Add array trimming before reshaping #731
Conversation
(cherry picked from commit 4673d19)
|
@ronpandolfi I wonder if this will blow up if the array is < the product of the array dimensions? I just have a question if we should trap that? This is very possible if the IOC is not configured correctly. |
ophyd/areadetector/base.py
Outdated
| array_shape = self.derived_shape[:self.derived_ndims] | ||
| if not any(array_shape): | ||
| raise RuntimeError(f"Invalid array size {self.derived_shape}") | ||
|
|
||
| array = self._derived_from.get() | ||
| return np.array(array).reshape(array_shape) | ||
| return np.array(value[:np.prod(array_shape)]).reshape(array_shape) |
There was a problem hiding this comment.
Shouldn't it be np.array(array[... using the array from L108? Not sure there the input value was used before.
There was a problem hiding this comment.
What type is value here? If it already an array (at least sometimes, in some circumstances) using np.asarray in place of np.array will save us from making a copy.
There was a problem hiding this comment.
Shouldn't it be
np.array(array[...using thearrayfrom L108? Not sure there the inputvaluewas used before.
The array value is actually passed into this method as the value arg; there's no need to get it again, so I've removed that line.
|
Qingteng Zhang has also encountered this issue with a Pilatus detector: |
|
@ronpandolfi I'm up for tagging a patch release to get this out. Can you look at @mrakitin's and my comments above when you get a chance? |
|
@danielballan Above comments have been addressed. |
|
Restarted the failed test. @mrakitin We don't have a good enough test rig for AD stuff to be worth reproducing this issue. |
Following discussion of areaDetector/ADFastCCD#13 with @mrakitin and @stuwilkins, it was decided that it is possible for EPICS to transmit more elements than the product of its shape, as with the FastCCD and "overscan rows".
To accommodate this, the image array must be trimmed before reshaping.