[OpenLayers-Dev] [Dev] Re: Vectors in OpenLayers?
Cameron Shorter
cameron.shorter at gmail.com
Fri Oct 20 16:01:06 EDT 2006
Paul, Bart,
After going back to the UML diagram
http://wiki.osgeo.org/index.php/AJAX_WebMapping_Vector_Rendering_Design#Vector_Layers_Class_Diagram
I've realised that there are holes in my arguements regarding
serialisation. Either we need to change our design, or we will need
serialisation as you suggest.
I remember Geotools has a concept of a DataSource. It would be good to
check their design and determine if there are any similarities we can
extract.
Paul Spencer wrote:
> I'm not keen on using GML as the only representation. For me, I rarely
> deal with GML and am more often dealing with WKT.
>
> I am also hesitant to agree to having the digitizing tools generate GML
> that is then processed into Geometry->Shape ... that seems quite
> awkward to me.
>
> I suggest that the architecture support serializing Geometries/Shapes
> into something ... those of us that are interested in implementing
> support for a particular format can implement it.
>
> Then digitizing/editing tools can generically operate on Shapes and be
> part of the generic graphics lib. Remember that one of our design
> goals was to build a graphics lib independent of the GIS part, that
> could be used as a stand-alone component.
>
> Cheers
>
> Paul
>
> On 20-Oct-06, at 7:15 AM, Cameron Shorter wrote:
>
>> Bart van den Eijnden (OSGIS) wrote:
>>
>>> Also, if somebody starts drawing new geometries from scratch, you
>>> will need a way to generate GML to insert them using a WFS. So not
>>> in all use cases will the client already have the data in GML format.
>>
>>
>> We already cover this use case in Mapbuilder, and I suggest we use
>> the same design pattern in this design.
>>
>> When a User draws on a map, we process the mouseActions and create a
>> GML Layer. The GML Layer is rendered the normal way GML->Geometry
>> Classes->Vector Renderer.
>>
>> Then when the user presses the SAVE button, you build a WFS
>> Transaction from the GML and send it to the server.
>>
>> So there is no need for serialising from Geometry->GML.
>>
>>> Cameron, are you suggesting to go for one format (WKT or GML), and
>>> then use the server-side to transform if necessary?
>>
>>
>> I think it will be a rare user requirement to convert from WKT->GML
>> or visa-versa in a Web Client. So I don't think it is worth spending
>> too much time addressing yet.
>>
>>> If so, would you go for WKT or GML? In my opinion GML is the most
>>> logical option, it can always be transformed server-side to WKT if
>>> you need to go to something like PostGIS directly (without a WFS),
>>> since mostly we will talk to OGC web services and not to something
>>> like PostGIS directly. Disadvantage is that its structures are more
>>> elaborate than WKT.
>>
>>
>> I agree that GML would probably be better.
>> We have quite a bit of experience and code dealing with GML and WFS
>> in Mapbuilder, and we should be able to reuse that.
>>
>>> Best regards,
>>> Bart
>>> Cameron Shorter schreef:
>>>
>>>> I'm wondering whether we will need to create a Geometry->GML
>>>> serialiser.
>>>> In an application which uses GML, the rendering logic will go:
>>>>
>>>> GML -> Geometry Classes -> Vector Rendering HTML
>>>>
>>>> Ie, the client will already have the data in GML format.
>>>>
>>>> If the user edits features, then we can have the tools update the
>>>> data source (GML in this case) which in turn will propagate the
>>>> update though to the Geometry and Renderer.
>>>>
>>>> The other reason you might want to serialise is so that you can
>>>> convert from one format to another, eg WKT->Geometry->GML or
>>>> similar. However I suggest we keep vector transformation on the
>>>> server, probably as a web service.
>>>>
>>>> Paul Spencer wrote:
>>>>
>>>>> For you maybe ;) GML would be another good output choice I guess
>>>>>
>>>>> Paul
>>>>>
>>>>> On 19-Oct-06, at 3:44 AM, Bart van den Eijnden (OSGIS) wrote:
>>>>>
>>>>>
>>>>>> Will there also be an option to serialize the objects as GML?
>>>>>>
>>>>>> Since interaction will mostly be against OGC Web Services.
>>>>>>
>>>>>> Best regards,
>>>>>> Bart
>>>>>>
>>>>>> --
>>>>>> Bart van den Eijnden
>>>>>> OSGIS, Open Source GIS
>>>>>> http://www.osgis.nl
>>>>>>
>>>>>>
>>>>>> --------- Oorspronkelijk bericht --------
>>>>>> Van: Paul Spencer <pspencer at dmsolutions.ca>
>>>>>> Naar: Pierre GIRAUD <bluecarto at gmail.com>
>>>>>> Cc: dev at openlayers.org
>>>>>> Onderwerp: Re: [OpenLayers-Dev] [Dev] Re: Vectors in OpenLayers?
