how to count lines in a file ?

[Delaney, Timothy]

> From: Richard Jones [mailto:rjones at ekit-inc.com]
> 
> On Thu, 25 Jul 2002 8:31 am, Delaney, Timothy wrote:
> > > From: Bo M. Maryniuck [mailto:b.maryniuk at forbis.lt]
> > >
> > > print len(open('/etc/passwd').readlines())
> 
> Sorry, I'm coming in to this thread halfway through, but I 
> need to know why 
> that code is broken? Is it the 
> "open('/etc/passwd').readlines()"? That's a 
> very common pattern, as you note, because people think it is 
> perfectly 
> legitemate. What about it is broken? Why isn't it pointed out 
> more often that 
> it _is_ broken?

We had this discussion just recently, but yes, that is precisely what is
broken.

What is broken? The file is never explicitly closed.

Why is it broken? There are no guarantees that the file will be closed
either:

1. Immediately (the assumption in code that uses the above construct);

2. In a timely manner;

3. Before the interpreter exits;

4. Before the computer is turned off.

You can usually gather a guarantee of 4 from the OS (all files opened by a
process are closed when that process exits) but you then have another
problem:

	open(filename).write(string)

Now you not only have no guarantees that the file is closed, but also no
guarantees that is the file is closed, it is flushed before closing. The
most likely case where this occurs is at program exit - all files are
closed, but not flushed.

It is explicitly documented that yo may not rely on reference-counting
semantics for such things (Paul Rubin - take note please).

Tim Delaney

To date, many people have relied in (non-) guarantee #1 - that the file will
be closed (and flushed) immediately.

This in currently not true in Jython, and would not have been true with the
proposed patch. Luckily for such people, there is a strong desire not to
break "working" code, even if such code really is broken anyway. What
happens if I try to use a module containing such code under Jython?

Tim Delaney