A "scopeguard" for Python

[Robert Kern]

On 2010-03-03 18:49 PM, Alf P. Steinbach wrote:
> * Robert Kern:
>> On 2010-03-03 15:35 PM, Alf P. Steinbach wrote:
>>> * Robert Kern:
>>>> On 2010-03-03 13:32 PM, Alf P. Steinbach wrote:
>>>>> * Robert Kern:
>>>>>> On 2010-03-03 11:18 AM, Alf P. Steinbach wrote:
>>>>>>> * Robert Kern:
>>>>>>>> On 2010-03-03 09:56 AM, Alf P. Steinbach wrote:
>>>>>>>>> * Mike Kent:
>>>>>>>>>> What's the compelling use case for this vs. a simple try/finally?
>>>>>>>>>
>>>>>>>>> if you thought about it you would mean a simple "try/else".
>>>>>>>>> "finally" is
>>>>>>>>> always executed. which is incorrect for cleanup
>>>>>>>>
>>>>>>>> Eh? Failed execution doesn't require cleanup? The example you
>>>>>>>> gave is
>>>>>>>> definitely equivalent to the try: finally: that Mike posted.
>>>>>>>
>>>>>>> Sorry, that's incorrect: it's not.
>>>>>>>
>>>>>>> With correct code (mine) cleanup for action A is only performed when
>>>>>>> action A succeeds.
>>>>>>>
>>>>>>> With incorrect code cleanup for action A is performed when A fails.
>>>>>>
>>>>>> Oh?
>>>>>>
>>>>>> $ cat cleanup.py
>>>>>>
>>>>>> class Cleanup:
>>>>>> def __init__( self ):
>>>>>> self._actions = []
>>>>>>
>>>>>> def call( self, action ):
>>>>>> assert( callable( action ) )
>>>>>> self._actions.append( action )
>>>>>>
>>>>>> def __enter__( self ):
>>>>>> return self
>>>>>>
>>>>>> def __exit__( self, x_type, x_value, x_traceback ):
>>>>>> while( len( self._actions ) != 0 ):
>>>>>> try:
>>>>>> self._actions.pop()()
>>>>>> except BaseException as x:
>>>>>> raise AssertionError( "Cleanup: exception during cleanup" )
>>>>>>
>>>>>> def print_(x):
>>>>>> print x
>>>>>>
>>>>>> with Cleanup() as at_cleanup:
>>>>>> at_cleanup.call(lambda: print_("Cleanup executed without an
>>>>>> exception."))
>>>>>>
>>>>>> with Cleanup() as at_cleanup:
>>>>>
>>>>> *Here* is where you should
>>>>>
>>>>> 1) Perform the action for which cleanup is needed.
>>>>>
>>>>> 2) Let it fail by raising an exception.
>>>>>
>>>>>
>>>>>> at_cleanup.call(lambda: print_("Cleanup execute with an exception."))
>>>>>> raise RuntimeError()
>>>>>
>>>>> With an exception raised here cleanup should of course be performed.
>>>>>
>>>>> And just in case you didn't notice: the above is not a test of the
>>>>> example I gave.
>>>>>
>>>>>
>>>>>> $ python cleanup.py
>>>>>> Cleanup executed without an exception.
>>>>>> Cleanup execute with an exception.
>>>>>> Traceback (most recent call last):
>>>>>> File "cleanup.py", line 28, in <module>
>>>>>> raise RuntimeError()
>>>>>> RuntimeError
>>>>>>
>>>>>>>> The actions are always executed in your example,
>>>>>>>
>>>>>>> Sorry, that's incorrect.
>>>>>>
>>>>>> Looks like it to me.
>>>>>
>>>>> I'm sorry, but you're
>>>>>
>>>>> 1) not testing my example which you're claiming that you're
>>>>> testing, and
>>>>
>>>> Then I would appreciate your writing a complete, runnable example that
>>>> demonstrates the feature you are claiming. Because it's apparently not
>>>> "ensur[ing] some desired cleanup at the end of a scope, even when the
>>>> scope is exited via an exception" that you talked about in your
>>>> original post.
>>>>
>>>> Your sketch of an example looks like mine:
>>>>
>>>> with Cleanup as at_cleanup:
>>>> # blah blah
>>>> chdir( somewhere )
>>>> at_cleanup.call( lambda: chdir( original_dir ) )
>>>> # blah blah
>>>>
>>>> The cleanup function gets registered immediately after the first
>>>> chdir() and before the second "blah blah". Even if an exception is
>>>> raised in the second "blah blah", then the cleanup function will still
>>>> run. This would be equivalent to a try: finally:
>>>>
>>>> # blah blah #1
>>>> chdir( somewhere )
>>>> try:
>>>> # blah blah #2
>>>> finally:
>>>> chdir( original_dir )
>>>
>>> Yes, this is equivalent code.
>>>
>>> The try-finally that you earlier claimed was equivalent, was not.
>>
>> Okay, but just because of the position of the chdir(), right?
>
> Yes, since it yields different results.
>
>>>> and not a try: else:
>>>>
>>>> # blah blah #1
>>>> chdir( somewhere )
>>>> try:
>>>> # blah blah #2
>>>> else:
>>>> chdir( original_dir )
>>>
>>> This example is however meaningless except as misdirection. There are
>>> infinitely many constructs that include try-finally and try-else, that
>>> the with-Cleanup code is not equivalent to. It's dumb to show one such.
>>>
>>> Exactly what are you trying to prove here?
>>
>> I'm just showing you what I thought you meant when you told Mike that
>> he should have used a try/else instead of try/finally.
>>
>>> Your earlier claims are still incorrect.
>>>
>>>> Now, I assumed that the behavior with respect to exceptions occurring
>>>> in the first "blah blah" weren't what you were talking about because
>>>> until the chdir(), there is nothing to clean up.
>>>>
>>>> There is no way that the example you gave translates to a try: else:
>>>> as you claimed in your response to Mike Kent.
>>>
>>> Of course there is.
>>>
>>> Note that Mike wrapped the action A within the 'try':
>>>
>>>
>>> <code author="Mike" correct="False">
>>> original_dir = os.getcwd()
>>> try:
>>> os.chdir(somewhere)
>>> # Do other stuff
>>> finally:
>>> os.chdir(original_dir)
>>> # Do other cleanup
>>> </code>
>>>
>>>
>>> The 'finally' he used, shown above, yields incorrect behavior.
>>>
>>> Namely cleanup always, while 'else', in that code, can yield correct
>>> behavior /provided/ that it's coded correctly:
>>>
>>>
>>> <code author="Alf" correct="ProbablyTrue" disclaimer="off the cuff">
>>> original_dir = os.getcwd()
>>> try:
>>> os.chdir(somewhere)
>>> except Whatever:
>>> # whatever, e.g. logging
>>> raise
>>> else:
>>> try:
>>> # Do other stuff
>>> finally:
>>> os.chdir(original_dir)
>>> # Do other cleanup
>>> </code>
>>
>> Ah, okay. Now we're getting somewhere. Now, please note that you did
>> not have any except: handling in your original example. So Mike made a
>> try: finally: example to attempt to match the semantics of your code.
>> When you tell him that he should 'mean a simple "try/else". "finally"
>> is always executed. which is incorrect for cleanup', can you
>> understand why we might think that you were saying that try: finally:
>> was wrong and that you were proposing that your code was equivalent to
>> some try: except: else: suite?
>
> No, not really. His code didn't match the semantics. Changing 'finally'
> to 'else' could make it equivalent.