>>>>>> Datum: 19/10/06 00:09
>>>>>>
>>>>>>
>>>>>>> Hi Pierre,
>>>>>>>
>>>>>>> I haven't looked at the code but I think that this will be
>>>>>>> needed in
>>>>>>> some sort of factory object that can create Geometry (and
>>>>>>> associated
>>>>>>> Shape) objects from WKT. We are a bit early for implementation
>>>>>>> details right now (despite the fact that I've already done a
>>>>>>> bunch of
>>>>>>> implementation), but when the design is finalized, we'll need this
>>>>>>> piece.
>>>>>>>
>>>>>>> This reminds me that Geometry objects need to be able to serialize
>>>>>>> themselves as WKT.
>>>>>>>
>>>>>>> Cheers
>>>>>>>
>>>>>>> Paul
>>>>>>>
>>>>>>> On 13-Oct-06, at 12:46 PM, Pierre GIRAUD wrote:
>>>>>>>
>>>>>>> > I wanna propose you some peace of code for this.
>>>>>>> > Hope you'll be able to have a quick look and check if I'm
>>>>>>> the good
>>>>>>> > way.
>>>>>>> >
>>>>>>> > I tried to follow the "factory" pattern, but I'm
>>>>>>> not sure I
>>>>>>> > implemented it right considering the rest of the architecture.
>>>>>>> >
>>>>>>> > Regards and nice week-end
>>>>>>> >
>>>>>>> > Pierre
>>>>>>> >
>>>>>>> > ps : this code might be useable in the current vector sandbox
>>>>>>> >
>>>>>>> > On 10/13/06, Pierre GIRAUD <bluecarto at gmail.com> wrote:
>>>>>>> >> I don't know if this is a needed functionnality, but it
>>>>>>> can
>>>>>>
>>>>>>
>>>>>>
>>>>>> probably
>>>>>>
>>>>>>> >> help. I was trying to think about and implement a way
>>>>>>> to add a
>>>>>>> >> feature
>>>>>>> >> by its WKT string writing.
>>>>>>> >>
>>>>>>> >> Something like :
>>>>>>> >>
>>>>>>> >> OpenLayers.Vector.WKT
>>>>>>> >>
>>>>>>> >> p4 = new OpenLayers.Vector.WKT('POINT(-40 30)');
>>>>>>> >>
>>>>>>> >> Regards
>>>>>>> >>
>>>>>>> >> Pierre
>>>>>>> >>
>>>>>>> >> On 10/13/06, Paul Spencer
>>>>>>> <pspencer at dmsolutions.ca> wrote:
>>>>>>> >> > Thanks Chris :)
>>>>>>> >> >
>>>>>>> >> > Performance really sucks if you add a lot of
>>>>>>> features. I
>>>>>>
>>>>>>
>>>>>>
>>>>>> wrote a
>>>>>>
>>>>>>> >> > quick loop to add a 1 degree box every 10 degrees and
>>>>>>
>>>>>>
>>>>>>
>>>>>> brought my
>>>>>>
>>>>>>> >> > firefox to a halt. Profiling reveals a number of
>>>>>>
>>>>>>
>>>>>>
>>>>>> problematic
>>>>>>
>>>>>>> >> areas,
>>>>>>> >> > the biggest of which is getPx (of course).
>>>>>>> >> >
>>>>>>> >> > I've done some optimization that improves
>>>>>>> performance in
>>>>>>
>>>>>>
>>>>>>
>>>>>> this case:
>>>>>>
>>>>>>> >> >
>>>>>>> >> > * remove the call to Pixel.add(x,y) and directly
>>>>>>> modify the
>>>>>>
>>>>>>
>>>>>>
>>>>>> pixel
>>>>>>
>>>>>>> >> > position - this was a big savings ...
>>>>>>> >> >
>>>>>>> >> > * cached px location in the Point object based on
>>>>>>> the bounds
>>>>>>
>>>>>>
>>>>>>
>>>>>> of the
>>>>>>
>>>>>>> >> > containing vectorLayer so getPx is only called
>>>>>>> when the
>>>>>>
>>>>>>
>>>>>>
>>>>>> bounds have
>>>>>>
>>>>>>> >> > actually changed.
>>>>>>> >> >
>>>>>>> >> > attached is a patch against the sandbox to
>>>>>>> incorporate these
>>>>>>
>>>>>>
>>>>>>
>>>>>>> >> changes.
