Grid speed comparison for 4 frameworks

[mankz]

http://www.dhtmlx.com/blog/?p=1525

[mitchellsimoens]

Would have liked to see them use 4.1.0 as you stated in a comment but oh well.

[joeri]

I have to agree with their perspective though: you don't use release candidates for benchmarking, only shipping software.

Anyway, why is it that their grid is faster? I thought the v4 grid in ExtJS was supposed to be engineered for better performance?

[mitchellsimoens]

It turns out to be a bad test also. Other frameworks as the number of rows get larger the speed stays pretty flat? Sounds like they are buffering then and the example they have of Ext JS is not a buffering one so of course the performance of Ext JS is going to be so bad.

[joeri]

I checked their code, and I didn't notice anything about buffered views.

But, it's still relevant, because they just implemented the grid like the Sencha examples recommend. I literally couldn't figure out how to do buffered grid views in ExtJS 4 (I only use v3 professionally) with 15 minutes of browsing the documentation. That's a problem with ExtJS and its documentation, not a problem with the benchmark.

[mitchellsimoens]

Have you see the buffered grid example? Basically the example in RC2 loads 5,000 records into the store and between the grid and the store it only shows a certain amount at a time.

[joeri]

I only looked in the API docs, not in the separate examples section. I assumed that the API docs would explain a core feature like buffered views.

Looking at the buffered example, I assume that the buffered view is being done by using a buffered store combined with a paginggridscroller. Looking at the benchmark code, that's actually what they're doing. On the other hand, when I scroll, there's no visible delay in rendering rows, so I'm not sure that the grid is actually rendering with a buffered view. Maybe someone with some hands-on ExtJS 4 knowledge needs to download the benchmark code and check what is going on?

[mitchellsimoens]

Nige (Animal) tore the test apart. It wasn't a fair test for what he says but something that will still motivate us to be better. 4.1.0 won't be the last of performance fixes for grids and other widgets, we have some big plans for 4.2.0 also!

[mcouillard]

While this test is simply against 1 of many components these frameworks use, looking only at the most simple test can be a useful comparison.

I downloaded this test and compared Ext 4.0.7 to 4.1.1 and 4.1.2a. I also tried dhtmlx3, though I was less interested in that.

Framework   Avg Load Time (Chrome; Test 1, 100 rows, 10 cols)
Ext 4.0.7   110ms
Ext 4.1.1   330ms (what performance improvement!?)
Ext 4.1.2a  340ms
    (debug js versions were NOT used, unlike the original tests)
dhxGridx3   75ms (SmartRendering OFF)

IE9 also proved to be around twice as slow using 4.1.1/2 compared to 4.0.

While I realize it's easy to disregard such a simple test, when comparing your own framework's versions against itself and performance gets worse over time... that's a huge red flag to me.

As a long-time user of ExtJS I'd prefer to see continued performance optimizations and lightweight/efficient themes (Neptune sounds like a good start) and build process improvements (Sencha Cmd is rough... but an improvement) over additional features - easily.