Source code for nifty8.operators.operator_adapter

# This program is free software: you can redistribute it and/or modify
# it under the terms of the GNU General Public License as published by
# the Free Software Foundation, either version 3 of the License, or
# (at your option) any later version.
#
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
# GNU General Public License for more details.
#
# You should have received a copy of the GNU General Public License
# along with this program.  If not, see <http://www.gnu.org/licenses/>.
#
# Copyright(C) 2013-2020 Max-Planck-Society
#
# NIFTy is being developed at the Max-Planck-Institut fuer Astrophysik.

from .linear_operator import LinearOperator




class OperatorAdapter(LinearOperator):
    """Class representing the inverse and/or adjoint of another operator.

    Objects of this class are created internally by `LinearOperator` whenever
    the inverse and/or adjoint of an already existing operator object is
    requested via the `LinearOperator` attributes `inverse`, `adjoint` or
    `_flip_modes()`.

    Users should never have to create instances of this class directly.

    Parameters
    ----------
    op : LinearOperator
        The operator on which the adapter will act
    op_transform : int
        1) adjoint
        2) inverse
        3) adjoint inverse
    """



    def __init__(self, op, op_transform, domain_dtype=float):
        self._op = op
        self._trafo = int(op_transform)
        if self._trafo < 1 or self._trafo > 3:
            raise ValueError("invalid operator transformation")
        self._domain = self._op._dom(1 << self._trafo)
        self._target = self._op._tgt(1 << self._trafo)
        self._capability = self._capTable[self._trafo][self._op.capability]

        try:
            import jax.numpy as jnp
            from jax import eval_shape, linear_transpose
            from jax.tree_util import tree_all, tree_map

            from ..nifty2jax import shapewithdtype_from_domain
            from ..re import Vector

            if callable(op.jax_expr) and self._trafo == self.ADJOINT_BIT:
                def jax_expr(y):
                    op_domain = shapewithdtype_from_domain(op.domain, domain_dtype)
                    op_domain = Vector(op_domain) if isinstance(y, Vector) else op_domain
                    tentative_yshape = eval_shape(op.jax_expr, op_domain)
                    if not tree_all(tree_map(lambda a,b : jnp.can_cast(a.dtype, b.dtype), y, tentative_yshape)):
                        raise ValueError(f"wrong dtype during transposition:/got {tentative_yshape} and expected {y!r}")
                    y = tree_map(lambda c, d: c.astype(d.dtype), y, tentative_yshape)
                    y_conj = tree_map(jnp.conj, y)
                    jax_expr_T = linear_transpose(op.jax_expr, op_domain)
                    return tree_map(jnp.conj, jax_expr_T(y_conj)[0])

                self._jax_expr = jax_expr
            elif hasattr(op, "_jax_expr_inv") and callable(op._jax_expr_inv) and self._trafo == self.INVERSE_BIT:
                self._jax_expr = op._jax_expr_inv
                self._jax_expr_inv = op._jax_expr
            else:
                self._jax_expr = None
        except ImportError:
            self._jax_expr = None


    def _flip_modes(self, trafo):
        newtrafo = trafo ^ self._trafo
        return self._op if newtrafo == 0 \
            else OperatorAdapter(self._op, newtrafo)



    def apply(self, x, mode):
        return self._op.apply(x,
                              self._modeTable[self._trafo][self._ilog[mode]])




    def draw_sample(self, from_inverse=False):
        if self._trafo & self.INVERSE_BIT:
            return self._op.draw_sample(not from_inverse)
        return self._op.draw_sample(from_inverse)




    def __repr__(self):
        from ..utilities import indent
        mode = ["adjoint", "inverse", "adjoint inverse"][self._trafo-1]
        res = "OperatorAdapter: {}\n".format(mode)
        return res + indent(self._op.__repr__())