Re: Bison 3.0: enhancement request proposal

[Akim Demaille]

Hi Francis,

> Le 30 août 2015 à 12:21, Francis ANDRE <address@hidden> a écrit :
> 
> Hi
> 
> I am using Bison for a grammar that brings conflicting names like the 
> following use case:
> 
> %token IDENTIFIER "IDENTIFIER"   // the keyword IDENTIFIER
> %token identifier                             // the terminal identifier for 
> lower case name
> 
> Bison  generates then
> 
> #ifndef YYTOKENTYPE
> # define YYTOKENTYPE
>  enum yytokentype
>  {
>    IDENTIFIER = 100,
>    identifier = 101,
> ...
> 
> fine... but I would like to use the name "identifier" in some parser C++ code 
> and that's not possible because it is already used in the enum yytokentype. 
> Thus I am proposing to add a directive to Bison for specifying the list of 
> token as an C++ enum class type as
> 
> % enum-class = "MyTokens";
> 
> so that Bison would generate
> 
> #ifndef YYTOKENTYPE
> # define YYTOKENTYPE
>  enum class MyTokens
>  {
>    IDENTIFIER = 100,
>    identifier = 101,
> ...
> 
> and replace yytokentype by MyTokens everywhere. As such, the token would need 
> to be qualified by MyTokens:: in the lexer and in semantic actions and that 
> exactly what is needed to keep the name "identifier" used out of the parser's 
> code.
> 
> Regards

For the records, Bison does that with C++ parsers.  Otherwise,
in C, you should have a look at %define api.token.prefix.

Cheers!