LabVIEW
Cluster addition bug
@smercurio_fc wrote:
easily bypassed by just using wires rather than unbundling the same element more than once,
Unfortunately, I had to conclude that this really was the source of the missing data in my code before I could come up with this solution (something which took close to an hour). 
Also, I hope I haven't done this anywhere else where it went unnoticed.
- If we unbundle the value exactly once (as any reasonable programmer would do), the error does not occur (A).
- If we get rid of that unecessary local variable (as any...), the error does not occur (B).
Message Edited by altenbach on 05-30-2006 09:04 AM
@altenbach wrote:
Still one wonders why anyone would use such a weird construct. 😮(Maybe LabVIEW should not even allow unbundling the same item more than once in a single unbundle node).
This specific construct was obviously just an example. The original was a much more complex version which included typedef clusters, LV2 globals and queues and did other things entirely.
Initially I thought the source of the problem might be the typedef, but I eventually managed to narrow it down to a single VI. The local was put there specifically to cause the problem. 
The main reason for doing this is that some reasonable programmers would prefer letting LV do a +1 calculation twice in order to get a cleaner diagram. Specifically, this was an incrementing counter which was compared against another value and then wired into the F input of a select primitive. Since both the comparison and the incrementing needed +1 I performed it twice.
I see no reason why LV shouldn't allow unbundling the same element twice - when using a large cluster you might wish to get the same element in more than one place on the screen and I prefer to avoid having wires cross each other.
Agreed. Similarly, bundling the same value multiple times is just a cause for confusion. Anyone take a guess what the value of the cluster will be here (... and will it still be the same result in the newest LabVIEW version 20 years from now?). 😉
Message Edited by altenbach on 05-30-2006 09:49 AM
In any event, I agree with tst's comments and would point out that we separate the example that shows the incorrect behavior from the applicability or practicality of such an example. I think we all agree that as shown it's not practical, but that wasn't the point of the example. As I noted in my first reply, you can obviously use wires instead of the duplicate unblundle, and I know tst knows this. However, that wasn't what tst was trying to show.
Also, at the risk of belaboring the issue, I agree with tst in that on occasion I have found it more convenient to use another unbundle with the same element being unbundled purely for aesthetic wiring reasons.
Message Edited by smercurio_fc on 05-30-2006 11:54 AM
@altenbach wrote:
Similarly, bundling the same value multiple times is just a cause for confusion.
Now you're just being unfair. There is a great difference between reading multiple times and writing multiple times. I see absolutely nothing wrong in unbundling the value multiple times for readability's sake, as long as the programmers make sure they selected the correct element.
Thank you, smercurio, for protecting the innocent users walking through the jungle of the web.
