single classed layers and GetLegendGraphic

[Havard Tveite]

Yewondwossen Assefa wrote:

> Kralidis,Tom [Burlington] wrote:
>
>>> -----Original Message-----
>>> From: Yewondwossen Assefa [mailto:assefa at dmsolutions.ca]
>>> Sent: Monday, 08 November, 2004 10:30
>>> To: Kralidis,Tom [Burlington]
>>> Cc: MAPSERVER-DEV at LISTS.UMN.EDU
>>> Subject: Re: [UMN_MAPSERVER-DEV] single classed layers and
>>> GetLegendGraphic
>>>
>>>
>>> Is it possible for you to use the RULE parameter (set to the
>>> class name
>>> ) ? That should retun only the icon for the class.
>>> If not, I am not sure what the best solution is. Maybe add a
>>> support to
>>> a vendor specific request parameter that alaways return an
>>> icon for the
>>> first class. I personnaly prefer this way rather than having
>>> a diffrent
>>> behaviour when there is one or more than one classes. Comments ?
>>>
>>
>>
>> We can use RULE=, however how would a WMS client know what the
>> rule/class names are?
>>
>> I have received some other responses which favour this approach because
>> one can get a 'full' legend without the need to print out text, as it's
>> handled in the output graphic.  This makes sense in this context.
>>
>> As a result, what do people think about:
>>
>> - making GetLegendGraphic always return icon ONLY for single-class
>> layers, full for multiple classes (as per now)
>> - adding in VendorSpecificCapabilities called "GetLegend" which would
>> return a 'full' legend, titles and all for multiple layers, so someone
>> can get a single legend icon for multiple WMS layers from the same WMS?
>> We could also extend so that such a request can also ingest a Web Map
>> Context Document and output a single legend graphic.
>>
>> Comments?
>>
>
> I waited a bit for others to respond.
>
> I personally like the idea of giving the control to the user to output
> full legend or icons based on a specific VendorSpecificCapabilities.
>   Saying that the implementation of the GetlegendGraphic has been
> discussed lengthly and we ended up with what we have now. What you are
> proposing is a behaviour change just before a release. It is not too
> late to make this kind of change and if it is better we certainly should
> do it before the release (since it is the first relase of Mapserver with
> the GetlegnedGraphic), but others involved in it should certainly
> approve this.  I have  cc'd Bart and Valik who were invloved in the
> GetLegend development/dicsussion in case they are not the the dev list.

I do not know the specifications in detail, but in my opinion,
the behaviour of the GetLegendGraphic request should not depend
on the number of classes in a layer.

By returning the graphic signature together with the describing
text, the legend can be useful for all cases.
By returning the graphic signature only, it is possible for
"powerful" clients to add the explaining text for layers with
only one class, but for layers with several classes, this will
not be possible. For less powerful clients, the graphics without
text is useless.

We should only change the behaviour if it is in conflict with
the specifications.


--
Håvard Tveite
IMT, NLH