Image array size doesn't match image1:ArraySize0 * image1:ArraySize1

[ronpandolfi]

I'm trying to use ophyd's shaped_image DerivedSignal, but I'm getting this error:

Subscription value callback exception (EpicsSignal(read_pv='ALS:701:image1:ArrayData', name='701_image1_array_data', parent='701_image1', value=array([882, 882, 882, ...,   0,   0,   0], dtype=int16), timestamp=1558385498.360814, auto_monitor=False, string=False, write_pv='ALS:701:image1:ArrayData', limits=False, put_complete=False))
Traceback (most recent call last):
  File "/home/rp/PycharmProjects/ophyd/ophyd/ophydobj.py", line 315, in inner
    cb(*args, **kwargs)
  File "/home/rp/PycharmProjects/ophyd/ophyd/signal.py", line 486, in _derived_value_callback
    value = self.inverse(value)
  File "/home/rp/PycharmProjects/ophyd/ophyd/areadetector/base.py", line 107, in inverse
    return np.array(value).reshape(array_shape)
ValueError: cannot reshape array of size 2361600 into shape (2050,960)

I can see that ...image1.shaped_image._shape[0] maps to ...image1:ArraySize1_RBV. These PVs don't match the size of the image (confirmed with caget).

Rather, ...cam1:ArraySizeX_RBV and ...cam1:ArraySizeX_RBV seem to match the shape of the image (2050,1152).

Is this an issue with the configuration of the detector, or the IOC?

[MarkRivers]

The size of the array in the NDStdArrays plugin (image1:ArraySize0_RBV x image1:ArraySize1_RBV) is not necessarily the same as the camera itself (cam1:ArraySizeX_RBV, x cam1:ArraySizeY_RBV) because there could be an ROI plugin, etc. in-between. It should be using the dimensions from the stdArrays
plugin because that is what is sent to the waveform record. Recall that the waveform record is a fixed length. What is that set to in your startup script?

I have not use ophyd so I don't know the details here. ophyd should only be requesting or using the number of elements image1:ArraySize0_RBV x image1:ArraySize1_RBV.

[ronpandolfi]

Hi Mark, my understanding of the internals here is limited; I'm still learning EPICS. I think my issue here is that the array I'm getting back from image1 does not match the shape associated with image1. I think this is not correct behavior.

Only as a curiosity, cam1's shape matches image1's array size; this was also unexpected.

[MarkRivers]

Hi Ron, I discussed this briefly with Dan and Tom yesterday. I think you should probably contact them directly.

[stuwilkins]

@ronpandolfi @mrakitin and I discussed this yesterday. I believe that Ron is seeing the expected behavior. He gets a numpy array which is of the size NELEM. This is of course bigger than the image size. Therefore the array should be truncated by image1:ArraySize0 * image1:ArraySize1 and reshaped by (image1:ArraySize0,image1:ArraySize1).