Dojo UE Features - Phase 1

[Evan]

The following features are added for base Grid:

1. Nested sorting
2. Indirect selection
3. Context menu
4. Swipe selection and DnD moving columns & rows

please refer to the design spec in ​http://docs.google.com/Doc?id=ddt3wpr4_298c9mtpqfr, and implementation spec in ​http://docs.google.com/Doc?id=dhht2z92_1c4jpgkcd

[Adam Peller]

To apply Evan's patch:

1. Apply "dojo-grid-ue-090512.patch" to dojox directory
2. Put *.png under {DOJO}/dojox/grid/resources/images/
3. Put "music-for-demo.part.csv" under {DOJO}/dojox/grid/tests/support/

[Adam Peller]

latest dojox.grid patch from Evan

[Adam Peller]

supporting image/test files from Evan. See README.txt

[Adam Peller]

latest update from Evan (2009-07-26) includes:

1. Removed all changes to Base Grid (by extending Builder - thanks to Nathan's new changes)
2. Refined nls loading for Nested Sorting
3. Fixed many a11y issues(keyboard focus) & other misc defects.
4. Changed the location of testing cases
5. Added more comments

The test case for EnhancedGrid? is at dojox/grid/tests/enhanced/test_enhanced_grid.html

[Adam Peller]

latest patch from Evan

[Adam Peller]

090730 patch:

1. Removed unused css classes in tundraEnhancedGrid.css
2. Used _setHeaderMenuAttr() like methods to avoid deprecation warning
3. Cleaned the code a bit
4. Added the "music-for-demo.part.csv" in the patch

[Nathan Toone]

Reassigning to me for tracking, since I've been working more with Evan....

[Nathan Toone]

I am not able to apply the latest patch. Would you mind creating a more recent one. I think that we are about to the point where we will be wanting to check some stuff into svn...

[Adam Peller]

update from Evan

[Adam Peller]

goes in dojox/grid/enhanced/resources/images

[Adam Peller]

latest patch from Evan at dojo-enhanced-grid-081709.patch (see also nestedSortArrows.png)

• Fixed some issues of programmatic row selection
• Added keyboard support for moving rows/columns (via Ctrl + Arrow keys)
• Synchronized with latest Dojo trunk
• Fixed several defects (DnD rendering, visual design related ...)

[Adam Peller]

(In [19901]) Initial patch from Evan Huang (IBM) for enhanced grid. Refs #9287 !strict

[Adam Peller]

(In [19911]) from Evan: added feature description to test case. Refs #9287

[Adam Peller]

(In [19912]) from Evan: added feature description to test case. Refs #9287

[Adam Peller]

(In [19934]) refactor enhanced grid css, from Evan. Refs #9287

[Adam Peller]

(In [19935]) style fixes from Evan. Refs #9287 !strict

[Adam Peller]

Drag/drop comments: I was surprised that dnd worked only on the selected portion and not on the column header/row header. Also, I expected the drop to occur where the mouse pointer was. It seems to use the top/left of the dragged box instead.

[David Schwartz]

There is a visual indication of the drop location during a drag. Also, as column selection is persistent, there needs to be a way of disambiguating column selection and drag. Hence, drags are initiated from the content area, not the header.

[Adam Peller]

I'm suggesting that the visual indication is not correct. And for column selection, is there a drag operation on the column header? If not, I don't see why disambiguation is necessary.

[Adam Peller]

the drop should be effective based on where the mouse is, not where the top-left corner is. I'm not sure how else to say it.

[David Schwartz]

I understand your point re the mouse position; seems like a reasonable argument. I'm investigating....

RE the drag issue, column header drag events are consumed by the swipe select feature.

[Adam Peller]

I suggest we open up separate tickets for features like swipe select (most features, actually) Swipe select currently works from headers only and not from cells.

[Adam Peller]

See #9919 for swipe select #9920 for drag/drop behavior

Closing ticket, since this patch was applied. Open up new issues in separate tickets.

[Adam Peller]

(In [20512]) Patch from Evan- fixes incorrect column mover width and style/spelling corrections. Refs #9287 !strict

[Adam Peller]

(In [20530]) From Evan. Refs #9287 !strict

1. Some formatting & spelling
2. Make the mover height more accurate for multiple views
3. Removed redundant 'tundraEnhancedGrid_rtl.css'(all the rtl fixes are in 'EnhancedGrid_rtl.css').

[Adam Peller]

(In [21664]) Various enhanced grid fixes from Evan, including:

• IE6 paging performance issue by fixing measurePage()
• No onSelected/onDeselected events will be fired after sorting
• Added keepSortSelection (false by default) which is used to persist selection across sortings (only applicable for client-side data stores)

Refs #9287 !strict

[bill]

Adding patch from Evan that fixes the following issues in EDG:

1.Memory leak

• Empty all connect/topics during destroy() phase
• Memory issues are fixed with this patch (pls see the attached "EDG-memory-testing.jpg") -

2.DnD API improvement

• Mapping info is add in published topic when col/row moved in form of {fromIdx: toIdx}

3.Requirement - Make the dummy plugin mechanism more pluggable in EDG

• User can either customize or add new plugins
• TODO - update campus doc( ​http://docs.dojocampus.org/dojox/grid/EnhancedGrid )

[bill]

memory leak test results

[bill]

updated patch from evan

[Bryan Forbes]

Could you post memory results after closing the browser window/tab the grid is running in? That will give us a real measure of leaks. Right now, it's just showing us how much resident memory is taken up by the page. Make sure you disable Firebug when testing Firefox because there is a slight leak when using FB.

[evan]

Sure, Bryan,

I tried that and is attaching the result "EDG-memory-testing-results-openclose.jpg", also changed this to "reopen" as we still have the following TODO items:

1.Update campus doc

2.Clean up the test cases

3.Investigate any ways to leverage both the indirect selection and Grid's _Selector

4.Keep improving performance & internal structure as possible.

[Bryan Forbes]

After an very cursory look over this patch, I see the following problems:

1. This code:

   dojo.hitch(this.grid.selection, dojox.grid.Selection.prototype[value ? 'addToSelection' : 'deselect'])(idx);
   

   Can be simplified to:

   dojox.grid.Selection.prototype[value ? 'addToSelection' : 'deselect'].call(this.grid.selection, idx);
   

   Please change any code that uses the first pattern use the second pattern.

2. The indirect selection code is still using a widget per row. Imagine 1,000 rows of data and you scroll through all 1,000 rows. This means that you will now have 1,000 instances of whatever widget you are using for the indirect selection.

I'll keep looking at the codebase and keep providing feedback.

[evan]

Thanks Bryan,

Actually #2 won't happen with this patch except user explicitly set "keepRows=1000" - widgets will be destroyed whenever their page gets destroyed. So the number of widgets exists at the same time == rowsPerPage * existed page number.

This patch is targeting 1.5.1 by addressing most urgent issues. I'm starting working on a separate patch to leverage indirect selection and _Selector way to fundamentally improve the performance, will post that once ready.

[Adam Peller]

In general, it's good to reuse tickets when we can, but at some point, we really need new tickets to track these defects, especially as they're going into patch releases.

[evan]

Got it, so we will track this with a new one ​#11633

[Adam Peller]

sounds like three independent things. normally I'd suggest three tickets, but let's see how this goes.

[Adam Peller]

new issues will be tracked in separate tickets