[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
Re: [cross-project-issues-dev] Open Editor timeouts

Ed,

Only the editors know what it is they're doing and the division between what should be done on the UI thread (creating widgets/controls), what could/should be deferred, and how that deferral as well as progress is reflected in the UI doesn't sound like something for which the platform can just provide a generic solution.  At best some type of scaffolding/framework, but that's not a prerequisite for solving the problem.  For example, for Oomph we load resource sets that contain a very large number of internet hosted resources, but the performance of that is generally so bad that we must take steps to do that on a background thread, i.e., it's not just 2 or 3 seconds but rather several 10s of seconds.  So we do several things:
  •  Keeping local caches of remote resources with etag/timestamp checking to reduce internet traffic.
  • Loading resources in parallel on many threads (because internet loading is mostly waiting for a response, so 10 parallel requests will generally produce results 10 times faster).
Of course it would be nice to feed this type of stuff back into EMF...

I must also say that personally I'm finding the UI freeze dialogs themselves the thing that is generally blocking and disrupting the UI in cases where I have no real personal perception that something is frozen.  Certainly shipping with this type of behavior will result in it generally being turned off, or in a ridiculous flood of bugzillas.


On 03/12/2014 8:44 AM, Ed Willink wrote:
Hi Lars

I think you're agreeing with me and identifying a useful criterion for must-use-Worker.

But while it's not a platform problem, only a platform solution makes sense. Why should many editor projects develop similar use-worker-to-open strategies? Much better at the platform level where there might even be integration with changing project dependencies during a slow open.

    Regards

        Ed Willink
   
On 02/12/2014 21:07, Lars Vogel wrote:
Hi Ed,

If an editor accesses the Internet I think it should use a worker thread to get the editor content. I have not seen reports for platform editor freezes, AFAIK the platform editors do not access the Internet for their content. 

I would think this is not a platform problem, an editor is IMHO responsible to get his data model and I'm unware of special support in the platform for asynchronous editor loading.

Best regards, Lars

2014-12-01 18:34 GMT+01:00 Ed Willink <ed@xxxxxxxxxxxxx>:
Hi

The Code Recommender's reports highlight a cross-project phenomenon that is difficult to allocate to a particular project.

If an editor is slow to open, the current opening strategy on the UI thread can lead to a UI freeze.

For platform editors, the 2s timeout may just be adequate, so, not a platform problem.

Bug 453590 causes the basic Ecore Editor to timeout if a user gives a stupid URI causing Internet failures.

By the time we get to Xtext and worse still my OCL Xtext editors, there is ample scope for opening to take more than 2 seconds.

I'm getting irritated by the number of UI freeeze duplicates and could attempt to override the platform/Xtext editor strategies to put up a loading welcome page while a worker thread does the necessary.

But I feel that provision of such a strategy is really a platform job rather than my job three levels of abstraction further on.

Is anyone else seeing this as a problem?

    Regards

        Ed Willink
_______________________________________________
cross-project-issues-dev mailing list
cross-project-issues-dev@xxxxxxxxxxx
To change your delivery options, retrieve your password, or unsubscribe from this list, visit
https://dev.eclipse.org/mailman/listinfo/cross-project-issues-dev



_______________________________________________
cross-project-issues-dev mailing list
cross-project-issues-dev@xxxxxxxxxxx
To change your delivery options, retrieve your password, or unsubscribe from this list, visit
https://dev.eclipse.org/mailman/listinfo/cross-project-issues-dev


No virus found in this message.
Checked by AVG - www.avg.com
Version: 2015.0.5577 / Virus Database: 4235/8669 - Release Date: 12/02/14




_______________________________________________
cross-project-issues-dev mailing list
cross-project-issues-dev@xxxxxxxxxxx
To change your delivery options, retrieve your password, or unsubscribe from this list, visit
https://dev.eclipse.org/mailman/listinfo/cross-project-issues-dev