# Copyright (c) ONNX Project Contributors

# SPDX-License-Identifier: Apache-2.0


import numpy as np

from onnx.reference.ops._op import OpRunReduceNumpy


def compute_log_sum_exp(data, axes, keepdims):
    data_max = data.copy()
    ind = np.isinf(data_max)
    data_max[ind] = -np.inf
    mx = data_max.max(axis=axes, keepdims=True)
    sub = np.subtract(data, mx)
    exp = np.exp(sub, out=sub)
    mxs = np.sum(exp, axis=axes, keepdims=True, dtype=data.dtype)
    res = np.log(mxs) + mx
    if not keepdims:  # type: ignore
        res = np.squeeze(res, axis=axes)
    return (res,)


class ReduceLogSumExp_1(OpRunReduceNumpy):
    def _run(self, data, axes=None, keepdims=None):  # type: ignore
        tax = tuple(axes) if axes is not None else None

        if data.size == 0:
            return self.reduce_constant(data, -np.inf, tax, keepdims)
        return compute_log_sum_exp(data, tax, keepdims)


class ReduceLogSumExp_18(OpRunReduceNumpy):
    def _run(self, data, axes=None, keepdims=1, noop_with_empty_axes=0):  # type: ignore
        if self.is_axes_empty(axes) and noop_with_empty_axes:  # type: ignore
            return (data,)

        axes = self.handle_axes(axes)
        keepdims = keepdims != 0  # type: ignore

        if data.size == 0:
            return self.reduce_constant(data, -np.inf, axes, keepdims)

        return compute_log_sum_exp(data, axes, keepdims)