pbrod / nvector Goto Github PK

This might not be a nvector code problem, but a numpy issue. But I thought I would mention it anyway.

I have the following code:

ecef = [[ 1782029.5702637 ]
[-5958967.13026865]
[ 1407960.53954478]]

ell = nv.FrameE(name='WGS84')
p_EB_E = ell.ECEFvector(ecef)
pnt = p_EB_E.to_geo_point()
lon, lat, z = pnt.latlon_deg
print(lon, lat)
print(type(lat))

With Windows, I get the following output:
[12.83821967] [-73.35070153]
<class 'numpy.ndarray'>

with linux, I get this:
12.838219667803369 -73.35070153096443
<class 'numpy.float64'>

Any idea why one returns single element array vs a floating point?

As of 0.7.7, Cartopy is a required dependency of nvector. This is sometimes problematic since Cartopy does not provide ready-built wheels on pypi, thus requiring every user to have a proper build environment (including a build of GEOS) to do pip install nvector.

My suggestion would be to put the plotting part of nvector as a an optional dependency so that users can do pip install nvector to get the core functionality or pip install nvector[plots] (or similar) to also require Cartopy (and perhaps matplotlib).

Right now, the "easy way out" for me is to do: pip install "nvector<=0.7.6"

The intersect method on GeoPath objects claims to return an instance of GeoPoint, which seems like the desired behavior (at least, that is what I want it to return too). However it returns an instance of Nvector.

From the last line of the intersect method...

return frame.Nvector(normal_c, z=depth)

Python 3.9.2
nvector 0.7.7

Following Example 1:

import numpy as np
import nvector as nv
wgs84 = nv.FrameE(name='WGS84')
pointA = wgs84.GeoPoint(latitude=1, longitude=2, z=3, degrees=True)
pointB = wgs84.GeoPoint(latitude=4, longitude=5, z=6, degrees=True)
p_AB_N = pointA.delta_to(pointB)
print(p_AB_N)

gives

Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/home/f380cedric/.local/lib/python3.9/site-packages/nvector/objects.py", line 138, in __repr__
    params = fmt.join(['{}={!r}'.format(name, dict_params[name])
  File "/home/f380cedric/.local/lib/python3.9/site-packages/nvector/objects.py", line 138, in <listcomp>
    params = fmt.join(['{}={!r}'.format(name, dict_params[name])
  File "/home/f380cedric/.local/lib/python3.9/site-packages/nvector/objects.py", line 138, in __repr__
    params = fmt.join(['{}={!r}'.format(name, dict_params[name])
  File "/home/f380cedric/.local/lib/python3.9/site-packages/nvector/objects.py", line 138, in <listcomp>
    params = fmt.join(['{}={!r}'.format(name, dict_params[name])
KeyError: 'point'

It goes like this: Pvector.__dict__["frame"] → FrameN.__dict__["point"]. FrameN._NAMES gives "point" while dict_params only contains key nvector.

Hi,

Using the latest version 0.7.5 the below example fails:

wgs84 = nv.FrameE(name='WGS84')
point1 = wgs84.GeoPoint(latitude=88, longitude=0, degrees=True)
point2 = wgs84.GeoPoint(latitude=89, longitude=-170, degrees=True)
s_12, _azi1, _azi2 = point1.distance_and_azimuth(point2)
p_12_E = point2.to_ecef_vector() - point1.to_ecef_vector()
d_12 = p_12_E.length[0]
msg = 'Ellipsoidal and Euclidean distance = {:5.2f} km, {:5.2f} km'
msg.format(s_12 / 1000, d_12 / 1000)

The error given is:

Attribute Error: 'list' object has no attribute 'T'

The example works correctly with version 0.7.4

Hello,

First of all, thank you for making this package opensource. It is very useful to me.
I came across the following issue:
on_path when method='exact' will not return consistently True on the positionA of a path.
Below is an illustrating example. Having had a quick look at the code under the hood, it may be the checking the difference of azimuth is not meaningful when point_a==point

import nvector as nv
import numpy as np

wgs84 = nv.FrameE(name='WGS84')
pointA = wgs84.GeoPoint(latitude=81, longitude=0, degrees=True)
pointB = wgs84.GeoPoint(latitude=80, longitude=0, degrees=True)
path = nv.GeoPath(pointA, pointB)
print(np.allclose(path.on_path(path.positionA, method='exact'), True))

wgs84 = nv.FrameE(name='WGS84')
pointA = wgs84.GeoPoint(latitude=79, longitude=0, degrees=True)
pointB = wgs84.GeoPoint(latitude=80, longitude=0, degrees=True)
path = nv.GeoPath(pointA, pointB)
print(np.allclose(path.on_path(path.positionA, method='exact'), True))

nvector.version == 0.7.4

I have run the same test as in test_n_E_and_wa2R_EL using the reference Matlab implementation (using Octave) and the result is not the one used as Python test

        R_EL1 = [[0, 1.0, 0],
                 [1.0, 0, 0],
                 [0, 0, -1.0]]

but rather

>> n_E_and_wa2R_EL([0 0 1]', wander_azimuth = (pi / 2))
ans =

  -6.1230e-17  -1.0000e+00   6.1230e-17
  -1.0000e+00   6.1230e-17  -7.4983e-33
   3.7492e-33  -6.1230e-17  -1.0000e+00

HTH

Hi,

I'm using this code for calculating the cartesian vector from "origin" to "target":
` wgs84 = nv.FrameE(name='WGS84')
origin = wgs84.GeoPoint(latitude=latitudeOrigin, longitude=longitudeOrigin, z=altitudeOrigin, degrees=False)
target = wgs84.GeoPoint(latitude=latitudeTarget, longitude=longitudeTarget, z=altitudeTarget, degrees=False)

    p_AB_N = origin.delta_to(target)
    x, y, z = p_AB_N.pvector.ravel()

`

Now, how can I reverse this operation? I have an origin and a cartesian vector and like to find the vector position as a GeoPosition (in my example "target").

In Pseudo-Code:
target = origin.to_geo_position(np.r_[x,y,z].reshape((-1, 1)))

Thank you,
Manfred

Would appreciate install availability with conda-forge. It's a little more stable with a conda environment than pypi. Thanks.

Hello, thanks for this great library! Would it be possible to release a new package on PyPi with documentation built using Sphinx>=3.0.4? The newer version uses an updated version of jQuery (3.5) that addresses a security vulnerability.

Hi, and thanks for this really nice concept and library. The paper is really interesting!

I ran into an inconsistency of return types from the GeoPath.track_distance-method:

• GeoPath.track_distance(method='great_circle') returns a single-element array
• GeoPath.track_distance(method='euclidian') returns a single-element array
• GeoPath.track_distance(method='exact') returns a scalar

From my point of view I would prefer it to consistently return scalars, that makes the most sense to me.

According to numpy documentation (v1.19), isclose should not be used with default atol value for comparing numbers close to 0. See https://numpy.org/doc/stable/reference/generated/numpy.isclose.html

However, this is what is done in function on_great_circle in _core.py (and maybe other areas).

Hi,

I was doing some quick experimentation with nvector, and found this calculation of the cross-track distance on the great circle.

The arccos function is inaccurate for small angles, as mentioned in the comment. When this function is called from cross_track_distance however, it is taken between the cross product of the n-vectors describing the path (c^E) and the position B. For small cross-track errors, shouldn't the expression with arccos be the most accurate (since arcsin is inaccurate at -pi/2 and pi/2)?

• Øystein