Question re angle wrapping

[ea42gh]

Should gv.Path unwrap the angles internally,
or should it be up to the user calling gv.Path?

[philippjfr]

It should handle longitude wrapping but out-of-bounds latitudes are not valid. I also doubt it will handle wrapping from paths which span more than 360 degrees in longitude properly.

[jbednar]

Not sure what you mean by out of bounds latitudes; those points all seem to be between +/- 60 in latitude.

[philippjfr]

Yes, I was just very confused by the plot on the left, I see now what it represents.

[ea42gh]

When you say it should handle longitude wrapping, do you intend to
modify gv.Path to remove that horizontal line?
Or is it still up to the user of hv.Path?
(Note the path in the figure is a great circle)

[philippjfr]

When you say it should handle longitude wrapping, do you intend to modify gv.Path to remove that horizontal line?

The project operation (which is called automatically) should do that for you. Could you provide the code used to generate the example? Discussing things in the abstract is rarely helpful.

[ea42gh]

lon = np.linspace(90, 360+90, 100)%360.
lat = np.linspace(-60,60, 100)
gf.coastline*gv.Path([(lon,lat)])+hv.Curve(lon)*hv.Curve(lat)

and

lon = np.linspace(0, 360, 100)
lat = np.linspace(0,180+60, 100) % 180 - 90
gf.coastline*gv.Path([(lon,lat)])+hv.Curve(lon)*hv.Curve(lat)

The problem is the branch cut: gv.Path draws connections that should not be there.
A fix would be to break the path into pieces each time we cross the boundary,
with some care taken to draw the Path to the edge of the plotted geographical area.

Nasty piece of code that calls for a utility function...

[philippjfr]

I think the problem here is simply that you are wrapping the coordinates with the modulo, if you leave off the modulo it will wrap it correctly for you:

lon = np.linspace(90, 360+90, 100)
lat = np.linspace(-60,60, 100)
gf.coastline*gv.Path([(lon,lat)])+hv.Curve(lon)*hv.Curve(lat)

[ea42gh]

It misses half the path that way

[philippjfr]

Okay, I see it now. I'll have to look into what cartopy is doing there.

[ea42gh]

continuity of paths (contd) Consider the following example

class PlaneWatcher(param.Parameterized):
    step = param.Integer(0, bounds=(0,None))
    
    def __init__(self, **params ):
        self.lonlat = (np.linspace(0, 40, 5), np.linspace(-90,10, 5))
        self.proj   = crs.LambertConformal()
        #self.proj = crs.PlateCarree()
        super(PlaneWatcher,self).__init__(**params)

    @param.depends("step")
    def view_plane(self):
        self.lonlat = ((self.lonlat[0]+1+180)%360-180, (self.lonlat[1]+1 + 90)%180-90)
        h = gf.coastline*gf.ocean *\
            gv.Points( self.lonlat ).options(size=5,color='darkblue', tools=['hover'])*\
            gv.Path( [self.lonlat] ).options(color='red', projection=self.proj)
        return h.relabel('step=%d'%self.step).options( global_extent=True, width=500, height=475, projection=self.proj )
       
    def view(self):
        return hv.DynamicMap( self.view_plane )

obj = PlaneWatcher()

player = Player(length=1000, interval=10)
player.width=400
player.loop_policy = 'loop'
player.link(obj, value='step')
pn.Column(obj.view(), pn.Row(pn.Spacer(width=200),player))

If you leave this running long enough, you will also see points coming in from the lower left
from outside the projected image

[ea42gh]

[ea42gh]

Don't know what happens to the coastline with proj LambertCylindrical()
Note quite a number of the projections also unnecessarily loose axes

[philippjfr]

Should be fixed by #278

lon = np.linspace(90, 360+90, 100)
lat = np.linspace(-60,60, 100)
gf.coastline*gv.Path([(lon,lat)])+hv.Curve(lon)*hv.Curve(lat)

[ea42gh]

Changed the lon/lat positioning logic:
I would like to keep the line connecting the points as the motion proceeds.

class PlaneWatcher(param.Parameterized):
    step = param.Integer(0, bounds=(0,None))
    
    def __init__(self, **params ):
        self.lon       = np.linspace(0, 40, 5)
        self.lat       = np.linspace(-90,10, 5)
        self.delta_lat = np.ones(5)

        #self.proj   = crs.LambertConformal()
        self.proj = crs.PlateCarree()
        super(PlaneWatcher,self).__init__(**params)

    @param.depends("step")
    def view_plane(self):
        self.lon += 1
        self.lat += self.delta_lat
        wrap = self.lat>90
        self.delta_lat[wrap]=-1
        self.lat[wrap] = 180-self.lat[wrap]

        wrap = self.lat<-90
        self.delta_lat[wrap]=1
        self.lat[wrap] =-180-self.lat[wrap]

        h = gf.coastline*gf.ocean *\
            gv.Points( (self.lon, self.lat) ).options(size=5,color='darkblue', tools=['hover'])*\
            gv.Path( [np.stack([self.lon,self.lat],axis=1)] ).options(color='red', projection=self.proj)
        return h.relabel('step=%d'%self.step).options( global_extent=True, width=500, height=475, projection=self.proj )
       
    def view(self):
        return hv.DynamicMap( self.view_plane )

obj = PlaneWatcher()

player = Player(length=1000, interval=10)
player.width=400
player.loop_policy = 'loop'
player.link(obj, value='step')
pn.Column(obj.view(), pn.Row(pn.Spacer(width=200),player))

[philippjfr]

With the small patch in #279 I get this: