What would be an implementation of a q kdb function that does the same thing as the rank function but returns duplicated rank numbers for equal values?
In the standard implementation
rank 2 7 4 3 19 16 15 20 25 16 19
gives:
0 3 2 1 7 5 4 9 10 6 8
But I am looking for:
0 3 2 1 6 5 4 7 8 5 6
Can do that by using distinct to remove duplicates, applying rank to that list and then creating a dictionary to map those rankings back to the original list.
q)list:2 7 4 3 19 16 15 20 25 16 19
/ creating the dictionary
q)d!rank d:distinct list
2 | 0
7 | 3
4 | 2
3 | 1
19| 6
16| 5
15| 4
20| 7
25| 8
/ wrapping this up in a function
q)rankDistinct:{(d!rank d:distinct x)x}
/ applying to the list
q)rankDistinct list
0 3 2 1 6 5 4 7 8 5 6
Or alternatively you could sort the distinct elements and then use find ? to get the ranking by finding the index of where list elements appear in the distinct sorted elements.
q)rankDistinct2:{(asc distinct x)?x}
q)rankDistinct2 list
0 3 2 1 6 5 4 7 8 5 6