Hash a Range of Values

Question

I know that I can hash singular values as keys in a dict. For example, I can hash 5 as one of the keys in a dict.

I am currently facing a problem that requires me to hash a range of values.

Basically, I need a faster way to to do this:

if 0 <= x <= 0.1:
    # f(A)
elif 0.1 <= x <= 0.2:
    # f(B)
elif 0.2 <= x <= 0.3:
    # f(C)
elif 0.3 <= x <= 0.4:
    # f(D)
elif 0.4 <= x <= 0.5:
    # f(E)
elif 0.5 <= x <= 0.6:
    # f(F)

where x is some float parameter of arbitrary precision.

The fastest way I can think of is hashing, but here's the problem: I can use (0.1, 0.2) as a key, but that still is going to cost me O(n) runtime and is ultimately no better than the slew of elifs (I would have to iterate over the keys and check to see if key[0] <= x <= key[1]).

Is there a way to hash a range of values so that I can check the hash table for0.15 and still get #execute B?

If such a hashing isn't possible, how else might I be able to improve the runtime of this? I am working with large enough data sets that linear runtime is not fast enough.

EDIT: In response to cheeken's answer, I must note that the intervals cannot be assumed to be regular. As a matter of fact, I can almost guarantee that they are not

In response to requests in comments, I should mention that I am doing this in an attempt to implement fitness-based selection in a genetic algorithm. The algorithm itself is for homework, but the specific implementation is only to improve the runtime for generating experimental data.

Answer 1

As others have noted, the best algorithm you're going to get for this is something that's O(log N), not O(1), with something along the lines of a bisection search through a sorted list.

The easiest way to do this in Python is with the bisect standard module, http://docs.python.org/library/bisect.html. Note, in particular, the example in section 8.5.2 there, on doing numeric table lookups -- it's exactly what you are doing:

>>> def grade(score, breakpoints=[60, 70, 80, 90], grades='FDCBA'):
...     i = bisect(breakpoints, score)
...     return grades[i]
...
>>> [grade(score) for score in [33, 99, 77, 70, 89, 90, 100]]
['F', 'A', 'C', 'C', 'B', 'A', 'A']

Replace the grades string with a list of functions, the breakpoints list with your list of lower thresholds, and there you go.

Answer 2

If your intervals are regular, you can scale and then floor your operands to the minimum of each range, then pass that result directly into a dict mapping those lower bounds to the appropriate handler.

An example implementation, using the ranges you provided.

# Integerize our 0.1 width intervals; scale by x10
handlerDict = {}
handlerDict[0] = lambda x: ... # 0.1
handlerDict[1] = lambda x: ... # 0.2
handlerDict[2] = lambda x: ... # 0.3
...

# Get the right handler, scaling x by x10; handle
handlerDict[int(10*x)](x, ...)

Answer 3

You don't necessarily need to hash an entire range of values. For example, in the scale given above, if you were given 0.15, you can round it up to the 0.2 (the first digit after decimal point), and hash 0.2 instead.

How efficient does this have to be? An alternative you can try is a binary search. Have the interval values stores in sorted order in a list, and do binary search on it. For example:

sorted_list = [ (0.1, function1), (0.2, function2), ....(0.6, function6) ] 

Then you simply do binary search to find the smallest element that is bigger than x. This will yield O(log(n)).

Answer 4

In order to improve the runtime, you could implement a bisection search.

Otherwise you can put the interval-thresholds on a trie.

EDIT: let me propose an implementation:

class IntervalHash():
    def __init__(self,SortedList):
        #check it's sorted 
        self.MyList = []
        self.MyList.extend(SortedList) 
        self.lenlist = len(self.MyList)
    def get_interval(self,a):
        mylen = self.lenlist 
        mypos = 0
        while mylen > 1:
            mylen = (mylen/2 + mylen % 2)
            if mypos + mylen > self.lenlist - 1:
                if self.MyList[self.lenlist - 1] < a:
                    mypos = self.lenlist - 1
                break
            if self.MyList[mypos + mylen] < a:
                mypos += mylen
        if mypos == 0:
            if self.MyList[0] > a: 
                return ("-infty",self.MyList[0],0)
        if mypos == self.lenlist - 1:
            return (self.MyList[mypos],"infty",0)
        return (self.MyList[mypos],self.MyList[mypos+1],0)

A = [0.32,0.70,1.13]
MyHasher = IntervalHash(A)
print "Intervals are:",A
print 0.9 ," is in ",MyHasher.get_interval(0.9)
print 0.1 ," is in ",MyHasher.get_interval(0.1)
print 1.8 ," is in ",MyHasher.get_interval(1.8)

further edits and improvements are welcome! The trie approach is much more involved, in my opinion it would be more appropriate for low level languages.