>>>>>>> >> >
>>>>>>> >> > Next steps for me are:
>>>>>>> >> >
>>>>>>> >> > * fix rendering in IE and Safari (that was the
>>>>>>> whole point
>>>>>>
>>>>>>
>>>>>>
>>>>>> of using
>>>>>>
>>>>>>> >> > excanvas)
>>>>>>> >> >
>>>>>>> >> > * further optimizations, including:
>>>>>>> >> >
>>>>>>> >> > o try caching the bounds of each vector and
>>>>>>> test for
>>>>>>> >> intersection
>>>>>>> >> > with layer bounds before rendering
>>>>>>> >> >
>>>>>>> >> > o tighten inner loops and remove deep call chains
>>>>>>
>>>>>>
>>>>>>
>>>>>> (possibly
>>>>>>
>>>>>>> >> > putting some of the pixel conversion stuff right into
>>>>>>
>>>>>>
>>>>>>
>>>>>> convertToPx)
>>>>>>
>>>>>>> >> >
>>>>>>> >> > o anything else you can think of
>>>>>>> >> >
>>>>>>> >> > * convenience functions for constructing polygons
>>>>>>> and lines
>>>>>>
>>>>>>
>>>>>>
>>>>>> from
>>>>>>
>>>>>>> >> > points without having to construct intermediate
>>>>>>> objects
>>>>>>> >> >
>>>>>>> >> > Other things that need to be done at some point,
>>>>>>> but not
>>>>>>> >> necessarily
>>>>>>> >> > for 2.3:
>>>>>>> >> >
>>>>>>> >> > * refactor the code (what, already?) so that the
>>>>>>> Vector
>>>>>>> >> architecture
>>>>>>> >> > can support multiple rendering types (Canvas, SVG,
>>>>>>> VML,
>>>>>>
>>>>>>
>>>>>>
>>>>>> possibly
>>>>>>
>>>>>>> >> > others in the future) by introducing Layer.Vector
>>>>>>> subclasses
>>>>>>
>>>>>>
>>>>>>
>>>>>> and a
>>>>>>
>>>>>>> >> > new Rendering class that knows how to render into the
>>>>>>
>>>>>>
>>>>>>
>>>>>> sub-classed
>>>>>>
>>>>>>> >> > layer types
>>>>>>> >> >
>>>>>>> >> > * refactor styling support to make it less
>>>>>>> specific to
>>>>>>
>>>>>>
>>>>>>
>>>>>> Canvas
>>>>>>
>>>>>>> >> > rendering stuff
>>>>>>> >> >
>>>>>>> >> > * add some Controls for digitizing and possibly
>>>>>>> editing
>>>>>>> >> >
>>>>>>> >> > Cheers
>>>>>>> >> >
>>>>>>> >> > Paul
>>>>>>> >> >
>>>>>>> >> > On 12-Oct-06, at 7:55 PM, Christopher Schmidt wrote:
>>>>>>> >> >
>>>>>>> >> > > On Thu, Oct 12, 2006 at 05:02:00PM -0400,
>>>>>>> Paul Spencer
>>>>>>
>>>>>>
>>>>>>
>>>>>> wrote:
>>>>>>
>>>>>>> >> > >> Schuyler, et al
>>>>>>> >> > >>
>>>>>>> >> > >> attached is a first cut at the vector
>>>>>>> stuff. It
>>>>>>
>>>>>>
>>>>>>
>>>>>> extracts
>>>>>>
>>>>>>> >> over an
>>>>>>> >> > >> existing openlayers folder and contains
>>>>>>> only the
>>>>>>
>>>>>>
>>>>>>
>>>>>> new files
>>>>>>
>>>>>>> >> I've added
>>>>>>> >> > >> plus a change to OpenLayers.js to load
>>>>>>> the extra
>>>>>>
>>>>>>
>>>>>>
>>>>>> files by
>>>>>>
>>>>>>> >> default
>>>>>>> >> > >> (except for excanvas for some reason ;)).
>>>>>>> >> > >>
>>>>>>> >> > >> There is a vector.html in the examples
>>>>>>> folder.