Okay, please show me what you mean by "changing 'finally' to 'else'." I think 
you are being hinty again. It's not helpful. The most straightforward 
interpretation of those words means that you literally just want to remove the 
word 'finally' and replace it with 'else' in Mike's example. Obviously you don't 
mean that because it is a syntax error. try: else: is not a construct in Python. 
There is a try: except: else:, but there is no point to doing that if you don't 
have anything in the except: clause. Neither Mike's example nor your original 
one have any except: clause. Why do you think that we would interpret those 
words to mean that you wanted the example you give just above?

>>>>> 2) not even showing anything about your earlier statements, which were
>>>>> just incorrect.
>>>>>
>>>>> You're instead showing that my code works as it should for the case
>>>>> that
>>>>> you're testing, which is a bit unnecessary since I knew that, but
>>>>> thanks
>>>>> anyway.
>>>>
>>>> It's the case you seem to be talking about in your original post.
>>>
>>> What's this "seems"? Are you unable to read that very short post?
>>
>> I say "seems" because my understandings of what you meant in your
>> original post and your response to Mike disagreed with one another.
>> Now I see that your later posts were talking about minor discrepancy
>> about which errors you wanted caught by the finally: and which you
>> didn't.
>
> It's absolutely not a minor discrepancy whether some code is executed or
> not. It can have arbitrarily large effect. And from my point of view the
> discussion of that snippet has not been about what errors I "want"
> caught by the 'finally'; it's been about whether two snippets of code
> yield the same effect or not: Mike's code was incorrect not because it
> did something else, but because as code that did something else it was
> not an equivalent to the code that I posted.
>
>
>> I was confused because it seemed that you were saying that try:
>> finally: was completely wrong and that "try/else" was right. It
>> confused me and at least one other person.
>>
>>>> , but I do ask you to acknowledge that you originally were talking
>>>> about a feature that "ensure[s] some desired cleanup at the end of a
>>>> scope, even when the scope is exited via an exception."
>>>
>>> Yes, that's what it does.
>>>
>>> Which is I why I wrote that.
>>>
>>> This should not be hard to grok.
>>>
>>>
>>>> Do you acknowledge this?
>>>
>>> This seems like pure noise, to cover up that you were sputing a lot of
>>> incorrect statements earlier.
>>
>> No, I'm just trying to clarify what you are trying to say. The above
>> statement did not appear to accord with your later statement: 'if you
>> thought about it you would mean a simple "try/else". "finally" is
>> always executed. which is incorrect for cleanup.' It turns out that
>> what you really meant was that it would be incorrect for cleanup to be
>> executed when an error occurred in the chdir() itself.
>>
>> Now, I happen to disagree with that.
>
> Well, I was pretty unclear, almost hint-like, sorry about that, mea
> culpa, but you have it slightly wrong. You wrote then "The example you
> gave is definitely equivalent to the try: finally: that Mike posted."
> And it isn't.

