ArrayCollection fast filtering

I have the following situation: have a screen with 4 charts, each being populated with 59 array collections. Each array collection has 2000+ data points.

This construct needs to be filtered(by time), but the problem I'm encountering is the fact that it takes a long time for the filtering to finish(which is to expected g开发者_如何学Pythoniven the amount of data points that need to filtered).

Filtering one chart at a time is not option, so I wanted to ask you what do you think would be the best approach? (should I use vector instead?). To generalize this question what would be the best way to filter large collections in flex/as3?

Thanks.

You'll have to squeeze out every bit of performance improvement that's possible and that's suited:

• use Vector if possible, and as much as you can. It has (contrary to what www.flextras.com posits) a filter property which accepts a filtering function. ArrayCollections are slow. (In general all flex native classes are unnecessarily slow) So if you really HAVE to use ArrayCollections, only use them to present the resulting Vectors.

• if the problem is that the application "freezes" you can look into green threading so you can present the user with a progress bar, that way they at least have a sense of progress. http://blog.generalrelativity.org/actionscript-30/green-threads/

I would suggest to filter large collections on the server. This has several benefits:

• you may be able to minimize network traffic because only the filtered data have to be transferred
• server side computing can be parallelized and is typically faster, because of more performant hardware and the server's runtime language (e.g. Java).
• services requests are done asynchronously, so your client application is not blocked

Use vector's wherever possible, use green threading if you still can't manage. Internally we use a lot of dictionaries to cache computed queries for later lookup. Dictionaries in as3 are one of the fastest objects around. So we pre-filter in the background and story the various filterted collections in a dictionary. Not sure if it works for your case.

Ok, so I googled around more for green threading and stumbled upon a project by gskinner (PerformanceTestv2). Testing the generation of data vs rendering time got me the following results:

[MethodTest name='Set gain series test: ' time=1056.0 min=1056 max=1056 deviation=0.000 memory=688] //filtering data source

[MethodTest name='Set gain series test: ' time=24810.0 min=24810 max=24810 deviation=0.000 memory=16488] //filtering + rendering.

Next I looked on how to improve rendering time of the chart, but not much improvement. However I did find a project based on Degrafa: Axis. This has been ported to flex 4, event 4.5([Axiss 4.5])3. I've integrated charts based on this framework and the results are really great so far.