I have a bunch of 3D vectors $\vec F_i = (F_{xi}, F_{yi}, F_{zi})$ which I would like to order. The ordering rule is the following:

$\vec F_i$ ">" $\vec F_j$ when:

  • $F_{xi} < F_{xj}$
  • $F_{yi} < F_{yj}$
  • $|F_{zi}| > |F_{xj}|$

My question is whether I can find a scalar function $g(\vec F)$ so that:

$g(\vec F_i) > g(\vec F_j)$ when $\vec F_i$ ">" $\vec F_j$


What makes this custom-ordering problem maybe a bit easier is that all vector components $F_{xi}, F_{yi}, F_{zi}$ assume their values from (0, -30, -60, ... -300).

So the first point would be (0, 0, 0) and the last (-300, -300, -300).

My tries have lead me nowhere. I basically tried multiplying the vector with an array whose coefficients I tried to "tune" but to no avail.

Note that there can be unorderable pairs as well. For example:

$F_1 = (-120, -90, -270)$


$F_2 = (-90, -120, -150)$

$F_{x1} < F_{x2}$ but $F_{y1} > F_{y2}$ and therefore neither can "win"

  • $\begingroup$ I think we should first try the 2-D case. $\endgroup$ – user456218 Jun 30 '17 at 8:23

Such a function doesn't exist.

The reason is that when you are comparing two vectors, you require all the 3 parameters of the one that will be larger, however you will need only two from the one that is smaller. So, you need such an $F$ such takes all three components in to evaluate to a scalar number but somehow magically ignores the third component when it is compared to a smaller number, no matter how large the $z$ component of the smaller one is.

To understand this let go of the function for a moment. Your order doesn't make sense; i.e it is possible to find elements $X,Y,Z$ such that $X>Y$ and $Y>Z$ but $X$ and $Z$ are not orderable . Try to find an example. If you can't then comment below.I will give you one. I am asking you to find to understand why there will be a problem. Once you find the example you can see that you can't find such a function. Because if $X>Y$ and $Y>Z$ then if such a function existed then $g(X)>g(Y)>g(Z)$.

(Hint: Try 2D first to find example)

  • $\begingroup$ convinced! Thanks! $\endgroup$ – Ev. Kounis Jun 30 '17 at 8:50

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.