[Python - persistence_graphical_tools] numpy (N x 3) a input could be convenient

[VincentRouvreau]

User reports they use numpy to load/save persistence diagrams with dimension information.

np.loadtxt('./rips.pers')                                                                                                                                          

array([[2.       , 0.138335 ,       inf],
       [1.       , 0.104347 ,       inf],
       ...,

       [1.       , 0.11169  , 0.111695 ]])

A patch to make it work:

def _array_handler(a):
    '''
    :param a: if array, assumes it is a (n x 2) or (n x 3) np.array and return a
                persistence-compatible list (padding with 0 or dimension), so that the
                plot can be performed seamlessly.
    '''
    if len(a[0]) == 3:
        if isinstance(a[0][2], np.float64) or isinstance(a[0][2], float):
            return [[int(x[0]), (x[1], x[2])] for x in a]
    if len(a[0]) == 2:
        if isinstance(a[0][1], np.float64) or isinstance(a[0][1], float):
            return [[0, x] for x in a]
    return a

Some documentation to be done.

[mglisse]

3 coordinates: dim, birth, death? That seems similar to a format that giotto likes, it is worth checking if that matches.

Another interesting format, produced for instance by the scikit-tda interface to ripser, is a list of nx2 arrays, one array per dimension.

At some point, with too many formats, it may become dangerous to guess, although for now it still seems doable.

Eventually, we may want to deprecate our current List[Tuple[int,Tuple[float,float]]], which is less convenient than the alternatives.