I’m building a project for the Raspberry Pi that turns a relay on and off random times in a specific time window. To manage the time slots, I want to use a two-dimensional array that’s generated daily. So, my Python application needs to empty the previous day’s array, and populate it with the on/off time slots being generated randomly. I can’t figure out how to append my time values to the array. Can someone help me? Here’s my code:

  daily_slots = np.empty([1], dtype=[('onTime', np.dtype(int)), ('offTime', np.dtype(int))])# numpy populates the array with whatever is in memory at that time,# so delete the existing, 'empty' array rowdaily_slots = np.delete(daily_slots, 0, 0)

With that in place, how do I append values? The numpy documentation says I should do something like the following:

daily_slots = np.append(daily_slots, [700, 800])

But that’s not working, I get

Traceback (most recent call last):File "./controller.py", line 351, in <module>init_app()File "./controller.py", line 128, in init_appbuild_daily_slots_array()File "./controller.py", line 307, in build_daily_slots_arraydaily_slots = np.append(daily_slots, [700, 800])File "/usr/lib/python2.7/dist-packages/numpy/lib/function_base.py", line 3555, in appendreturn concatenate((arr, values), axis=axis)
TypeError: invalid type promotion

I've tried everything I can think of and I've poked and prodded at the numpy documentation and I've not found the answer.

Am I not declaring the array correctly? I want each array 'row' to consist of a two-element array (the ontime and offtime).

np.append uses np.concatenate. It isn't a true copy of list append. It turns the new value into an array and does concatenate. But to do concatenate with structured arrays, all arrays have to have the same dtype (or at least compatible ones, hence the 'promotion' error.

This works:

In [4]: np.append(daily_slots, np.array((700, 800), dtype=daily_slots.dtype))
Out[4]: 
array([(  0, 1075970048), (700,        800)], dtype=[('onTime', '<i4'), ('offTime', '<i4')])

This also works. I had to add the [] so that the 2nd array was 1d like the first. Without them it is 0d - in effect that's all that np.append adds to the game.

In [6]: np.concatenate((daily_slots, np.array([(700, 800)], dtype=daily_slots.dtype)))
Out[6]: 
array([(  0, 1075970048), (700,        800)], dtype=[('onTime', '<i4'), ('offTime', '<i4')])

For growing and shrinking, a list is usually better. In this case a list of 2 element tuples, or may be a list of a custom class.

If it is a list of tuples, you could wrap it in np.array to turn it into a 2d array for ease of doing calculations.