>>>>>>> >> > >
>>>>>>> >> > > This code, with some improvements from
>>>>>>> Schuyler and I,
>>>>>>
>>>>>>
>>>>>>
>>>>>> is now
>>>>>>
>>>>>>> >> in place
>>>>>>> >> > > in a sandbox in the OpenLayers SVN. You can
>>>>>>> check out
>>>>>>
>>>>>>
>>>>>>
>>>>>> with:
>>>>>>
>>>>>>> >> > >
>>>>>>> >> > > svn co http://svn.openlayers.org/sandbox/
>>>>>>> sderle/ vector/
>>>>>>> >> > >
>>>>>>> >> > > You can try it out at:
>>>>>>> >> > >
>>>>>>> >> > >
>>>>>>
>>>>>>
>>>>>>
>>>>>> http://openlayers.org/~crschmidt/vector/examples/vector.html
>>>>>>
>>>>>>> >> > >
>>>>>>> >> > > Click once to start drawing a line -- click
>>>>>>> again to
>>>>>>
>>>>>>
>>>>>>
>>>>>> end it.
>>>>>>
>>>>>>> >> > >
>>>>>>> >> > > FF only at the moment.
>>>>>>> >> > >
>>>>>>> >> > > Cool stuff, Paul, thanks for getting the
>>>>>>> ball rolling
>>>>>>
>>>>>>
>>>>>>
>>>>>> on this.
>>>>>>
>>>>>>> >> > > --
>>>>>>> >> > > Christopher Schmidt
>>>>>>> >> > > Web Developer
>>>>>>> >> >
>>>>>>> >> >
>>>>>>
>>>>>>
>>>>>>
>>>>>> +----------------------------------------------------------------- +
>>>>>>
>>>>>>> >> > |Paul Spencer
>>>>>>
>>>>>>
>>>>>>
>>>>>> pspencer at dmsolutions.ca |
>>>>>>
>>>>>>> >> >
>>>>>>
>>>>>>
>>>>>>
>>>>>> +----------------------------------------------------------------- +
>>>>>>
>>>>>>> >> > |Chief Technology Officer
>>>>>>
>>>>>>
>>>>>>
>>>>>> |
>>>>>>
>>>>>>> >> > |DM Solutions Group Inc
>>>>>>
>>>>>>
>>>>>>
>>>>>> http://www.dmsolutions.ca/ |
>>>>>>
>>>>>>> >> >
>>>>>>
>>>>>>
>>>>>>
>>>>>> +----------------------------------------------------------------- +
>>>>>>
>>>>>>> >> >
>>>>>>> >> >
>>>>>>> >> >
>>>>>>> >> >
>>>>>>> >> >
>>>>>>> >> >
>>>>>>> >> >
>>>>>>> >> > _______________________________________________
>>>>>>> >> > Dev mailing list
>>>>>>> >> > Dev at openlayers.org
>>>>>>> >> > http://openlayers.org/mailman/listinfo/dev
>>>>>>> >> >
>>>>>>> >> >
>>>>>>> >> >
>>>>>>> >> >
>>>>>>> >>
>>>>>>> >> <wkt.html>
>>>>>>> >> <WKT.js>
>>>>>>> > _______________________________________________
>>>>>>> > Dev mailing list
>>>>>>> > Dev at openlayers.org
>>>>>>> > http://openlayers.org/mailman/listinfo/dev
>>>>>>>
>>>>>>> +---------------------------------------------------------------- -+
>>>>>>> |Paul Spencer pspencer at dmsolutions.ca |
>>>>>>> +---------------------------------------------------------------- -+
>>>>>>> |Chief Technology Officer |
>>>>>>> |DM Solutions Group Inc http:// www.dmsolutions.ca/ |
>>>>>>> +---------------------------------------------------------------- -+
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> _______________________________________________
>>>>>>> Dev mailing list
>>>>>>> Dev at openlayers.org
>>>>>>> http://openlayers.org/mailman/listinfo/dev
>>>>>>>
>>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>
>>>>> +-----------------------------------------------------------------+
>>>>> |Paul Spencer pspencer at dmsolutions.ca |
>>>>> +-----------------------------------------------------------------+
>>>>> |Chief Technology Officer |
>>>>> |DM Solutions Group Inc http://www.dmsolutions.ca/ |
>>>>> +-----------------------------------------------------------------+
>>>>>
>>>>>
>>>>>
>>>>>
>>>>> _______________________________________________
>>>>> Dev mailing list
>>>>> Dev at openlayers.org
>>>>> http://openlayers.org/mailman/listinfo/dev
>>>>>
>>>>
>>>>
>>
>>
>> --
>> Cameron Shorter
>> http://cameron.shorter.net
>
>
> +-----------------------------------------------------------------+
> |Paul Spencer pspencer at dmsolutions.ca |
> +-----------------------------------------------------------------+
> |Chief Technology Officer |
> |DM Solutions Group Inc http://www.dmsolutions.ca/ |
> +-----------------------------------------------------------------+
>
>
>
>
>
--
Cameron Shorter
http://cameron.shorter.net
More information about the Dev
mailing list