[OpenLayers-Dev] OpenLayers performance

[tomas.novotny at tmapy.cz]

Boys,

the problem I am talking about is not based on number of layers and map
size. I have no problem with loading time of map tiles as well. It's just
the way map dragging is implemented.
The link to OL example Guillaume posted is also the case I wrote about.
Let's try it (http://www.openlayers.org/dev/examples/fullScreen.html).
Please switch off overlay for fair-play and zoom in the map. Then make a
few large circles with map using mouse. Try the same with GM. And don't
try to tell me you have the same feeling (at least in FF 2 and IE7).
And I've made lot of such tests with same result.
I have absolutely nothing personal against OL, I really like them from
many points of view. I also understand and agree with Christopher when he
talks about freedom of influencing OL code. But if I try to solve that
panning problem I would do so significant change that I would be really
worried about the rest of the code. I'm open for discussion about such
possible change with OL core developers but I don't have capacity to do it
on my own.

Tomas Novotny

> I wouldn't say OpenLayers performances are slower than GM. You have to
> compare things the same way. If your OpenLayers app loads several layers
> (say 10), then yes, the pan is becoming slower, a bit like if you were
> carrying a heavy weight, which is indeed what you are doing, and what GM
> can't do. Furthermore, if your map is very large, say 1680 x 1050, yes
> it can be a bit slower too. But most of the time it can be considered at
> least as fast as GM (see
> http://www.openlayers.org/dev/examples/fullScreen.html for a good example)
> But as OpenLayers can do much more than a GoogleMap app, there is
> another point to consider which is tuning. When you have many layers,
> you have to take care on how each of them is loaded and used in the app,
> and set meaningfully options such a SingleTile or the way you access the
> data (wms, wms + tilecache, wfs, geoJSON). each of these has its
> benefits and drawbacks. But it's not openlayers architecture anymore,
> it's your app architecture, your app tuning and settings, and finally,
> your job !
> And it's an exciting part of it !
>
> Guillaume
>
> Christopher Schmidt a écrit :
>> On Mon, Apr 07, 2008 at 05:34:41PM +0200, Tomas Novotny wrote:
>>> I'm after few months of using OpenLayers (OL). I've built some
>>> application on OL and appreciated it's architecture for good
>>> extensibility.
>>> But when things come to performance of such common task as panning
>>> (moving a map with mouse), I am really dissatisfied.
>>
>> Patches welcome.
>>
>>> The way the panning is implemented is everything but not
>>> browser-friendly. There is coord2pixel and pixel2coord processing on
>>> several places in those moveTo() and setCenter() methods on every mouse
>>> move and it is only part of the problem. A tried to prune OL for tests
>>> to just moving map div with one empty layer and things were still the
>>> same. You can feel lags as you are moving map from side to side, up and
>>> down. When you try to compare it with ie. GoogleMaps you can really
>>> feel
>>> the difference.
>>
>> I'll state for the record that I don't feel any difference -- if
>> anything, I typically feel like GMaps is slower than OpenLayers.
>> However, there are a number of patches in trac which have been stated by
>> their authors to be 'faster' or 'improve performance', which are
>> currently marked for further investigation in OpenLayers 2.7.
>>
>>> I haven't seen any OL example with better feeling so I think it's a
>>> matter of source architecture, not bug.
>>
>> The architecture of OpenLayers is not designed to be slow. If you are
>> finding that it is slow, there is no reason that a refactoring of that
>> aspect of the code would not be appreciated so long as it did not
>> regress in behavior in some important way.
>>
>>> What I'm trying to say? I found out that OL are good for building
>>> robust
>>> applications where wide functionality is the goal, not the performance.
>>> If I'm not right and making something wrong please tell me.
>>
>> I don't think that OpenLayers tries to be slow in any way. If you feel
>> there are obvious deficiencies in the way that OpenLayers does things,
>> and see a better way, please feel free to offer them up as comments to
>> the community.
>>
>> Regards,
>
>
>
>
>