Comments on: Using BitmapData for Array manipulation in AS3

By: adrien

adrien — Sun, 22 Nov 2009 09:09:19 +0000

this is great i was looking exactly for this kind of benchmark before continuing with my project,
i only need comparison to tell me if 2 bytearray vars are the same, no reunion intersection.
do you think bytearray.compress before compare would speed up ? i have flash player 10
and all tests were faster for me using bytearr and not bitmapdata

By: mzx

mzx — Thu, 29 Jan 2009 17:47:05 +0000

nice catch. There are some “optimized” pieces in flash player that works better than other. Images/filters likely uses some parallel calculations

By: Mark

Mark — Sun, 21 Dec 2008 04:07:43 +0000

Nice visual solution. I see interesting potential combining visual with functional.

By: todd

todd — Wed, 17 Sep 2008 19:33:42 +0000

Seems like a sparse matrix could be useful in this type of solution. But, I wonder 1) if there’s an Actionscript implementation anywhere, and 2) how it’d fare in this test regarding speed 3) assume just a couple of on/off state of what you did with the bitmapdata. Maybe not, though.

By: Hadrien

Hadrien — Mon, 15 Sep 2008 18:10:06 +0000

Nice hack!
Have you tried BitVector class from http://lab.polygonal.de/ds/ ?

By: Alex Bustin

Alex Bustin — Mon, 15 Sep 2008 05:14:25 +0000

I wanted to imitate your BitmapData method thus the three ByteArrays. So yes, I’m sure it could be optimized .

Anyway, glad to have added to the pot. See you at MAX!

By: Doug

Doug — Mon, 15 Sep 2008 03:01:24 +0000

@Alex – I updated the test app to include a test that uses ByteArray. I also modified your code to use writeBoolean() and readBoolean() instead of writeByte() and readByte(). I did brief testing and I think that squeezed a little more speed out of it (but I could be wrong). I also didn’t create the comparison ByteArray like you had in your code, since I don’t think that was really needed.

By: Doug

Doug — Mon, 15 Sep 2008 02:33:06 +0000

@Alex – nice! I’m going to use that bytearray method and put together another side-by-side test. My tests indicate that it’s about the same speed as the others on the 10k and 100k tests, but when you get into the 1 million items test it shaves off a solid second.

The test you posted doesn’t calculate the processing time in the same way my tests do though. Your test doesn’t account for the filtering time in the total duration, and since a big chunk of the work required for the algorithm to run (calling byteArray.writeByte) happens in the filtering function, that needs to be included in the duration benchmark as well. Once that gets thrown in the numbers look more like the other tests (but like I said, when you get into millions of records the bytearray method pulls ahead)

By: Alex Bustin

Alex Bustin — Mon, 15 Sep 2008 01:10:47 +0000

Not tested fully yet … but I seem to have gained good performance with ByteArray ..

http://thebackbutton.com/filtertest/Filter.html
http://thebackbutton.com/filtertest/srcview/

By: Rad

Rad — Sun, 14 Sep 2008 21:33:23 +0000

Thanks for sharing cool idea.

Another option to consider is to encode filtered arrays as a bitarray and then use bitwise operations. Using your data as an example

Filtered array can be encoded as:
f1=01101
f2=11110

then

~f1=10010
~f2=00001

then answer to “unique values in Filtered Array #1 that do not appear in #2 (Item 5)” would be:

f1 & ~f2
01101
00001
——-
00001.