[vwnc] [workaround] for a refactoring "bug"

> Here's a workaround for extracting pieces of a long series of binary
> message sends into temps & methods. Example method:
>
> giantSpeak
>        Giant new
>         hollar:
>             'fee' , 'fie' , 'foe' , 'fum' , 'I' , 'smell' , 'the' ,
> 'blood' , 'of' ,
>                     'an' , ' Englishman'.
>     Jack new
>         say: 'Do' , 'you' , 'know' , 'an' , 'Englishman' , '?'
>         to: GiantWife new
>
> Being a good programmer, I might want to extract    'an', 'Englishman'  
>    to another method or temp for reuse in my story. But when I try to
> select that code & do Extract Method, I get a "Could not extract code
> from method" error. "Extract to Temporary" grabs the entire series of
> binary messages, which is wrong.
>
> The problem is that the refactoring engine creates a parse tree and
> does parse tree matching. And basically makes a message node with
>
>   'Do', 'you', 'know', 'an' as the receiver
> #, as the selector
> 'Englishman' as the argument
>
> Then it never matches the selected string.
>
> This is not a refactoring from a smalltalk parsing point of view, but I
>   know that semantically my "refactoring" is equivalent.
>
> So, the workaround is to recompile the method with ()'s around the part
> to extract. Then select inside the parens and extract. The RB likes
> this is much better & complies with your directive.
>
> It would be interesting to try to do this automagically if an
> extraction failed. Insert ()'s in the source at the selection point,
> and if the resulting source is valid syntax, the match should work and
> the "refactoring" could continue.
>
> See ExtractMethodRefactoring>>extractMethod and step down the long
> parse tree matching to see what I'm talking about.
>
> eric
>
> _______________________________________________
> vwnc mailing list
> [hidden email]
> http://lists.cs.uiuc.edu/mailman/listinfo/vwnc
>
>
>  

>
> The problem is that you understand the semantics of the , message but
> the Refactoring Browser doesn't.  The RB doesn't know that the , message
> is associative.  It has no reason to believe that
>     ('a' , 'b') , 'c'
> is equivalent to
>     'a' , ('b' , 'c')
>
> Smalltalk obviously interprets
>     'a' , 'b' , 'c'   as ('a', 'b') , 'c'
>
> You are selecting the 'b' , 'c' part and asking the RB to trust you that
> the operation is associative.
>
> What would you expect the RB to do in this case:
>
>     a * b + c
>
> if you select b + c and ask it to do an extract method?
>
> David Buck

> The problem is that you understand the semantics of the , message but
> the Refactoring Browser doesn't.  The RB doesn't know that the ,
> message is associative.  It has no reason to believe that
>    ('a' , 'b') , 'c'
> is equivalent to
>    'a' , ('b' , 'c')
>
> Smalltalk obviously interprets
>    'a' , 'b' , 'c'   as ('a', 'b') , 'c'
>
> You are selecting the 'b' , 'c' part and asking the RB to trust you
> that the operation is associative.
>
> What would you expect the RB to do in this case:
>
>    a * b + c
>
> if you select b + c and ask it to do an extract method?

>
> David Buck
> Simberon Inc.
> www.simberon.com
>
> Eric Winger wrote:
>> Here's a workaround for extracting pieces of a long series of binary
>> message sends into temps & methods. Example method:
>>
>> giantSpeak
>>        Giant new
>>         hollar:
>>             'fee' , 'fie' , 'foe' , 'fum' , 'I' , 'smell' , 'the' ,
>> 'blood' , 'of' ,
>>                     'an' , ' Englishman'.
>>     Jack new
>>         say: 'Do' , 'you' , 'know' , 'an' , 'Englishman' , '?'
>>         to: GiantWife new
>>
>> Being a good programmer, I might want to extract    'an',
>> 'Englishman'     to another method or temp for reuse in my story. But
>> when I try to select that code & do Extract Method, I get a "Could
>> not extract code from method" error. "Extract to Temporary" grabs the
>> entire series of binary messages, which is wrong.
>>
>> The problem is that the refactoring engine creates a parse tree and
>> does parse tree matching. And basically makes a message node with
>>
>>   'Do', 'you', 'know', 'an' as the receiver
>> #, as the selector
>> 'Englishman' as the argument
>>
>> Then it never matches the selected string.
>>
>> This is not a refactoring from a smalltalk parsing point of view, but
>> I   know that semantically my "refactoring" is equivalent.
>>
>> So, the workaround is to recompile the method with ()'s around the
>> part to extract. Then select inside the parens and extract. The RB
>> likes this is much better & complies with your directive.
>>
>> It would be interesting to try to do this automagically if an
>> extraction failed. Insert ()'s in the source at the selection point,
>> and if the resulting source is valid syntax, the match should work
>> and the "refactoring" could continue.
>>
>> See ExtractMethodRefactoring>>extractMethod and step down the long
>> parse tree matching to see what I'm talking about.
>>
>> eric
>>
>> _______________________________________________
>> vwnc mailing list
>> [hidden email]
>> http://lists.cs.uiuc.edu/mailman/listinfo/vwnc
>>
>>
>>
>
>