Add conversion of vector wrt to basis sets

Added the ability to convert the coefficients of a vector inside
one basis to another basis of possible different size.

iodata/overlap.py

@@ -28,7 +28,9 @@
 from .basis import HORTON2_CONVENTIONS as OVERLAP_CONVENTIONS
 
 
-__all__ = ['OVERLAP_CONVENTIONS', 'compute_overlap', 'gob_cart_normalization']
+__all__ = [
+    'OVERLAP_CONVENTIONS', 'compute_overlap', 'gob_cart_normalization', "convert_vector_basis"
+]
 
 
 def compute_overlap(obasis: MolecularBasis, atcoords: np.ndarray) -> np.ndarray:
@@ -159,3 +161,63 @@ def gob_cart_normalization(alpha: np.ndarray, n: np.ndarray) -> np.ndarray:
     vfac2 = np.vectorize(factorialk)
     return np.sqrt((4 * alpha)**sum(n) * (2 * alpha / np.pi)**1.5
                    / np.prod(vfac2(2 * n - 1, 2)))
+
+
+def convert_vector_basis(
+        coeffs1: np.ndarray,
+        basis2_overlap: np.ndarray,
+        basis21_overlap: np.ndarray
+) -> np.ndarray:
+    r"""
+    Convert a vector from basis 1 of size M to another basis 2 of size N.
+
+    Basis vectors are defined to be linearly independent wrt to one another
+     and need not be orthogonal.
+
+    Parameters
+    ----------
+    coeffs1:
+        Coefficients of the vector expanded in basis set 1. Shape is (M,).
+    basis2_overlap:
+        Symmetric matrix whose entries are the inner-product between basis vectors
+        inside basis set 2. Shape is (N, N).
+    basis21_overlap:
+        The overlap matrix between basis set 1 and basis set 2. Shape is (N, M).
+
+    Returns
+    -------
+    coeffs2 :
+        Coefficients of the vector expanded in basis set 2. Shape is (N,).
+
+    Raises
+    ------
+    ValueError :
+        If shapes of the matrices don't match the requirements in the docs.
+    LinAlgError :
+        If least squared solution does not converge.
+
+    Notes
+    -----
+    - `basis2_overlap` is the matrix with (i, j)th entries
+    :math:`\braket{\psi_i, \psi_j}` where :math:`\psi_i` are in basis set 2.
+
+    - `basis21_overlap` is the matrix whose (i, j)th entries are
+    :math:`\braket{\psi_i, \phi_j}` where :math:`\phi_j` are in basis set 1 and
+    :math:`\psi_i` are in basis set 2.
+
+    - If `basis2_overlap` is not full rank, then least-squared solution is solved instead.
+
+    """
+    if basis2_overlap.shape[0] != basis2_overlap.shape[1]:
+        raise ValueError("The `basis2_overlap` should be a square matrix.")
+    if np.any(np.abs(basis2_overlap.T - basis2_overlap) > 1e-10):
+        raise ValueError("The `basis2_overlap` should be a symmetric matrix.")
+
+    b = basis21_overlap.dot(coeffs1)
+    try:
+        # Try solving exact solution if `basis12_overlap` is full rank.
+        coeffs2 = np.linalg.solve(basis2_overlap, b)
+    except np.linalg.LinAlgError:
+        # Try least-squared solution.
+        coeffs2, _, _, _ = np.linalg.lstsq(basis2_overlap, b)
+    return coeffs2