[vwnc] Fundamental flaw in Wrapper architecture

Previous Topic Next Topic
 
classic Classic list List threaded Threaded
2 messages Options
Reply | Threaded
Open this post in threaded view
|

[vwnc] Fundamental flaw in Wrapper architecture

Terry Raymond

I made a simple modification to the File browser that uses a dataset

in place of the file list. This would permit me to see file dates and sizes

in addition to the file name. Additionally, the dataset is able to sort on a

selected column, which I find rather helpful when there are a lot of files.

 

The problem I encountered was when I removed a file. Due to the

“unsynchronous” nature of the Wrapper architecture the dataset

attempted to display the file that was just removed. One can easily

reproduce this simply by changing AbstractFileItem>>displayString

to the following and then remove a file, no other changes to the

FileBrowser are necessary.

 

displayString

 

            ^filename tail , ' - ' , filename modificationTimestamp printString

 

The actual problem I encountered is a little more complex because I am

using a modified Aragon dataset. In this situation the dataset’s ‘model’

and ‘sequence’ are different objects because, in the general case, when

a sorted list is presented the model’s list should not be reordered.

 

I say that the architecture is flawed because change/update mechanism

in concert with #invalidateNow causes the system to attempt to redisplay

portions of the application before all the underlying objects have been

updated to reflect the new state. I say “underlying objects” rather than the

model because the various views and widgets may also have supporting

objects, i.e. a sorted list, that must be updated to display properly.

 

I suggest that eliminating all immediate display updates would be the

first step in correcting this problem.

 

 

Terry

===========================================================
Terry Raymond
Crafted Smalltalk
80 Lazywood Ln.
Tiverton, RI  02878
(401) 624-4517      [hidden email]
<http://www.craftedsmalltalk.com>
===========================================================

 


_______________________________________________
vwnc mailing list
[hidden email]
http://lists.cs.uiuc.edu/mailman/listinfo/vwnc
Reply | Threaded
Open this post in threaded view
|

Re: [vwnc] Fundamental flaw in Wrapper architecture

Cesar Rabak
Terry Raymond escreveu:

> I made a simple modification to the File browser that uses a dataset
>
> in place of the file list. This would permit me to see file dates and sizes
>
> in addition to the file name. Additionally, the dataset is able to sort on a
>
> selected column, which I find rather helpful when there are a lot of files.
>
>  
>
> The problem I encountered was when I removed a file. Due to the
>
> “unsynchronous” nature of the Wrapper architecture the dataset
>
> attempted to display the file that was just removed. One can easily
>
> reproduce this simply by changing AbstractFileItem>>displayString
>
> to the following and then remove a file, no other changes to the
>
> FileBrowser are necessary.
>
>  
>
> displayString
>
>  
>
>             ^filename tail , ' - ' , filename modificationTimestamp
> printString
>
>  
>
> The actual problem I encountered is a little more complex because I am
>
> using a modified Aragon dataset. In this situation the dataset’s ‘model’
>
> and ‘sequence’ are different objects because, in the general case, when
>
> a sorted list is presented the model’s list should not be reordered.
>
>  
>
> I say that the architecture is flawed because change/update mechanism
>
> in concert with #invalidateNow causes the system to attempt to redisplay
>
> portions of the application before all the underlying objects have been
>
> updated to reflect the new state. I say “underlying objects” rather than the
>
> model because the various views and widgets may also have supporting
>
> objects, i.e. a sorted list, that must be updated to display properly.
>
>  
>
> I suggest that eliminating all immediate display updates would be the
>
> first step in correcting this problem.
>
>  
>
>  
>
Terry,

I hada my fingers burned some years ago by the same cause you report.

At that time my co-workers even remembered me the announcement mechanism
is not OO.

However, in our case when we tried to go to a message passing only
solution the application seemed to crawl like a slug in at the time very
capable machines.

The time that ocurred the expectation with the coming "Pollock" (a.k.a.
Panda) (renamed "Widgetry" ), so we waited but new GUI model did not come...

--
Cesar Rabak
GNU/Linux User 52247.
Get counted: http://counter.li.org/
_______________________________________________
vwnc mailing list
[hidden email]
http://lists.cs.uiuc.edu/mailman/listinfo/vwnc