Approximate search for a prefixed line in a sorted file (ex. a huge log file)

bsearch-huge-file.py

import os.path
import sys


def bsearch(prefix, filename, steps, lines):
    lo = 0
    hi = os.path.getsize(filename)

    with open(filename, 'r') as fp:
        while steps > 0:
            mid = (lo + hi) / 2

            fp.seek(mid)
            fp.readline()
            line = fp.readline()
            if line < prefix:
                lo = mid
            elif line > prefix:
                hi = mid

            steps -= 1

        sys.stdout.write(line)
        for lineno, line in enumerate(fp):
            sys.stdout.write(line)
            if lineno + 1 >= lines:
                break


if __name__ == '__main__':
    prefix = sys.argv[1]
    filename = sys.argv[2]
    lines = int(sys.argv[3])
    bsearch(prefix, filename, 20, lines)