PDA

View Full Version : performance question



costa_basil
25 Nov 2009, 11:09 AM
In the GXT grid plugins example, in the second grid, why is it taking a good second to select or deselect all the rows? Compare that to the ExtJs version. In the ExtJs version the operation is instantaneous! Ok, in the GXT version there are 44 rows and in the ExtJs there are only 30. But it shouldn't matter.


GXT:
http://www.extjs.com/examples/#gridplugins

EXTjs:
http://www.extjs.com/deploy/dev/examples/grid/grid-plugins.html

Thanks

sven
25 Nov 2009, 2:11 PM
Its a FF issue with huge dom layouts. You normally should not hit that in normal applications. Try the code of the checkboxgrid in a simple example, it will work with the same speed as extjs. You should not test speed at an app that shows _every_ single component and some even more than one time.

costa_basil
25 Nov 2009, 2:53 PM
You should not test speed at an app that shows _every_ single component and some even more than one time.

What ?! Perhaps I don't understand what you are saying, but I think this is the best test for an app. My application will have a zillion grids, forms, panels, trees. How do you want me to test the performance?

sven
25 Nov 2009, 3:03 PM
If your app has that all initiated and all on the same page, you should rethink your application design.

Also this is clearly not an issue of GXT. You will reach that always in FF with that big domlyout the explorer generates.

You can only compare the same with the same. So when comparing Ext JS examples with GXT one, you should compare the same example. When doing perfomnce tests you should create the layout your app will actually have. You will never have the layout of the explorer.


GXT also provides ways to minimize the domlayouts of the grid. However, this example does not integrate this.

costa_basil
25 Nov 2009, 3:33 PM
If your app has that all initiated and all on the same page, you should rethink your application design.

You are right. I don't use only one page for the entire application, but I use a page per module. I create the panels lazily as they are accessed by the user. But if the user accesses every panel in the page, then all the items in the UI will be created or rendered.


Also this is clearly not an issue of GXT. You will reach that always in FF with that big domlyout the explorer generates.

Ok, I get this with regards to FF. And I tried the selection with IE and it was faster than FF - surprisingly.


You can only compare the same with the same. So when comparing Ext JS examples with GXT one, you should compare the same example.

I thought I tested the same. In the demo I opened only the Grid Plugins tab. I assumed that the demo app doesn't create every single tab until the user clicks on the associated icon or menu item. I agree that this particular ExtJs sample is thin in the sense that it includes only the grids. Are you saying that, in the GXT sample, because we have the menu panel & the panels that surround the actual individual example panels, that can slow down the selection of rows?


You will never have the layout of the explorer.

Actually I do. My layout has an accordion panel on the left side with tree panels inside each accordion child.


Thank you

sven
25 Nov 2009, 3:37 PM
Many people thing that IE is slower, but FF can also be slower, especially when it goes to dom operations. Hopefully 3.6 solves this.

I looked into a workaround for this, and probably also have found one. If this cannot be integrated into GXT, i will post it here so you can add it to your app (if you can add it there)

costa_basil
25 Nov 2009, 3:40 PM
Thank you.