Hi Even,

Thank you for the detailed response, please find my comments inline below:

(I'd appreciate if RFCs could also be kept proportional (=small) in length
> to the change being discussed to remain accessible to humans that don't use
> LLMs)
>
That makes sense, I'll follow that policy in the subsequent efforts :-)

> - I'm wondering if we really need the Cartesian behavior to be preserved.
> Does it make sense at all ? if not this could save us and users from
> handling an extra configuration parameter. I'd say if the project has a CRS
> use the geodetic scale, otherwise fallback to Cartesian method.
>
Yes, it makes sense as a product/API simplification, but I would be
cautious about removing the configurability entirely. I'm quite afraid of
breaking backwards compatibility for a minor release like 8.8.  Existing
MapServer users may have scalebars in projected CRSs where the current
cartesian behavior is expected, visually tested, or intentionally tied to
projected coordinates. Switching all projected maps to geodesic by default
could change rendered output in a minor-looking but user-visible way. That
is especially sensitive for regression tests, printed map workflows etc. I
think the geodesic default might be the right long term direction for the
next major release, where the backwards compatibility is not so important.

> - I'm not sure it makes sense to correlate the sample position with the
> location of the scale bar. Looking at QGIS scalebar in layout, it has the
> following options "average top , middle and bottom scales", "top of map",
> "middle of map", "bottom of map", "always at the equator".   We don't
> necessarily need to offer all that complexity, so perhaps just middle of
> the map ?
>
I can agree with the proposal. For MapServer’s first implementation,
center-of-map matches the motivating OpenLayers-style use case (according
to https://github.com/MapServer/MapServer/issues/7397) and avoids
surprising changes when users move the scalebar visually.
>
> - the geodesic mode, if taking into account distances along the X axis, as
> I see in
> https://github.com/MapServer/MapServer/pull/7507/changes#diff-968c7ca959d226116368f37b62faf245fff8049e3e29fd03ab6914766d61d3d5R238
>  doesn't
> always make sense because the scale, even when mapping locally can be very
> different along the horizontal and vertical axis. Conformal projections
> (Mercator, Transverse Mercator, LCC, etc.) have indeed identical scale in
> both directions, but if you use let's say Equal Earth that can be very
> different. Maybe we should use the geometric average sqrt ( scale_along_x *
> scale_along_y )  which should yield at least something better if doing area
> measurement, which will also often be the Cartesian distance. If we don't
> do that, we should be very clear in the documentation that the value only
> makes sense in the horizontal direction.
>
> Ex: Equal Earth, 10 km in projected distance along X and Y:
>
> $ echo  0 7700000 | gdaltransform  -s_srs +proj=eqearth  -t_srs EPSG:4326
> 0 70.1915580220322 0
> $ echo  0 7710000 | gdaltransform -s_srs +proj=eqearth  -t_srs EPSG:4326
> 0 70.3458420298602 0
> $ echo 10000 7700000 | gdaltransform -s_srs +proj=eqearth  -t_srs
> EPSG:4326
> 0.153998673719061 70.1915580220322 0
> $ echo 70.1915580220322 0 70.3458420298602 0 | geod -I +ellps=WGS84
> 0d    -180d    17212.757
> $ echo 70.1915580220322 0 70.1915580220322 0.153998673719061  | geod -I
> +ellps=WGS84
> 89d55'39.204"    -89d55'39.204"    5826.666
>
> So the 10 km have become 17.2 along X and 5.83 along Y.    But sqrt(17.2 *
> 5.83) = 10.0 km
>
This is a reasonable point technically, but it would open up further
technical considerations and decision requirements. With the current
solution, I just wanted to address the  common web mapping use cases, like
the Web Mercator/local-map use case where a horizontal scalebar should
reflect the real ground distance at the current map location, instead of
using distorted projected units. For that purpose, measuring geodesic
distance along the scalebar’s horizontal axis is natural, because it keeps
the rendered bar literal: its on-screen length corresponds to the ground
distance between its left and right endpoints.
Using a geometric average of horizontal and vertical scale could be useful
as a representative or area-oriented scale, especially for non-conformal
projections, but it would no longer describe the actual horizontal length
of the scalebar. So if we keep the current approach, we should document
clearly that geodesic scalebars represent horizontal distance at the sample
point, not vertical distance or general area scale.
>
>
Best regards,

Tamas
_______________________________________________
MapServer-dev mailing list
[email protected]
https://lists.osgeo.org/mailman/listinfo/mapserver-dev

Reply via email to