Re: Parsing input from a string...

[Dave Trombley]

Dave Trombley wrote:

>    I'm trying to do some stuff with bison and flex for the first time, 
> and can't seem to figure this out:  What I'd like to do is to parse 
> input from a char*, rather than a FILE*.  Of course, I could write my 
> string to a temporary file, but this is very poor for obvious reasons. 
> Is there any way at all for me to accomplish this without hand editing 
> the yylex() function?  (I'll need to also not use global variables, as 
> I have a multithreaded program, and many threads will be parsing 
> potentially many different grammars.)


   I've actually made some progress with this, but I've run into a 
problem I don't know how to work around; however it's a bit more 
concrete than what I posted above!

  I'm using

#define YYPARSE_PARAM parm
#define YYLEX_PARAM parm

  to create those parm variables, which I intend to be of a type that 
looks something like this:

typedef struct {
   char *target;
   int pos;
   int length;
   } parsebuf;

  The idea was to pass a structure of that defined type to yyparse(), 
and in term yyparse() would pass it on to yylex().  I had hoped that 
this would suffice it for use in my YY_INPUT macro, which looks like this:

#define YY_INPUT(buf,res,msize)                 \
   {                             \
     parsebuf *pb = (parsebuf*)parm;            \
     int chars = pb->len - pb->pos;            \
     if (msize < chars) chars = msize;            \
     memcpy(buf,&pb->buf[pb->pos],chars);            \
     pb->pos += chars;                    \
     if (pb->pos >= pb->length) res = YY_NULL;        \
     res = chars;                        \
   }

   It seemed I was all set!  But alas:

mungbean:~/schema$ make
gcc -g  -c lex.yy.c schema.tab.c
lex.yy.c: In function `yy_get_next_buffer':
lex.yy.c:929: `parm' undeclared (first use in this function)
lex.yy.c:929: (Each undeclared identifier is reported only once
lex.yy.c:929: for each function it appears in.)
lex.yy.c:929: parse error before `('
lex.yy.c:929: structure has no member named `length'
make: *** [yyparse.a] Error 1

YY_INPUT is called from yy_get_next_buffer(), and there doesn't seem to 
be a way to tell yylex() to pass on parm to it.  So, my questions:

   1) Is there any cleaner way to make the YYLEX_PARAM variable 
available for use in YY_INPUT than running a script on lex.yy.c to make 
yylex() pass the parameter onto yy_get_next_buffer()?
  
   2) Is it inadvisable to want to do this for some reason?  (say, 
supposing there is an easier way to do this sort of thing then I'm 
attempting...)

   -dj trombley
     <address@hidden>