[pLog-svn] r6452 - in plog/trunk/class: . object

[Mark Wu]

> >>> 2. Oscar remove "all" lt_include(), so, if you do this,
> >> some test can
> >>> not run well..
> >>  	If I am reading the code correctly, it looks through 
> the map, and 
> >> if it can't be found, then looks through all of the 
> directories.  So, 
> >> the tests should still work correctly.
> >
> > Well, yes .... The side effect is "bad performance" coming 
> back again.
> >
> > I don't like this. I prefer use "a small memory usage" to 
> get  "better 
> > performance".
>  	I don't understand.  The code currently looks in the 
> map file, and if it isn't found, looks through the 
> directories alphabetically.  I am saying that it shouldn't 
> ever look through the test directories until the very end, 
> since real files shouldn't ever get to those directories, 
> unless they are not going to be found anyway.
>  	I think the directory list should not be alphabetical, 
> but in an order that makes the most sense from an efficiency 
> standpoint.  I'd want to see when that directory search code 
> is used, and make it work as quickly as possible, and at 
> least that means moving the test cases to the end.  I am not 
> sure what other changes should be done.

For the directory lookup order change is okay for me. 

I just don't like removing those files from class loader map in development
stage.

Of course our current implementaton can find the class file and include it
correctly, but you never know how much time it will take.

That's why I said, I prefer those files keep  in there.

I really don't like to take another full day to find out why some scripts
run very slow ... This kind of bug is really annoying.