Re: context dependencies

[Ricky Zhang]

Hi John,

Your suggestion inspired me. I came up a simpler solution: instead of
eating spaces after NUMBER or ID, I did it in binary operators (such as
PLUS_SIGN '+') and special signs (such as LEFT_ROUND_BRACKET '(' ).

In Flex file only,

space       [ ]

blank       {space}*

...

%%

...

{blank}"+"{blank}    {return PLUS;}

"("{blank}           {return LEFT_ROUND_BRACKET;}

...

{space}+             {return SINGLE_SPACE_OP;}


The reason I do it this way is because expr non-terminal in my
production rule is very complicated.

It is difficult for me to trace and tag them. But the number of signs
separating them are limited and clear.

I tested a couple of cases. It can parse. I need more test to confirm this.


In any case, thanks Ron and John for your help!


Cheers,

Ricky




On Fri, Dec 2, 2016 at 2:55 AM, John P. Hartmann <address@hidden>
wrote:

> You will have to do that in the scanner using exclusive start states. When
> it meets blanks after a NUMBER or ID, it emits the SPACE token if the
> beginning of a number or identifier follows the blank(s):
>
> <afternumid>{
> [ \t]+/[a-z0-9]         BEGIN(INITIAL); return SPACE;
> [ \t]+                  BEGIN(INITIAL);
>
> }
>
> On 12/02/2016 05:19 AM, Ricky Zhang wrote:
>
>> I have read through 'handling context dependencies' section Bison manual.
>> But I still haven't figured out the proper solution to my problem.
>>
>> We have production rule:
>>
>> expr : expr SINGLE_SPACE expr { /* some operation */ }
>>         | expr + expr
>>         ...
>>         ;
>>
>> expr: NUMBER
>>        | ID
>>        ;
>>
>> and also for all white space [SINGLE_SPACE | \t]* in the language should
>> be
>> ignored. Are there any easy way to solve it?
>>
>> Thanks,
>> Ricky
>> _______________________________________________
>> address@hidden https://lists.gnu.org/mailman/listinfo/help-bison
>>
>>