[mapserver-dev] RFC 94: Shape Smoothing

[Alan Boudreault]

See comments...

On 13-03-30 12:20 PM, Stephen Woodbridge wrote:
> On 3/30/2013 11:26 AM, Alan Boudreault wrote:
>> Hi devs,
>>
>> I'd like to request some review on a pending RFC regarding support for
>> shape smoothing.
>>
>> Please see http://mapserver.org/development/rfc/ms-rfc-94.html
>>
>> It's a first draft, we might have to correct/improve some points.
>
> Overall, it is a very nice RFC and explains the problem well along with
> your proposed solution. Thanks.
>
> Questions:
>
> You mentioned that there are a lot of pathological cases and ways to
> work around them. I think that the reality will be that in a single file
> we will probably have some of all the bad cases so how can we deal with
> this likely case?
>
> Is there an easy (fast) way to guess what cases we are dealing with?
> Then set the parameters for that case?
>

Unless than experimenting with its data and trying different 
combinations of the available parameters, no. There are not that many.

> For example:
>
> * check if the start and end are the same: likely loops and apply the
> angle case
>
> * compute the average deviation from the smoothing average and ...
> * compute num_pnts/length over your window or your window+2 points and ...
>
> Maybe this is covered in the paper you referenced, I did not read it,
> but it seems that if we know some condition will generate bad results,
> the we should be able to also detect or guess that condition is in play
> for a given segment then apply the fix, maybe this is called the
> "auto_adjust" options.
>

auto_adjust is a very good idea. I'm just not 100% sure what should be 
the good conditions and behaviors for them. Note that if this cannot be 
determined at this moment, it can also be a future improvements. With 
more testers, we could find a better auto_adjust mode. I can also push a 
first implementation in my github fork and let you play with it if you 
want to help?

> Regardless, good work on this, and identifying the issues is half the
> problem of solving them.
>
> -Steve W
> _______________________________________________
> mapserver-dev mailing list
> mapserver-dev at lists.osgeo.org
> http://lists.osgeo.org/mailman/listinfo/mapserver-dev


-- 
Alan Boudreault
http://www.mapgears.com/