Making a series montonically decreasing in pyspark

Say we have a pyspark dataframe that looks like

I want to add another column such that the pct column is monotonically decreasing

Would we be able to do this in pyspark?

Using transform,

df = (df.withColumn('Mon_dec_pct', 
                    F.expr("""
                           transform(Pct, (x, i) -> 
                                     case
                                         when i = 0 then x
                                         when Pct[i-1] > x then x 
                                         else Pct[i-1]
                                     end)
                           """)))

========================================

UPDATE

@busfighter brought up a good point.

If an array looks like [50,10,15,12,2], the simple transform doesn't work because it can only compare the value against 1 previous value.

So, you need to carry the minimum value.

This is with using the reduce to carry both the updated array and the minimum value.

df = (df.withColumn('Mon_dec_pct', 
                    F.expr("""
                           reduce(Pct, 
                                  struct(cast(array() as array<int>) as result, Pct[0] as min_val),
                                  (acc, x) -> (array_append(acc.result, least(acc.min_val, x)), 
                                               least(acc.min_val, x)),
                                  x -> x.result
                           )
                           """)))

Input

+---+-------------------+
| ID|                Pct|
+---+-------------------+
|  1|[50, 10, 15, 12, 2]|
|  2| [42, 10, 15, 5, 3]|
+---+-------------------+

Result

+---+-------------------+-------------------+
| ID|                Pct|       Mon_dec_pct |
+---+-------------------+-------------------+
|  1|[50, 10, 15, 12, 2]|[50, 10, 10, 10, 2]|
|  2| [42, 10, 15, 5, 3]| [42, 10, 10, 5, 3]|
+---+-------------------+-------------------+