Why is break allowed in finally, but continue is not?

[Ben Finney]

Ned Batchelder <ned at nedbatchelder.com> writes:

> For testing coverage.py, I wrote a program to generate
> randomly-structured Python functions.  When compiling
> the results, I got a message I'd never seen before:
>
> SyntaxError: 'continue' not supported inside 'finally' clause
>
> I guess this makes sense, when cleaning up from an
> exception, continuing the loop seems an odd thing to do.
> But 'break' is allowed in 'finally' clauses!  I tried this:
>
>     # Huh? This prints "here", and no exception is raised.
>
>     for i in range(1):
>         try:
>             1/0
>         finally:
>             # If you change this to "continue", you get:
>             # 'continue' not supported inside 'finally' clause
>             break
>     print "here"
>
> The finally is perfectly willing to have a 'break', but it's
> a syntax error to have 'continue'?  Why? I don't see a
> difference between the two when it comes to cleaning up
> exceptions.

Raymond Hettinger's answer is:

    The use of continue in a finally-clause is forbidden because its
    interpretation would have been problematic. […]

    <URL:https://stackoverflow.com/questions/8302293/why-is-continue-not-allowed-in-a-finally-clause-in-python#answer-8302601>

The example he uses::

    for i in range(10):
        print i
        try:
           raise RuntimeError
        finally:
           continue        # if the loop continues, what would happen to the exception?
        print i

What, in your opinion, should the above code do if instead of ‘continue’
some other flow-control statement is used?

> There are other things you can do in a finally clause that
> will prevent the exception from being raised, like 'return',
> and 'break' works just fine.

Hettinger doesn't defend those, and is dubious about ‘return’'s
candidacy:

    Interestingly, you can put a return inside a finally-clause and it
    will swallow all exceptions including KeyboardInterrupt, SystemExit,
    and MemoryError. That probably isn't a good idea either ;-)

> So why treat 'continue' specially?

I am inclined to agree, but in the opposite direction: a case should be
made for allowing *any* flow-control statement in an exception-handler's
‘finally’ clause.

-- 
 \             “We can't depend for the long run on distinguishing one |
  `\         bitstream from another in order to figure out which rules |
_o__)               apply.” —Eben Moglen, _Anarchism Triumphant_, 1999 |
Ben Finney