Optimization of checking on conditions in a 3d array

General Usage
1

I have an example of a python code snippet. I need to check all the numbers from the map_spectrum array for 5 conditions, then put the new value in another array according to them. How can this be done in a more optimized version than in python code ?

x = y = np.arange(0.0, 1.0, 0.1)
map_spectrum = np.ones( (len(x), len(y), 6) )
map_color = np.ones( ( len(x), len(y) ) )

for ix, xv in enumerate(x):
    for iy, yv in enumerate(y):
            if map_spectrum[ix, iy, 0] > 0 and map_spectrum[ix, iy, 1] > 0 and map_spectrum[ix, iy, 2] == 0 and map_spectrum[ix, iy, 3:] < 0 :
                        map_color[ix, iy] = 1

In the real case, this array has a size of 101x101x6, that is, conditionally there are 10201 cells containing 6 values.
All the cases of conditions in a short form:

First case +, +, 0, -, -, -
Second case +, 0, -, - ,- ,-
Third case 0, 0, -, -, -, -
Four case 0, -, -, -
Another

In addition, how to take into account machine zero when checking conditions.
Example real values from array
image

After creating a new array, it will be used as colors in heatmap that is

heatmap(x, y, map_color)

I tried to cycle through all the values and check them for 5 conditions, but it takes quite a long time

2

this loop will run pretty fast in Julia, have you tried?

3

With a large array, this may take some time, also I would like to avoid using a loop or at least 5 conditions in a row in a loop. but even if we leave everything in this form, there remains the problem of how to check whether an infinitesimal value is really zero.

4

how else do you want to do this?

use a cutoff

julia> isapprox(1e-15, 0.0; atol=1e-16)
false

julia> isapprox(1e-17, 0.0; atol=1e-16)
true
1 Like
5

I’m thinking of using the equivalent switch statement from c++, but Julia doesn’t have it therefore, apparently there remains only the option of placing 5 conditions in the function.

6

The value can have, for example, 1 e4 or 1e-6 or 1.0762649529273773e-6 and this should be considered a machine zero
I can only specify which precision is considered zero?

7

atol is the absolute tolerance

8

you have five conditions ANDed together, how are you gonna use switch?

9

try using ternary operator in this way

First case +, +, 0, -, -, -
Second case +, 0, -, - ,- ,-
Third case 0, 0, -, -, -, -
Four case 0, -, -, -
Another

ms0>0 && ms1>0 && ms2==0 && ms3<0   ? mp = "first" :
ms0>0 && ms1==0                     ? mp = "second" :
ms0==0 && ms1==0                    ? mp = "third" :
ms0==0                              ? mp = "four" :
                                      mp = "another"
1 Like
10

Thank you, i read it and in general I understood how to use it except for one moment. I have not found an example in which the result would be written to a new variable and not the current one

11

I’m not sure what your doubt is. If you are referring to the variable mp = 1, mp = 2 …, in its place you can put any expression with any variable

12

I understanded, sorry. I checked this with my code and it’s work.
It remains to check isapprox on my data

13

isaprox don’t work with number 1.9184694551117396e-5

isapprox(1.9184694551117396e-5, 0.0; atol = 1e-8)

how to fix it or I don’t fully understand something

14

your tolerance is smaller than the difference, make the tolerance larger