[mapserver-dev] RFC-60: Label enhancement to skip ANGLE FOLLOW labels with overlapping chars

[Jeff McKenna]

Hi Daniel and Alan,

Thanks for producing this RFC and doing the background tests.  (I 
especially like the map image showing the BBoxes of the individual label 
characters, which really helps to understand the angle between some 
label characters 
(http://trac.osgeo.org/mapserver/attachment/ticket/3523/mapserver-label-boxes.png).

I have the benefit of sitting beside Andrea from GeoServer right now, 
and I have asked him specific details of how GeoServer handles this same 
issue.  As I understand, this is how it works for GeoServer:
1) user can modify angle to use between characters with a MAXANGLEDELTA 
parameter 
(http://docs.geoserver.org/2.0.x/en/user/styling/sld-reference/labeling.html#labeling-max-angle-delta)
2) by default GeoServer uses a hardcoded value of something like 25 
(Andrea wasn't sure about exact value at the time) and discards labels 
with angles higher than that

So, regarding RFC-60, I support the addition of the MAXOVERLAPANGLE 
parameter (I like this name better than GeoServer's, at least I can 
understand what the parameter means by the name).

However I disagree with the RFC's statement "The default behavior of the 
software if this parameter is not set will remain unchanged, i.e. this 
test will be disabled by default and no labels will be skipped.".  I 
don't see how any user would ever want a high angle between their label 
characters, so let's set a default value (22.5) and discard labels with 
higher angles between characters.

Aside: in the case of labelling contours for the benchmarking exercise, 
Andrea showed me his config file and he sets a MAXANGLEDELTA of 30.

-jeff



On 10-08-26 10:25 PM, Daniel Morissette wrote:
> Hi Devs,
>
> As part of the FOSS4G 2010 Benchmarking exercise, we have discovered
> that MapServer did a poor job at labeling contour lines whereas others
> such as GeoServer didn't suffer of the same problem. It turns out that
> GeoServer detects and skips those problematic labels, leaving room for
> other/better labels to fall in the spot left available by the skipped label.
>
> Alan has been doing some experiments around this problem and we have
> come up with a RFC that we would like to discuss. To make a long story
> short we are proposing that we do the same as GeoServer and offer an
> optional mechanism to detect and skip those labels.
>
> Please have a look at RFC-60 and let us know what you think:
>
>    http://mapserver.org/development/rfc/ms-rfc-60.html
>
> The discussions in ticket 3523 may also be of interest:
>
>    http://trac.osgeo.org/mapserver/ticket/3523
>
> And finally there is also ticket 2221 discussing more general labeling
> enhancements that this RFC does not aim to address:
>
>    http://trac.osgeo.org/mapserver/ticket/2221
>