crop error of real track

[caver456]

First guess: maybe this is due to a flaw in removeSpurs.. The log shows that several spurs were removed, including at coordinates very close to the split. Try the same crop operation without doing removeSpurs first.

[caver456]

commented out the removeSpurs call for the tracks, and the crop error persists - for all tracks:

That's more than a coincidence... looking in the log, there's now just one occurrence of 'spur removed' and it appears to be talking about a spur on the boundary shape at the opposite corner of the assignment:

That might or might not be relevant. But - take a look at the corner below the bad crops - it's a crossover, so that is probably where the problem resides. Looks like the 'beyond' argument is resulting in an inward growth where the crossover happens, rather than an outward growth. So - need to look at the coordinates at each step around the buffer command, around line 1961:

        if boundaryType=='Polygon':
            cgc=cg['coordinates'][0]
            cgc=self.removeSpurs(cgc)
            boundaryGeom=Polygon(cgc).buffer(beyond) # Shapely object

[caver456]

Sure enough, adding the grown boundary to the sartopo map shows it:

[caver456]

The shapely docs say that doing a buffer of zero will split bowties into two polygons at the crossover point. So, split the bowties, do a buffer on each resulting polygon, then OR the results together into one polygon.

Holy cow that worked:

    def buffer2(self,boundaryGeom,beyond):
        a=boundaryGeom.buffer(0) # split bowties into separate polygons
        merged=unary_union(a)
        return merged.buffer(beyond)

...
in crop:
            # boundaryGeom=Polygon(cgc).buffer(beyond) # Shapely object
            boundaryGeom=self.buffer2(Polygon(cgc),beyond)

[caver456]

if this fails in the future, note that the shapely docs say that results may vary when trying to fix bowties using zero size buffer and also when using unary_union, and that some users reported better success when using small positive values. So, if needed, could try progressively larger buffer values until is_valid is true for each resulting polygon, or something to that effect.