python - Split array into evenly-distributed chunks -

i'm looking efficient way split array chunks have similar histogram.

for example, given array:

l = np.array([1, 2, 3, 4, 1, 1, 1, 2, 3, 4, 55, 5, 5, 5, 5, 5, 3, 2, 2, 21, 2]) 

i obtain:

c1 = np.array([1, 4, 1, 5, 5, 3, 2]) c2 = np.array([2, 1, 3, 4, 5, 5, 2]) c3 = np.array([3, 1, 2, 55, 5, 2, 21]) 

not each chunk should have similar size , similar sum on given function f:

1. |sum(ci, f) - sum(cj, f)| < e, != j 2. |len(ci) - len(cj)| < e, != j 

where

sum(c, f) = f(c[0]) + ... + f(c[len(c)]) 

edit: clarify intention of this. want parallelize process on list n sub-processes cost has distributed among each sub-process evenly. cost of processing element in list function f of integer @ same position in array l, f computational complexity of process. example, f(i)=i^2. so, want processes have same computational cost , not have processes finish while others last forever.

let's start weak assumption similar histograms defined in following rudimentary way: set of integers s1, histogram h(s1) similar histogram h(s2) of set s2 if sum(s1) = sum(s2).

generally after finding subsets s1, s2, ..., sn of array a such f(s1) = f(s2) = ... = f(sn), , under our assumptions f=sum. unfortunately you've got k-partition problem, np-hard, , if you'll find efficient way (ie poly-time) that, requested, consequence p=np proven true first on stackoverflow!