The input to Mapper is a numpy.ndarray with double data type. If it is two-dimensional (say of shape (m,n) with m rows and n columns), the data is interpreted as vector data: m data points in \(\mathbb R^n\). If the array is one-dimensional, it is interpreted as pairwise distances. The array is the flattened, upper triangular part of the pairwise distance matrix. For m data points \(x_0,\ldots, x_{m-1}\), this vector contains \(\tbinom m 2=m(m-1)/2\) entries. The first entry is the distance \(d(x_0, x_1)\), the (m−1)-th entry is \(d(x_0, x_{m-1})\), the m-th entry is \(d(x_1, x_2)\), and so on.