I agree that it does behave differently with respect to when an exception is 
raised in chdir(). I was wrong on that point. I thought you were claiming that 
it behaved differently when there was an exception in the "# Do other stuff" 
block because you were being (and are still being) unclear.

>> There are a couple of ways to do this kind of cleanup depending on the
>> situation. Basically, you have several different code blocks:
>>
>> # 1. Record original state.
>> # 2. Modify state.
>> # 3. Do stuff requiring the modified state.
>> # 4. Revert to the original state.
>>
>> Depending on where errors are expected to occur, and how the state
>> needs to get modified and restored, there are different ways of
>> arranging these blocks. The one Mike showed:
>>
>> # 1. Record original state.
>> try:
>> # 2. Modify state.
>> # 3. Do stuff requiring the modified state.
>> finally:
>> # 4. Revert to the original state.
>>
>> And the one you prefer:
>>
>> # 1. Record original state.
>> # 2. Modify state.
>> try:
>> # 3. Do stuff requiring the modified state.
>> finally:
>> # 4. Revert to the original state.
>>
>> These differ in what happens when an error occurs in block #2, the
>> modification of the state. In Mike's, the cleanup code runs; in yours,
>> it doesn't. For chdir(), it really doesn't matter. Reverting to the
>> original state is harmless whether the original chdir() succeeds or
>> fails, and chdir() is essentially atomic so if it raises an exception,
>> the state did not change and nothing needs to be cleaned up.
>>
>> However, not all block #2s are atomic. Some are going to fail partway
>> through and need to be cleaned up even though they raised an
>> exception. Fortunately, cleanup can frequently be written to not care
>> whether the whole thing finished or not.
>
> Yeah, and there are some systematic ways to handle these things. You
> might look up Dave Abraham's levels of exception safety. Mostly his
> approach boils down to making operations effectively atomic so as to
> reduce the complexity: ideally, if an operation raises an exception,
> then it has undone any side effects.
>
> Of course it can't undo the launching of an ICBM, for example...
>
> But ideally, if it could, then it should.

I agree. Atomic operations like chdir() help a lot. But this is Python, and 
exceptions can happen in many different places. If you're not just calling an 
extension module function that makes a known-atomic system call, you run the 
risk of not having an atomic operation.

> If you call the possibly failing operation "A", then that systematic
> approach goes like this: if A fails, then it has cleaned up its own
> mess, but if A succeeds, then it's the responsibility of the calling
> code to clean up if the higher level (multiple statements) operation
> that A is embedded in, fails.
>
> And that's what Marginean's original C++ ScopeGuard was designed for,
> and what the corresponding Python Cleanup class is designed for.

And try: finally:, for that matter.

>> Both formulations can be correct (and both work perfectly fine with
>> the chdir() example being used). Sometimes one is better than the
>> other, and sometimes not. You can achieve both ways with either your
>> Cleanup class or with try: finally:.
>>
>> I am still of the opinion that Cleanup is not an improvement over try:
>> finally: and has the significant ugliness of forcing cleanup code into
>> callables. This significantly limits what you can do in your cleanup
>> code.
>
> Uhm, not really. :-) As I see it.

Well, not being able to affect the namespace is a significant limitation. 
Sometimes you need to delete objects from the namespace in order to ensure that 
their refcounts go to zero and their cleanup code gets executed. Tracebacks will 
keep the namespace alive and all objects in it.

-- 
Robert Kern

"I have come to believe that the whole world is an enigma, a harmless enigma
  that is made terrible by our own mad attempt to interpret it as though it had
  an underlying truth."
   -- Umberto Eco