Exceptions are not just for errors

[Gregory Ewing]

Ben Finney wrote:

> Another way of thinking about it is that there's no sensible sequence of
> bytes to return at EOF, so the Pythonic thing to do is to raise an
> exception for this exceptional circumstance.

But this is *not* what Python does, so it's obviously
not Pythonic. :-)

If f.read(n) is to mean "read n bytes, or however many are
left", and there are no bytes left, then the consistent
thing to do is to return a zero-length sequence of bytes.

I came across a situation recently where Microsoft got this
badly wrong. I was using a language that didn't have very
good file-access capabilities, and I wanted to compare the
contents of two files. So I used Scripting.FileSystemObject
via COM, and wrote something like the equivalent of

    f = open(...)
    g = open(...)
    if f.read() == g.read():
       ...

This worked fine... except when one of the files was empty,
in which case the read() call *raised an exception*!

I am very glad that Python didn't go down that route.

-- 
Greg