dolfin team mailing list archive

["Garth N. Wells" <gnw20@xxxxxxxxx>]

On 06/03/11 23:01, Anders Logg wrote:
> On Sun, Mar 06, 2011 at 09:45:20PM +0000, Garth N. Wells wrote:
>>
>>
>> On 06/03/11 21:39, Anders Logg wrote:
>>> On Sun, Mar 06, 2011 at 10:29:10PM +0100, Marie E. Rognes wrote:
>>>> On 03/06/2011 10:22 PM, Anders Logg wrote:
>>>>> On Sun, Mar 06, 2011 at 09:08:10PM +0000, Garth N. Wells wrote:
>>>>>>
>>>>>>
>>>>>> On 06/03/11 21:02, Marie E. Rognes wrote:
>>>>>>> On 03/06/2011 09:30 PM, noreply@xxxxxxxxxxxxx wrote:
>>>>>>>>     if (parameters["max_dimension"].change_count()>   0
>>>>>>>>         &&   V.dim()>   max_dimension)
>>>>>>>> +  {
>>>>>>>>       return true;
>>>>>>>> -
>>>>>>>> -  // Otherwise, not done.
>>>>>>>> -  return false;
>>>>>>>> +  }
>>>>>>>> +  else
>>>>>>>> +    return false;
>>>>>>>>   }
>>>>>>>
>>>>>>>
>>>>>>> I notice that my early returns keep getting moved into else clauses... I
>>>>>>> find this approach less readable, especially when there are nested ifs.
>>>>>>> Why is it the preferred way?
>>>>>>>
>>>>>>
>>>>>> Because your comment basically says else, so I'd say it's better to have
>>>>>> the code say it consistently.
>>>>>>
>>>>>> I find it easier to follow, because it's clear that the function exits
>>>>> >from the conditional block. The return value is either true or false
>>>>>> depending on the one true/false evaluation.
>>>>
>>>>
>>>> The code is an if -- else if -- else. I don't see how moving that
>>>> into an if, if -- else increases consistency.
>>>>
>>>>
>>>>>
>>>>> I'm not sure we're talking about the same thing here, but I think the
>>>>> following examples are *good* practice:
>>>>>
>>>>> 1. Early return to get rid of special cases
>>>>>
>>>>> void foo()
>>>>> {
>>>>>   // Get rid of some case
>>>>>   if (...)
>>>>>     return;
>>>>>
>>>>>   // Get rid of another case
>>>>>   if (...)
>>>>>     return;
>>>>>
>>>>>   // Do main bulk of work here
>>>>>   ...
>>>>> }
>>>>>
>>>>> Alternative using if/else will force main code in one indentation
>>>>> level.
>>>>>
>>>>> 2. Fallback return at end of block
>>>>>
>>>>> bool foo()
>>>>> {
>>>>>   // Handle case 1
>>>>>   if (...)
>>>>>   {
>>>>>     ...
>>>>>
>>>>>     return true;
>>>>>   }
>>>>>
>>>>>   // Handle case 2
>>>>>   if (...)
>>>>>   {
>>>>>     ...
>>>>>
>>>>>     return true;
>>>>>   }
>>>>>
>>>>>   // No foos found
>>>>>   return false;
>>>>> }
>>>>
>>>>
>>>> Yep, I prefer 2. to an if -- else if (-- else if) -- else.
>>>
>>> It's also logical since the 'else' has no effect. So the policy can be
>>> to remove code that has zero effect.
>>>
>>
>> If it's a straight if - else, I have a pretty strong view that 'else'
>> should be used. C/C++ don't provide 'else' for nothing.
> 
> 'else' does have an effect when there is no return inside the 'if'.
> 
> This works:
> 
>   int a = 1;
>   if (...)
>     a += 1;
>   else
>     a += 2;
> 
> This doesn't work:
> 
>   int a = 1;
>   if (...)
>     a += 1;
>   a += 2;
> 
> But consider this:
> 
>   if (...)
>     return 1;
>   else // this line can be remove and has absolutely no effect
>     return 2;
> 
>> Related, one strict school of thought says that both 1. and 2. are bad
>> since a function should only have one exit. We have had memory leaks
>> because of premature exits.
> 
> I don't think that's a good policy. It would increase the number of
> indentations in general.
> 

I'm not advocating it, but it does have some merits, i.e. using if-else
in simple statements that return (like the one I changed) make it
unambiguous that a function returns from the control block, and it clear
from the one code block how the return value is determined.

Garth


> --
> Anders