[QGIS-Developer] Is the measurement tool using proj library?

[Andreas Neumann]

Hi Nyall and others, 

I was wrong, there is no difference in behaviour between proj493 and
proj520. Nor is there a difference between Linux and Windows. 

We, as a user, made the mistake of creating test data in a new empty
layer at around (0,0) and my colleague in a totally different area in
the coordinate system in an area that isn't valid for the EPSG:2056
projection. The EPSG:2056 projection only works reasonably well in the
bound of approx. 2485000 1075000 2828500 13000000 

It produces bogus results if the user creates data outside of this
boundary, like at around 0,0. 

I was so embarrassed that it took me so long to find out our mistake.
For the longest time I thought it was some bug in QGIS or proj, only to
find out that both myself and my colleague created our test data outside
of the valid bounds of the projection at totally different locations,
but both outside of Switzerland. If we had some base layer below our
test data rather than just creating empty project with no background
maps, it wouldn't happen. 

When I moved the testdata into the valid bounds of the projection, the
area difference between planimetric and ellipsoidic area was less than a
square meter. 

For myself, I decided to never make fun again about other people not
understanding projections, given that I make such mistakes myself ... 

I wonder if it would make sense to issue a warning if a user creates
data outside of the valid bounds of a projection? It would definitely
make sense for stupid people like myself ... 

Thanks and greetings, 

Andreas 

On 2019-07-19 01:18, Nyall Dawson wrote:

> On Fri, 19 Jul 2019 at 00:38, Andreas Neumann <a.neumann at carto.net> wrote: 
> 
>> Hi,
>> 
>> We are having issues on the QGIS 3.4 release that the difference between and ellipsoidical and planimetric distance (and area) is huge. Too huge.
>> 
>> We have this issue on Windows 3.4.9, but on Linux (a bit newer commit) I don't have the same huge difference, but rather a very reasonable one.
>> 
>> I noticed that while the Windows version was built with proj 520 and the Linux version with 493, so a bit older.
>> 
>> But I don't know if the measurement tool in QGIS is using proj at all or not? If yes, could it be that newer proj releases made things worse regarding ellipsoidical measurements?
> 
> It's possible. The actual measurement calculation uses code ported
> from GRASS (and not the proj implementation of these algorithms), but
> proj IS used in transforming map/measure coordinates prior to
> calculating the distance. Possibly something has changed there which
> is impacting this.
> 
> Could you please test:
> 
> 1. Do you see the same distance values when measuring feature lengths?
> I.e. setup a layer with a feature with the same line as you are
> measuring, and check the identify tool derived attributes or $length
> value for it.
> 
> 2. Can you confirm that the project ellipsoid settings are identical
> between the windows & linux builds?
> 
> 3. Can you share an example I can test with?
> 
> Nyall
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/qgis-developer/attachments/20190719/6b269fc2/attachment-0001.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Ellipsoid_troubles_Sandra.zip
Type: application/zip
Size: 12199 bytes
Desc: not available
URL: <http://lists.osgeo.org/pipermail/qgis-developer/attachments/20190719/6b269fc2/attachment-0002.zip>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: ellipsoid_troubles.zip
Type: application/zip
Size: 9296 bytes
Desc: not available
URL: <http://lists.osgeo.org/pipermail/qgis-developer/attachments/20190719/6b269fc2/attachment-0003.zip>