Just one question about ( min function) why it works only with scalars ... whenever I want to use it with vectors or matrics it doesn't work ...
How to overcome this ?
Should I always use programming ?
For example the following program : in each (qc;gem;cor ) I have 3 matrics each one has ten values.... what I need is to write a program that will compare the both variables and choose the minimum value of each one ( compare the first raw with the first raw and chooses the min ...and so on ) ...and the result should be also one matric with ten values. However, it chooses only the minimum value of each matrics
ACCEPTED SOLUTION
Accepted Solutions
It seems to me that @YA_10963798 would need you function but for only two inputs.
Applied to the simpler example posted in the other thread this would mean
BTW, I seem not to grasp the necessity of your local function minv?
Prime 9 sheet attached.
Try something like the function Min shown below. It recursively calculates the minimum of each nested array ... well, it's supposed to!
The last one is messy but interesting.
Stuart
Hi Stuart,
I tried it
Actually what I need is something different I need the program to compare qc;gem;1 and qc;gem;2 and choose the minimum between both of them.
So , it goes to the first raw in both variables and choose the min and then goes to the second one and does the same.
but what we did is just choose the min value in each matric
Ah, my apologies. I was distracted by another problem and didn't read your message correctly. I just caught sight of the word "minimum", saw a nested vector, and ran off like a dumb dog chasing a rabbit instead of a hare (not that a modern, socially aware. dog ought to be chasing any fellow creatures, of course).
Is this any better?
Stuart
It seems to me that @YA_10963798 would need you function but for only two inputs.
Applied to the simpler example posted in the other thread this would mean
BTW, I seem not to grasp the necessity of your local function minv?
Prime 9 sheet attached.
@Werner_E wrote:
It seems to me that @YA_10963798 would need you function but for only two inputs.
Applied to the simpler example posted in the other thread this would mean
BTW, I seem not to grasp the necessity of your local function minv?
Probably due to sheer laziness on my part. You've externally defined and vectorized minv (as minvecs2) whereas I put it inside minvecs; AFAICT, it's needed either way. The lazy way means minvecs is self-contained and I don't need to remember to vectorize minvec2 or, worse, potentially vectorize it more than once!! 😱 I have a self-imposed per diem keystroke limit - my personal contribution to saving energy and reducing anthropogenic global warming. 😇
Then again, I might very well have missed something that you've spotted and am still failing to see it. ☹️
Stuart
I've miscopied vectorized expressions more than once because I didn't fully enclose the expression, so I tend to avoid explicit vectorization, preferring to hide it away in a function where possible.
@StuartBruff wrote:
@Werner_E wrote:
Ahh, I missed the fact that in your version its not necessary to vectorize the function call when used with vector arguments.
Usually I prefer to explicitly vectorize when I call a function with vector elements and want the operation to be done element-wise - not least to remind myself of what is happening at that point.
Horses for courses. Mathcad, like many other such applications, vectorizes many of its built-in functions, and so it makes sense to me to do the same. In addition, the fact that function arguments are effectively global variables allows the programmer to remove some of the burden of trying to vectorize operations over (knowingly) disparate array dimensions. I frequently use this ability to handle pseudo-vectorization of functions of the form f(n, v) where not only does f vectorize scalar n over vector v, but it can also vectorize over vector n as well. Most users do not want to know about such degrees of complexity and simply want "it" to happen.
But, this approach can catch people out if they're not expecting it, and it makes the functions a lot more complex in themselves and, therefore, harder to verify and understand. Not to mention usually taking up more valuable whitespace than a simpler function would! 😱
Here's an example of making things more complex but with added generality of function. Which output would people prefer, if either? And which would they like to see as a solution to their specific problem? Note: structure-preserving means that the overall shape of the nested array is preserved, ie the fact that an element was an array or not an array. It does not preserve the size of the nested elements, nor should it, as min is a reducing function.
The first variant of Min could be readily made into a structure-preserving function by copying over the last line in the function body of the second variant.
Stuart
I like it, but I am not sure if many people would expect a 1x1 matrix as the result of a vector or matrix argument.
But of course that is what structure preserving means.
Nonetheless I'm sure that some people would prefer
But then, its all a matter of what you need the function/result for in the further calculation.
Its similar to match or lookup, where you often just need the index number or the matching element and not a 1x1^matrix, which is returned by the built-in functions. They do this for good reason, as there could be more than one match and it would complicate a lot of applications if the functions would return a scalar instead if a 1x1 matrix.
@Werner_E wrote:
I like it, but I am not sure if many people would expect a 1x1 matrix as the result of a vector or matrix argument.
But of course that is what structure preserving means.
Nonetheless I'm sure that some people would prefer
But then, its all a matter of what you need the function/result for in the further calculation.
Its similar to match or lookup, where you often just need the index number or the matching element and not a 1x1^matrix, which is returned by the built-in functions. They do this for good reason, as there could be more than one match and it would complicate a lot of applications if the functions would return a scalar instead if a 1x1 matrix.
Indeed. When I could reasonably expect to see a vector result, I normally ensure that valid results return a vector, even if is a 1x1. I frequently use 0 (zero) to indicate that there was no reasonable vector result, ie 0 acts a stand in for the empty array (*). An example is when I want to filter a list using, say, match but there are no elements in the list that meet the match criterion. In that case, 0 usually makes far more sense than raising an error (as built-in match does) - it tells me exactly what I need to know: there were zero matches. As even a single match returns a vector, it's easy to check whether there are any matches. And iI can do so without having to build in error traps and consume unnecessary vertical whitespace using the try-else operator.
Stuart
(*) It doesn't cover all cases, which is why we still need an empty array. Ideally accompanied by a true multi-dimensional array (MDA) data type! 😁
