NTNDArray post() method does not check if what is passed in is a Value

[sfeister]

The NTScalar class method wrap() method begins with:

    def wrap(self, value, **kws):
        ....
        if isinstance(value, Value):
            pass

However, the NTNDArray method is missing this pass statement. Because of this, the example in the documentation about modifying a Value before posting it doesn't work for NTNDArrays, which is too bad as I'd like to be able to set the fields "uniqueId" and "dataTimeStamp" myself.

I've made a workaround by creating a subclass of NTNDArray and writing my own wrap method, but I thought I'd bring it to your attention!

In case anyone is looking for a workaround, here's one example:

import numpy as np
from p4p import Value
from p4p.nt import NTNDArray
from p4p.server.thread import SharedPV

class ModifiedNTNDArray(NTNDArray):
    """ Modified to allow for more customization """
    def wrap(self, value, **kws):
        if isinstance(value, Value):
            return value
        else:
            return super().wrap(value, **kws)
        
if __name__ == "__main__": 
    pv = SharedPV(nt=ModifiedNTNDArray(), initial=np.zeros(500))
    assert pv.current().raw['uniqueId'] == 0


    val = pv._wrap(np.ones(500))
    val['uniqueId'] = 23
    pv.post(val)
    assert pv.current().raw['uniqueId'] == 23

[sfeister]

It might also be nice also to put the optional keywords "uniqueId" and "dataTimeStamp" in the wrap method as optional keywords, like I did in this second example modification to the NTNDArray class:

import time
from datetime import datetime
import numpy as np

from p4p import Value
from p4p.nt import NTScalar, NTNDArray
from p4p.server.thread import SharedPV

class ModifiedNTNDArray(NTNDArray):
    """ Modified to allow for uniqueId to be set as a keyword"""
    def wrap(self, value, uniqueId=0, dataTimeStamp=None, **kws):
        if isinstance(value, Value):
            return self._annotate(value, **kws)
        
        else:
            value = super().wrap(value, **kws)
            value['uniqueId'] = uniqueId
            
            if dataTimeStamp is not None: # this code section is copied from the NTBase class
                # dataTimeStamp may be: datetime, seconds as float or int, or tuple of (sec, ns)
                if isinstance(dataTimeStamp, datetime):
                    dataTimeStamp = datetime.timestamp(dataTimeStamp)

                if isinstance(dataTimeStamp, (int, float)):
                    sec, ns = divmod(dataTimeStamp, 1.0)
                    dataTimeStamp = (int(sec), int(ns*1e9))

                # at this point timestamp must be a tuple of (sec, ns)
                value['dataTimeStamp'] = {'secondsPastEpoch':dataTimeStamp[0], 'nanoseconds':dataTimeStamp[1]}
            
            return value

if __name__ == "__main__": 
    pv = SharedPV(nt=ModifiedNTNDArray(), initial=np.zeros(500))
    assert pv.current().raw['uniqueId'] == 0

    pv.post(np.ones(500), uniqueId=23, dataTimeStamp=time.time(), timestamp=time.time())
    assert pv.current().raw['uniqueId'] == 23

[mdavidsaver]

However, the NTNDArray method is missing this pass statement. ...

This seems a reasonable change to me. Would you like to create a PR?

... nice also to put the optional keywords "uniqueId" and "dataTimeStamp" ...

While I have never been sure of the distinction between timeStamp and dataTimeStamp, I also don't have an objection.

Any opinions from someone who uses AD regularly?

[sfeister]

@mdavidsaver I will create a PR, thanks!