Testing pier-magma

[Keith Hodges]

Findings so far.

I have persisted 248 pages so far (a far cry from the promised 5000!) 
but I am doing so slowly.

Magma reports 84818 objects from aRepositoryController objectCount.
That is an average of 342 objects per page. Although Pier pages do 
contain a lot of objects to describe what most wiki's store in a single 
string. 342 seems to be quite a lot.

Disk usage is 3Mb for 250 pages which is 12k per page, i.e. 60Mb for a 
site of 5000 pages with one page generated per Squeak class.

So I opened an explorer on the repository and took a look at a simple page.
The page generated for Squeak class 'Password' seemed to be quite 
typical, not too large. (I didn't even know squeak had a class called 
Password).

The page consists of headings, the class comment, a list of subclasses 
(Password has none)
an unordered list of instance methods each being a [broken] link (i.e. 
pages for individual methods have not been filled in) and a similar 
unordered list of class methods.

Counting objects in the repository explorer I get approximately 102 
unique Objects, 50 nil's, and 8 object references. Just for one page! 
160 Objects that I can explicitly count, I am not sure what Magma is 
doing behind the scenes in terms of overhead.

The only thing that I can see that could be optimised is the appearance 
of 9 empty arrays ( #decoration and #children ) which could be lazily 
initialialized. I doubt if it is worth doing what is the overhead of 
storing an empty array, as opposed to nil?

Keith









		
___________________________________________________________ 
Try the all-new Yahoo! Mail. "The New Version is radically easier to use" – The Wall Street Journal 
http://uk.docs.yahoo.com/nowyoucan.html