
>
> --
>   Richard Gaskin


Richard,

Back in 2004 we had a discussion on the improve-list very much about the 
same problem. At that time The Rev-IDE was unable even to cope with 
cards that contained around of 3000 objects. The Metacard IDE had no 
problems with this. Creating a standalone in the Rev IDE lasted 
"forever", again , no problem in the Metacard IDE.

Here is some of the relevant text on my "RevTestPage" 
<http://www.sanke.org/RevtestPage> from 2005:

> Revolution Test Stacks
>
> Following the recent discussion about the "RevTable bug" (Bugzilla 
> 2019) and Standalone Builder problems (Bugzilla 2217) on the 
> improve-revolution list (August/September, 2004) I add two stacks
>
> <http://www.sanke.org/Software/RevTestStacks.zip>
>
> and a readme file for interested testers of the Revolution IDE that 
> demonstrate the bugs and related problems with stacks that contain 
> larger number number of controls, especially larger number of fields. 
> Details - i.e. the complete bug reports of Bugzilla 2019 and 2217 - 
> are to be found by clicking the "readme"-buttons of the test stacks, 
> which however when opened in the Rev environment need some time to 
> display the contents of the connected field in the larger 3300-fields 
> stack that may bring down the Revolution IDE almost to freezing point 
> on medium-fast computers.
>
> Read below what has been fixed in the meantime:
>
> All the described problems do not appear when the stacks are opened in 
> the Metacard IDE. You can test this by downloading the Metacard IDE 
> (1.4 MB zipped) from here.
>
> Thanks to the continued efforts to improve Revolution some of the 
> described problems have now been resolved with version 2.5.1. (March 
> 2005). The speed of the test stack is now almost identical in both 
> environments with even a small advantage for Revolution - when run in 
> the development environment..
>
> However, the problem of building a standalone of this stack 
> unfortunately has not dealt with successfully even with Revolution 
> version 2.6 (June 2005). Building the standalone in the Rev IDE takes 
> "forever", whereas building a standalone of the test stack with the 
> Metacard "Standalone Builder" takes less than a second. (see comments 
> in Bugzilla 2217).
>
> Version 2.7-GM-1 of Revolution (February 2006) has also not yet 
> resolved this standalone-building problem.
>   

Responding to your post, I have expanded my test stack of 2004 to now 
containing 10000 (ten thousand) fields.

I will post the URL of the new test stack in my next post to this list.

Preliminary results of testing with the 10,000 fields' test stack:

Creating the 10000 fields takes around 200 seconds in the Rev IDE, 130 
seconds in the MC IDE.
Deleting the 10000 fields in the Rev IDE takes 420 seconds in the Rev 
IDE, 6 seconds in the MC IDE

The three buttons for creating color patterns need about 2.3 seconds in 
the MC IDE, 3.5 seconds in the Rev IDE

When the 10000 fields have been created, the Rev IDE is very much 
unresponsive, no Problems with the Metacard IDE (with the exception of 
the Metacard "Control Bowser" which needs also more time to respond when 
10000 fields are present)

I did not yet test to build standalones.

So far for now.

Best regards,

Wilhelm Sanke
Prof. emeritus
University of Kassel, Germany
Education/Educational Technology
<http://www.sanke.org/MetaMedia>