def __call__()

in core/maxframe/tensor/arithmetic/clip.py [0:0]

• 50 lines of code
• 18 McCabe index (conditional complexity)

    def __call__(self, a, a_min, a_max, out=None):
        a = astensor(a)
        tensors = [a]
        sparse = a.issparse()

        if isinstance(a_min, Number):
            if a_min > 0:
                sparse = False
            a_min_dtype = np.array(a_min).dtype
        elif a_min is not None:
            a_min = astensor(a_min)
            tensors.append(a_min)
            if not a_min.issparse():
                sparse = False
            a_min_dtype = a_min.dtype
        else:
            a_min_dtype = None
        self.a_min = a_min

        if isinstance(a_max, Number):
            if a_max < 0:
                sparse = False
            a_max_dtype = np.array(a_max).dtype
        elif a_max is not None:
            a_max = astensor(a_max)
            tensors.append(a_max)
            if not a_max.issparse():
                sparse = False
            a_max_dtype = a_max.dtype
        else:
            a_max_dtype = None
        self.a_max = a_max

        if out is not None:
            if isinstance(out, Tensor):
                self.out = out
            else:
                raise TypeError(f"out should be Tensor object, got {type(out)} instead")

        dtypes = [dt for dt in [a.dtype, a_min_dtype, a_max_dtype] if dt is not None]
        dtype = np.result_type(*dtypes)
        # check broadcast
        shape = broadcast_shape(*[t.shape for t in tensors])

        setattr(self, "sparse", sparse)
        inputs = filter_inputs([a, a_min, a_max, out])
        t = self.new_tensor(inputs, shape)

        if out is None:
            setattr(self, "dtype", dtype)
            return t

        # if `out` is specified, use out's dtype and shape
        out_shape, out_dtype = out.shape, out.dtype

        if t.shape != out_shape:
            t = self.new_tensor(inputs, out_shape)
        setattr(self, "dtype", out_dtype)

        out.data = t.data
        return out