def landscapes_approx(diag,p_dim,x_min,x_max,nb_nodes,nb_ld):
    """Compute a dicretization of the first nb_ld landscape of a 
    p_dim-dimensional persistence diagram on a regular grid on the 
    interval [x_min,x_max]. The output is a nb_ld x nb_nodes numpy
    array
    + diag: a persistence diagram (in the Gudhi format)
    + dim: the dimension in homology to consider
    """
    landscape = np.zeros((nb_ld,nb_nodes))
    diag_dim = []
    for pair in diag: #get persistence points for homology in dimension dim
        if (pair[0] == p_dim):
            diag_dim.append(pair[1])
    
    step = (x_max - x_min) / (nb_nodes - 1)
    #Warning: naive and not the most efficient way to proceed!!!!!
    for i in range(nb_nodes):
        x = x_min + i * step
        t = x / np.sqrt(2)
        event_list = []
        for pair in diag_dim:
            b = pair[0]
            d = pair[1]
            if b <= t <= d:
                if t >= (d+b)/2:
                    event_list.append((d-t)*np.sqrt(2))
                else:
                    event_list.append((t-b)*np.sqrt(2))
        event_list.sort(reverse=True)
        event_list = np.asarray(event_list)
        for j in range(nb_ld):
            if(j<len(event_list)):
                landscape[j,i]=event_list[j]
    
    return landscape