import collections,cStringIO
def rev_readlines(arg):
f=open(arg,'rb')
f.seek(0,2)# go to the end
line=collections.deque()
while f.tell():
f.seek(-1,1)
c=f.read(1)
f.seek(-1,1)
line.appendleft(c)
if c =='\n':
yield ''.join(line).strip()
line.clear() #clear for next line
yield ''.join(line).strip()
#bit of optimization, load groups of bytes from disk into memory
def rev_readlines2(arg,bufsize=8192):
f1=open(arg,'rb')
f1.seek(0,2)# go to the end
leftover=''
while f1.tell():
print f1.tell()
if f1.tell()<bufsize: bufsize=f1.tell()
f1.seek(-bufsize,1)
in_memory=f1.read(bufsize)+leftover
f1.seek(-bufsize,1)
buffer=cStringIO.StringIO(in_memory)
buffer.seek(0,2)# go to the end
line=collections.deque()
while buffer.tell():
buffer.seek(-1,1)
c=buffer.read(1)
buffer.seek(-1,1)
line.appendleft(c)
if c =='\n':
yield ''.join(line).strip()
line.clear()
leftover=''.join(line).strip()
yield leftover
#different approach and much faster
def rev_readlines3(arg,bufsize=8192):
f1=open(arg,'rb')
f1.seek(0,2)# go to the end
leftover=''
while f1.tell():
if f1.tell()<bufsize: bufsize=f1.tell()
f1.seek(-bufsize,1)
in_memory=f1.read(bufsize)+leftover
f1.seek(-bufsize,1)
lines=in_memory.split('\n')
for i in reversed(lines[1:]): yield i
leftover=lines[0]
yield leftover
for i in rev_readlines(filename): print i
