[OpenLayers-Users] Markers on Google move around N/S

[Matt Williamson]

Ok,

So I'm trying to do all this with sphericalMercator on...now, if I  
start my map with a EPSG:4326 layer, and then switch to a Google EPSG: 
900913 layer (using the layer switcher), my GML layer doesn't  
automatically transform itself...fine, but what's the correct way to  
make it do so? Or should it do it automatically?

Or, am I not supposed to try to make maps that can switch between base  
layers of different projections? Maybe that's why I'm having so much  
trouble with this...?

Thanks,

-Matt


On Sep 2, 2008, at 9:45 AM, Christopher Schmidt wrote:

> On Tue, Sep 02, 2008 at 09:32:43AM -0600, Matt Williamson wrote:
>> Hello List,
>>
>> Having a strange behavior with the following page:
>>
>> http://s89238293.onlinehome.us/w/index.php?title=Special:EditData/BirdCountGoogle/Google_form_test
>>
>> ... click on the "Show map" link, and you'll get an OpenLayers map
>> with a Google base layer. What's happening is, whenever the map is
>> panned or zoomed, the marker icon moves around North or South from
>> where it's supposed to be. Stranger, the pop-up remains in the  
>> correct
>> place!! I'm not using "sphericalMercator" on purpose, because it made
>> a mess of setting extents, etc.
>
>> It seems this should work, especially
>> since the pop-up knows how to stay put...and markers loaded with the
>> older Layer.Text mechanism (e.g. here: http:// 
>> s89238293.onlinehome.us/
>> w/ , bottom of the page) don't have this issue.
>
>> What gives?? Please  tell me there's something simple I can do to fix
>> this.
>
> You're not using Spherical Mercator. That's the way the cookie  
> crumbles.
>
> Specifically, the reason for this difference in behavior is becasue
> markers and popups are single entities: having them place themselves  
> by
> doing a request to the base layer is not too expensive.
>
> However, vectors can be thousands of points: and when rendering many
> points, this additional work (which, with Google, actually equates  
> to an
> additional *15 level deep* call stack, due to the way the GMaps API
> works) slows down drawing of vectors by as much as 50%.
>
> Rather than maintaining two sets of code -- one that does ask the Base
> Layer, one that doesn't -- we went with the path of providing a way to
> make a GMaps layer behave the way every other base layer in OpenLayers
> works, where you actually know where a given pixel is.
>
> OpenLayers offers the ability to reproject extents from Spherical
> Mercator to LonLat and reverse:
>
>  var bounds = new OpenLayers.Bounds(-180,-90,180,90).transform(
>    new OpenLayers.Projection("EPSG:4326"),
>    new OpenLayers.Projection("EPSG:900913")
>  );
>
> This is also true of LonLat objects and geometries.
>
> I would highly recommend going this route; my expectation is that in  
> 3.0
> (~6-12 months from now is my personal guess on the first release of  
> 3.0,
> though I don't really decide, just a guess at this point) that we will
> remove 'markers' as they currently exist and replace them entirely  
> with
> vectors. (Some of the spherical mercator things will probably get  
> easier
> at the same time -- but not all.)
>
> If you're using projected data, treat it as if it's projected.  
> Google's
> LonLat view of the world doesn't represent the data you're interacting
> with. Get to the roots!
>
> (This is an expansion of
> http://faq.openlayers.org/vector-related-questions/why-dont-my-vector-features-work-over-google-yahoo-virtual-earth-etc/
> )
>
> Regards,
> -- 
> Christopher Schmidt
> MetaCarta