Is there a disadvantage in using the fft for this? If yes, then I think, I would be more in favor of adding a 1d np.fftconvolve, similar to the duplication 1d <-> nd between np.correlate/convolve and ...

% jc_convolve_hanning Smooth data by convolving a hanning function of length 'win'. % jc_convolve_hanning(x) uses a default window of 5.