import numpy as np

def process(shots, names, kinds):
    channel_indicies = [i for i, x in enumerate(kinds) if x == 'channel']
    out = np.full(len(channel_indicies)+1, np.nan)
    channel_indicies.pop(0)
    out_names = []
    # signal diff
    #            A B C D
    # chopper 1: - + + -
    # chopper 2: - - + +
    #   we want A-B+C-D
    c1 = shots[-2]
    c2 = shots[-1]
    a = np.mean(shots[0, (c1==-1)*(c2==-1)])
    b = np.mean(shots[0, (c1==+1)*(c2==-1)])
    c = np.mean(shots[0, (c1==+1)*(c2==+1)])
    d = np.mean(shots[0, (c1==-1)*(c2==+1)])
    out[0] = a-b+c-d
    out_names.append('signal_diff')
    # signal mean
    out[1] = np.mean(shots[0])
    out_names.append('signal_mean')
    # others
    for i in channel_indicies:
        out[i+1] = np.mean(shots[i])
        out_names.append(names[i])
    # finish
    return [out, out_names]