Use ndarray.size instead of numpy.prod

[zariiii9003]

The numpy.prod() call seems quite expensive:

The _calculate_num_samps_per_chan() call could be skipped entirely if the user requested to read all available data (=-1), right?

[zhindes]

The _calculate_num_samps_per_chan() call could be skipped entirely if the user requested to read all available data (=-1), right?

You mean in the StreamReaders? If so, no, _calculate_num_samps_per_chan converts the -1 to a real value.

[zhindes]

Good fix, thanks!

[zariiii9003]

You mean in the StreamReaders? If so, no, _calculate_num_samps_per_chan converts the -1 to a real value.

_read_analog_f_64() calls DAQmxReadAnalogF64 which accepts -1 to read all values. So unless _verify_array_shape is True, that seems unnecessary.
You could save ~30% time by checking _verify_array_shape first:

        if self._verify_array_shape:
            number_of_samples_per_channel = (
                self._task._calculate_num_samps_per_chan(
                    number_of_samples_per_channel))
    
            self._verify_array(data, number_of_samples_per_channel, True, True)

        return _read_analog_f_64(
            self._handle, data, number_of_samples_per_channel,
            timeout)

Edit: Seems like DAQmxReadAnalogF64 and numSampsPerChan=-1 is blocking when the task is not continuous. I guess that's undesirable.

[zhindes]

Oh, right, great point. I was going to recommend that we move _calculate_num_samps_per_chan into _verify_array, but that wouldn't be quite right unless we made number_of_samples_per_channel and out-param. We want to use a consistent snapshot of the number_of_samples_per_channel to read for both _verify_array and _read_analog_f_64.

Our stream tests don't seem to cover anything useful for -1, so I'd want to include some new tests, some pseudocode:

for verify_array_shape in [True, False]:
   for acq_mode in [Finite, Continuous, On Demand]:
      # setup task
      # if continuous, wait for 0.1s
      if verify_array_shape:
         # allocate not enough data
         expect throw:
            read(-1)

      # allocate plenty of data
      read(-1)
      # validate we got as much data as we wanted (all samples, ~0.1s worth, or 1 sample depending on acq_mode)
      # validate good data