This might be done with a rolling function in pandas probably, not sure but I would like to apply the following function for a list, the current state S in position x is defined as

S[x] = if S[x-1] > 0 S[x-1] -1 S[x] else S[x] -1 for x > 1

It can be understood as the current state -1 and the current state... This is because I need to do a kind of cumulative sum of all the previous positions -1 the current positon.

An example for the list

[1,1,2,0,0,2]

returns this values

[0,0,1,0,-1,1]

because:

S[0] = 1 - 1 = 0

S[1] = S[1] - 1 S[0] = 1 - 1 0 = 0

S[2] = S[2] - 1 S[1] = 2 - 1 0 = 1

S[3] = S[3] - 1 S[2] = 0 - 1 1 = 0

S[4] = S[4] - 1 S[3] = 0 - 1 0 = -1

S[5] = S[5] - 1 (no S[4] because the else rule being smaller than 0) = 2 - 1 = 1

I am pretty sure this can probably be done in pandas but I am also open to a standard python function I send a list to (prefer pandas though).

Have been trying recursion and failed miserably.

CodePudding user response：

subtract 1 then use cumsum

(s-1).cumsum()

0    0
1    0
2    1
3    0
4   -1
5    0

here you go, revised solution to accommodate of condition in calculating cumulative sum

np.where(((s.shift(1) - 1).cumsum()) > 0, 
         (s-1).cumsum(), 
          s-1)

[ 0,  0,  1,  0, -1,  1]