First pass at doing a live "cat" in python

livecat.py

'''
Live cat of file. Uses mmap to try to be efficient with memory,
though reads a byte at a time, so it burns through CPU by hitting the
python interpreter loop N times. We could add buffering, but that
would just make the code more complex, and who wants that?

stdout gets the default buffering, so consider invoking with
-u flag/PYTHONBUFFERED environment variable if not writing to
a file.

Requires fsmonitor: https://github.com/shaurz/fsmonitor

 ...which unfortunately lacks a decent monitor for non-Windows/Linux
systems, so burns up CPU unnecessarily on those platforms. Despite its
reputation, python makes it surprisingly difficult to write clean and
correct code.

No, this is *NOT* tail. Tail looks at the end of a file, and just
happens to support live updates to the end of a file.

'''

# all the python things we need
from contextlib import closing
from itertools import chain
from mmap import mmap, ACCESS_READ, ALLOCATIONGRANULARITY
from os import fstat
from sys import argv, exit, stdout

# the magic
from fsmonitor import FSMonitor


def dump_contents(fd, offset=0, output=stdout):
    '''
    Dump contents attached to a filedescriptor.

    fd - a file descriptor, needs to support mmap
    offset - offset into the file object to start at
    '''

    filesize = fstat(fd).st_size
    gap = filesize - offset

    if gap < 0:
        exit(-1)  # file shrank, can't do much useful

    if gap > 0:
        # some logic here to paper over mmap API's inherent ugliness
        pageoffset = offset % ALLOCATIONGRANULARITY
        mmapoffset = offset - pageoffset
        with closing(mmap(fd, filesize - mmapoffset, access=ACCESS_READ, offset=mmapoffset)) as mapped:
            mapped.seek(pageoffset)
            for _ in xrange(gap):
                output.write(mapped.read_byte())
    return filesize


def cat(infile=__file__, outfile=None):
    m = FSMonitor()
    offset = 0
    m.add_file_watch(infile)

    # arbitary buffering size set to 4096 bytes
    with open(outfile, 'wb', 4096) if outfile else stdout as output:
        while True:
            with open(infile, 'rb') as f:
                file_loop = chain((None,), m.read_events())  # prepend initial iteration
                for _ in file_loop:
                    offset = dump_contents(f.fileno(), offset, output)


if __name__ == '__main__':
    cat(*argv[1:])