def to_numpy()

in src/datasets/features/features.py [0:0]

• 34 lines of code
• 9 McCabe index (conditional complexity)

    def to_numpy(self, zero_copy_only=True):
        storage: pa.ListArray = self.storage
        null_mask = storage.is_null().to_numpy(zero_copy_only=False)

        if self.type.shape[0] is not None:
            size = 1
            null_indices = np.arange(len(storage))[null_mask] - np.arange(np.sum(null_mask))

            for i in range(self.type.ndims):
                size *= self.type.shape[i]
                storage = storage.flatten()
            numpy_arr = storage.to_numpy(zero_copy_only=zero_copy_only)
            numpy_arr = numpy_arr.reshape(len(self) - len(null_indices), *self.type.shape)

            if len(null_indices):
                numpy_arr = np.insert(numpy_arr.astype(np.float64), null_indices, np.nan, axis=0)

        else:
            shape = self.type.shape
            ndims = self.type.ndims
            arrays = []
            first_dim_offsets = np.array([off.as_py() for off in storage.offsets])
            for i, is_null in enumerate(null_mask):
                if is_null:
                    arrays.append(np.nan)
                else:
                    storage_el = storage[i : i + 1]
                    first_dim = first_dim_offsets[i + 1] - first_dim_offsets[i]
                    # flatten storage
                    for _ in range(ndims):
                        storage_el = storage_el.flatten()

                    numpy_arr = storage_el.to_numpy(zero_copy_only=zero_copy_only)
                    arrays.append(numpy_arr.reshape(first_dim, *shape[1:]))

            if len(np.unique(np.diff(first_dim_offsets))) > 1:
                # ragged
                numpy_arr = np.empty(len(arrays), dtype=object)
                numpy_arr[:] = arrays
            else:
                numpy_arr = np.array(arrays)

        return numpy_arr