jitter_coords.py

import numpy as np


def main():
    coord_and_names = np.genfromtxt('coordinates.txt',
                                    dtype=[('lat', 'float'), ('lng', 'float'), ('name', 'U20')],
                                    delimiter=',', skip_header=True)

    # make lat and lng into 2D array
    coords = np.column_stack((coord_and_names['lat'], coord_and_names['lng']))

    uniques, indexes, counts = np.unique(coords, axis=0, return_index=True, return_counts=True)

    # get mask for all rows in coords not in indexes[counts == 1]
    # meaning, we're creating a mask matching coord_and_names (or coords), that
    # is True for all duplicated coordinates.
    duplication_mask = np.ones(coords.shape[0], dtype=bool)
    duplication_mask[indexes[counts == 1]] = False

    # this will now only affect the duplicates
    coords[duplication_mask] += np.random.uniform(-0.001, 0.001, coords[duplication_mask].shape)

    # save the new coordinates
    # https://stackoverflow.com/a/26019097/3717467
    from numpy.lib.recfunctions import append_fields
    latlng = np.rec.array(coords, dtype=[('lat', 'float'), ('lng', 'float')])
    mixed_coords = append_fields(latlng, 'name', coord_and_names['name'], usemask=False)

    # alternatively, only store the jittered coordinates:
    # latlng = np.rec.array(coords[duplication_mask], dtype=[('lat', 'float'), ('lng', 'float')])
    # mixed_coords = append_fields(latlng, 'name', coord_and_names['name'][duplication_mask], usemask=False)

    np.savetxt('jittered_coordinates.txt', mixed_coords,
               delimiter=',', fmt=['%.9f', '%.9f', '%s'],
               header='lat,lng,name', comments='')


if __name__ == '__main__':
    main()