Simple Python implementation of 1BRC

1brc.py

from collections import defaultdict
from functools import reduce
try:
    from itertools import pairwise
except ImportError:
    # Simple alternative for pypy
    def pairwise(seq):
        seq = list(seq)
        return zip(seq, seq[1:])
import multiprocessing
import mmap
import os
import sys

class StatEntry:
    def __init__(self):
        self.min = 1000000.0
        self.max = -1000000.0
        self.sum = 0.0
        self.count = 0

    def update(self, measurement):
        if self.min > measurement:
            self.min = measurement
        if self.max < measurement:
            self.max = measurement
        self.sum += measurement
        self.count += 1

    def __add__(self, other):
        s = StatEntry()
        s.min = min(self.min, other.min)
        s.max = max(self.max, other.max)
        s.sum = self.sum + other.sum
        s.count = self.count + other.count
        return s

def merge(a, b):
    return {key: a[key] + b[key] for key in set(a.keys()).union((b.keys()))}

def collect_stats(filename, start, end):
    with open(sys.argv[1], 'r+b') as f:
        mm = mmap.mmap(f.fileno(), 0)
        mm.seek(start)
        stats = defaultdict(StatEntry)
        while mm.tell() < end:
            city, measurement = mm.readline().split(b';', 1)
            stats[city].update(float(measurement))
        return stats

def align(f, pos):
    f.seek(pos)
    f.readline()
    return f.tell()

def get_segments(filename, n):
    with open(sys.argv[1], 'r+b') as f:
        l = os.path.getsize(filename)
        return pairwise([0] + [align(f, int(i*l/n)) for i in range(1,n)] + [l])

if __name__ == '__main__':
    filename = sys.argv[1]
    num_procs = multiprocessing.cpu_count()
    pool = multiprocessing.Pool(num_procs)
    stats = reduce(merge, pool.starmap(collect_stats, [(filename, start, end)
        for start, end in get_segments(filename, num_procs)]))

    print(", ".join(f"{c.decode('utf8')}: {s.min:0.1f}/{s.sum / s.count:0.1f}/{s.max:0.1f}"
                    for c,s in sorted(stats.items())))