Dan Colesworthy wrote:
>
> Hello,
>
>
>
> I started with the example in the scipy cookbook which acquires data
> from a NI DAQ via the NIDAQmx DLL. Attempting to perform the inverse
> function i.e.:
>
>
>
> # initialize variables
>
> taskHandle = TaskHandle(0)
>
> max_num_samples = 1000
>
> dat = numpy.zeros(max_num_samples,dtype=numpy.float64)
>
> for i in numpy.arange(1000):
>
> dat[i] = 9.95*math.sin(i*2.0*PI*1000.0/16000.0)
>
>
>
> After initializing a task etc, I tried:
>
>
>
> CHK(nidaq.DAQmxWriteAnalogF64(taskHandle,
>
> 1000,
>
> 0,
>
> float64(10.0),
>
> DAQmx_Val_GroupByChannel,
>
> ctypes.byref(dat),
>
> None,
>
> None ) )
>
>
>
> Which generates the error:
>
>
>
> Traceback (most recent call last):
>
> File "./ni_2.py", line 76, in ?
>
> ctypes.byref(dat),
>
> TypeError: byref() argument must be a ctypes instance, not 'numpy.ndarray'
>
>
>
> The call requires a float64 write_array[] argument.
>
>
>
Generally ctypes will not convert a python type into a C type. Only a
few Python types like str and int are assigned default C types by
ctypes. numpy arrays and float64s are not. But the address of a numpy
array's data can be retrieved and used as a pointer argument to a
foreign function.

DOUBLEPTR = ctypes.POINTER(ctypes.c_double)

CHK(nidaq.DAQmxWriteAnalogF64(taskHandle,

1000,

0,

ctypes.c_double(10.0),

DAQmx_Val_GroupByChannel,

ctypes.cast(dat.ctypes.data, DOUBLEPTR),

None,

None ) )


-- 
Lenard Lindstrom
<len-l@xxxxxxxxx>