Proper use of case structures for array insertion

[StudentSmith]

Salutations,

I have an array that looks of the form:

0 0 0 0
1 2 3 0
0 1 0 3
0 0 0 0

I want to be able to remove the lines that are full of zeros.  My idea was to use a summation of the row and check to see if the value was greater than zero.  This True or False statement would be sent to a case structure that would then insert the rows that are non-zero'd (true) when added together and just ignore when the summation was zero (false).  This idea has failed me, because the case structure requires information to be sent for each case.  I was hoping someone would have a clever trick for removing the 0'd rows.

Thank you,
E. Smith

[Darren]

Hello,

I have attached a VI that does what you're looking for.  Notice in the VI I use the Array Max and Min function to determine if the row is full of zeroes.  Adding the row elements won't work...what if you have a row that has 1, -1, 1, -1 in it?  Notice I also use the Delete From Array function and a shift register, and I keep track of the current row index based on whether or not I have deleted a row.

Hope this helps...and don't worry, the more you program in LabVIEW, the easier it will be for you to come up with stuff like this... 🙂

-D

[WNM]

In the attached VI are a couple more ways of doing it...

[altenbach]

To test a row for all zeroes, simply feed it through a "equal zero" from the comparison palette, then through a "AND array elements" from the boolean palette. Summing array elements will potentially fail once you allow negative numbers.

Attached is one other possibility.

[Darren]

So who wants to benchmark and see which of the myriad options presented here is the fastest one?  🙂

-D

[altenbach]

😄

Here's my wild guess, assuming very large arrays:

Darren: "delete from array" is potentially problematic, because for each row you delete, ALL higher rows will need to be moved down a row. This is a lot of data shuffling. if you do this thousands of times.

Warren #1: Building an array in a loop is potentially expensive, because of memory reallocations.

Warren #2: Same problem as Darren.

Altenbach: All array operations are "in place", followed by a final trimming ar the end. One might try to use index array inside the loop instead of the autoindexing, but causal tests in the past did not show a difference.

I could be way off, of course. Anyone want to benchmark them and post the results? 🙂

[WNM]

---

@altenbach wrote:

To test a row for all zeroes, simply feed it through a "equal zero" from the comparison palette, then through a "AND array elements" from the boolean palette. Summing array elements will potentially fail once you allow negative numbers.

Attached is one other possibility.

---

You are of course right about the flaw with the summing. I just wasn't thinking past the problem as it was posed. I'm also impressed by your method of sifting thru the array in place. Later in the day yesterday I happened to notice that your method would benefit one of the OpenG library routines and suggested as much. It looks like it will be incorporated there.

Thanks!