Quantize inputs like dtrace does

quantize.py

"""
Re-bucket input histogram in linear or exponential buckets.

Input: 
value\tcount

Output:
quantized_value\tcount
"""

import itertools
import sys
from collections import OrderedDict

parser = argparse.ArgumentParser(description='Quantize input data')
parser.add_argument('-t', '--type', choices=('q', 'l', 'll'), default='q',
                    help='type of quantization')
parser.add_argument('params', type=int, nargs='*',
                    help='type of quantization')

args, more_args = parser.parse_known_args()
if args.type == 'q':
    def gen():
        for i in itertools.count():
            yield 2 ** i
elif args.type == 'l':
    lower, upper, step = args.params
    def gen():
        for i in xrange(lower, upper, step):
            yield i

elif args.type == 'll':
    factor, lower, upper, step = args.params
    def gen():
        for i in xrange(lower, upper):
            bottom = factor ** i
            top = bottom * factor
            lstep = max(top//step, 1)
            for j in xrange(bottom, top, lstep):
                yield j

buckets = OrderedDict()
for l in sys.stdin:
    val, count = map(int, l.split('\t', 1))

    first = True
    last = None
    for bucket in gen():
        if last is None:
            last = '<' + str(bucket)

        buckets.setdefault(last, 0)
        if val < bucket:
            buckets[last] += count
            break
        last = bucket

    else:
        buckets.setdefault(last, 0)
        buckets[last] += count

first = True
for k, v in buckets.iteritems():
    if first and not v:
        continue
    first = False
    print '{}\t{}'.format(k, v)

see quantize and lquantize in dtrace literature, and llquantize here