[Qgis-developer] QGIS Crash - Serious problem in 2x

[Bernhard Ströbl]

Hi Matthias,

Am 24.06.2014 16:46, schrieb Matthias Kuhn:
> On 24.06.2014 16:10, Bernhard Ströbl wrote:
>>
>> Hi Matthias,
>>
>> In your mail I attached two files (not for the list). One is North
>> America in a Lambert Conformal Conic projection, the other one in
>> WGS84. Both show a grid with 10 degrees distance between parallels and
>> meridians.
>>
>> Am 24.06.2014 14:02, schrieb Matthias Kuhn:
>>> Hi Bernhard,
>>>
>>> On 24.06.2014 11:06, Bernhard Ströbl wrote:
>>>> Hi Matthias,
>>>>
>>>> probably this is academical...
>>>>
>>>> Am 24.06.2014 10:42, schrieb Matthias Kuhn:
>>>>> Hi Bernhard,
>>>>>
>>>>> I wouldn't say no sense at all. It strongly depends on the
>> context, but
>>>>> if you have for example a lesson for geography students and are
>>>>> introducing CRS/projections and their properties one could want to
>> add a
>>>>> scale bar in degrees.
>>>>
>>>> But would that scalebar show the degrees for lon or lat?
>>>
>>> Maybe I am wrong, but I assume that there is no difference. One unit
>>> (degree) will represent the same amount of pixels/points horizontal and
>>> vertical.
>>
>> Well, you are wrong because one degree in lat is always ~110km whereas
>> one degree in lon is ~110 km at the equator and e.g. in Zurich ~75km
>> (for calculation see [1]). So how many pixels are 1 degree? Depends on
>> the projection; in WGS84 it is the same amount no matter if for lon or
>> lat, in Lambert it is not.
>
> For Lambert neither one nor the other makes sense. The only appropriate
> solution here would be some kind of "Lambert unit" whatever that may be.
> But for the WGS84 map you sent the original statement above holds true:
> the same amount of pixels matches the same amount of degrees everywhere
> on the map (In terms of lat/lon, not in terms of degrees on the sphere
> though). While a scalebar in km as provided is subject to distortion for
> the exact reason you noted.

Agreed. But a scale bar is used to measure distances (and IMHO distances 
are in miles, km,..., not in degrees) If a scale bar makes sense depends 
on the projection and the area covered (as I stated some mails ago: "The 
fact that a map is suitable to measure and compare distances is not 
decided by the map units but by the used projection and the covered 
area."). If it does not make sense one should not put a scale bar on the 
map.

>
>>
>>>
>>>> If the first (lon) for which latitude?
>>> It doesn't matter in degrees. But it really matters when trying to
>> put a
>>> scalebar in meters.
>>
>> It does also matter in degrees, depending on the projection. same in
>> meters: 1 cm on the map represents always a certain distance in
>> reality (though this distance varies troughout the map depending on
>> the projection and the area covered). If you look at the Lambert map,
>> you realize that the distance between two parallels (10 degrees!)
>> increases towards the pole, although in reality it is always (10*110km
>> =) 1100 km. In the WGS84 map the distance between the parallels is
>> constant but so is the distance between the meridians, but this is
>> false as the distance gets less towards the pole in reality. So a
>> scalebar (in m) being accurate in the middle of the map becomes less
>> accurate towards the edges. Hence my question on which base the
>> scalebar is calculated.
>
> The question absolutely makes sense but I don't know the answer :)

Could you check? or whom would we have to ask?

>
>>
>>>> Either of the two: how do you want to tell people that this scalebar
>>>> is only true for North-South (lat) or East-West (lon) measurements and
>>>> must not be used in any other direction? IMHO a scale bar is to enable
>>>> readers to use their ruler to measure a distance on the map in _any_
>>>> direction.
>>>>
>>>>> I agree that it's not very common and most people
>>>>> are probably unused, but if you explicitly state the fact that the
>> map
>>>>> is in degrees you might even avoid confusion and prevent people from
>>>>> trying to compare distances.
>>>>
>>>> But adding a scale bar encourages users to compare distances! The fact
>>>> that a map is suitable to measure and compare distances is not decided
>>>> by the map units but by the used projection and the covered area. If
>>>> your map is in degrees just enable the graticules and (if useful) add
>>>> a scalebar in m/km/miles (does that work with degrees? I have not
>>>> tried. If not this would be a feature request.)
>>>>
>>> Doesn't really make sense to me. Graticules are just another reference
>>> for distances (in degrees in this case) and an alternative or addition
>>> to scale bars. What problem exactly would the combination of a grid in
>>> degrees and a scalebar in meters solve?
>>
>> a scale bar makes distances measurable while a graticule helps
>> localizing a point. In certain cases (projections) the graticule could
>> be used for measuring, though.
> You are right here. It's not a replacement but a bit of a different
> thing (because graticules are not necessarily required to be straight).
> So for the ship navigation example before, graticules in degrees for
> localization (non-straight) combined with a meaningful scalebar (given a
> suitable projection) make absolutely sense. My main point was, that a
> graticule doesn't compensate for a scalebar on an unsuitable projection.

OK, total agreement, all is related to the projection, so the maker of 
the map has to use care to choose a suitable one.

>>
>>>
>>>> BTW: for which point of a map is the scale bar currently created
>>>> (thinking of non-distance-true projections and large areas e.g.
>>>> continents)?
>>> No idea.
>>> But if there should be proper support for scalebars in meters on
>>> degree-based maps, then it has to be configurable. And also the two
>>> different scalebars (horizontal vs. vertical) that you mentioned. Then
>>> it could be that there is a small enough area that this can be
>>> considered accurate enough to be useful. And there should be
>> warnings to
>>> inform the mapper that he might be misleading readers and should
>>> consider to reproject.
>>
>> I cannot imagine any use case for measuring distances in degrees. If
>> you look at either of my maps you see that the south of Greenland is
>> apporximately 40 degrees north of Cuba and that Canada covers almost
>> 90 degrees in east-west direction. But why should someone measure this
>> in degrees and not in km or miles? Would you measure the distance
>> between your place and your favourite bar in degrees?
>
> That's not the right question to ask. Instead it should be: "Why would
> someone want to measure distances on an unsuitable projection, with a
> ruler that he has no idea if it has any meaning for the location he
> tries to measure?".

Again, its up to the maker of the map to provide a scalebar if it is 
meaningful and none otherwise, same with the graticule.

> The result is the same as when a friend measures the distance to his
> favorite bar, find out that it's 2.8 miles and then telling me that it's
> 2.8 km. I'd rather know that it's 2.8 "units" and therefore be aware of
> the requirement to be careful with the interpretation of the result.
> Or even better: have somebody warn my friend that he might be using the
> wrong tool for the job he tries to accomplish :)

or even better: go with him to the bar to show you the way :-)

Bernhard


__________ Information from ESET Mail Security, version of virus signature database 9993 (20140624) __________

The message was checked by ESET Mail Security.
http://www.eset.com