2138 Rate this article:
No rating

The minimum and maximum operators

Mark Alonzo
Have you used the minimum (<) and maximum (>) operators in IDL? They're among my favorites; they're binary operators that return the smaller (minimum) or larger (maximum) of their operands. For example, 5 is larger than 2, so the minimum operator returns 2:
IDL> print, 5 < 2
       2
whereas the maximum operator returns 5:
IDL> print, 5 > 2
       5
Note that these differ from the IDL relational operators LT and GT, which are also binary operators, but instead return a true (1B) or false (0B) value:
IDL> print, 5 lt 2
   0
IDL> print, 5 gt 2
   1
It gets better: like most operators in IDL, the minimum and maximum operators also work on arrays. For example, a sill operation raises the minimum value of an array to the value of the sill. Here, the array new is created by applying the maximum operator to the array test with the value sill:
IDL> test = indgen(5)
IDL> sill = 2
IDL> print, test
       0       1       2       3       4
IDL> new = test > sill
IDL> print, new
       2       2       2       3       4
Note that no value of new is less than sill. The minimum and maximum operators can also be used in masking. Imagine data, stored in an array, that has bad values. I want to mask out and replace the bad values. For this example, I'll use RANDOMN to generate sample data:
IDL> sample = randomn(1, 5, 5)
IDL> print, sample
    -0.836854    -0.172280     0.187117      1.61544    -0.176774
     0.653145    -0.546364     0.194146     0.925709      1.20432
      1.53055     -1.35556    0.0514889      1.02018     -1.22616
     0.708497     0.871673    -0.789721     0.332079     0.205603
    -0.169367    -0.318417    -0.295643     0.522291     -2.23105
Say the bad values are less than zero and I want to replace them with the value zero. You could use the WHERE function to identify the locations of these values in the array:
IDL> i_bad = where(sample lt 0.0, /null)
IDL> help, i_bad
I_BAD           LONG      = Array[11]
and then, by subscripting, replace them with zeros:
IDL> sample[i_bad] = 0.0
IDL> print, sample
     0.000000     0.000000     0.187117      1.61544     0.000000
     0.653145     0.000000     0.194146     0.925709      1.20432
      1.53055     0.000000    0.0514889      1.02018     0.000000
     0.708497     0.871673     0.000000     0.332079     0.205603
     0.000000     0.000000     0.000000     0.522291     0.000000
But the maximum operator, used this time in a compound operator, provides a one-line solution:
IDL> sample = randomn(1, 5, 5) ; reconstitute the sample
IDL> sample >= 0.0
IDL> print, sample
     0.000000     0.000000     0.187117      1.61544     0.000000
     0.653145     0.000000     0.194146     0.925709      1.20432
      1.53055     0.000000    0.0514889      1.02018     0.000000
     0.708497     0.871673     0.000000     0.332079     0.205603
     0.000000     0.000000     0.000000     0.522291     0.000000
Cool! Do you have an example in your work where you use the minimum or maximum operator?

Please login or register to post comments.