van Wijngaarden grammars

[Jose E. Marchesi]

I see each struct hypernotion has hlen numbers of struct hitems linked
in hdef, where:

struct hitem
  { enum symbol sy;		/* s_ssm or s_meta   */
    int it_s;			/* if sy == s_ssm    */
    struct dictnode *it_z;	/* if sy == s_meta   */
    bool isbound;		/* used by checkgr   */
  };

So each ssm is an int.

>> Why not tokenizing protonotions rather than small syntactic marks?
>
> The problem is, I don't understand the program enough to make big
> changes.  The Fisher-parser operates on metanotions and ssms.
> Checkll1 computes starter sets for each  alternative by setting a
> bit for 1st ssm in each and checks, that these are disjoint. So the
> number of ssms should fit into a machine word.  Today we may safely
> assume that 64 bits are available.

I see that checkll1 does:

 For each metanotion in the metalist,
   For each rhs that produces the metanotion,
     hn = hypernotion in the rhs
     m1 = starters (hn, 0)

where starters returns a bitmap reflecting the presence of all ssm in
the rhs hypernotion, as well as the starters of the metanotions in the
rhs hypernotion.  But I don't see it checks the "1st ssm in each".  The
loop in `starters' process all the ssm/metanotions of the given
hypernotions, starting at k...

> My 1st idea to do it was to compute a list of all tokens occurring
> in the a68 grammar, by using blanks as delimiters. But this list
> was about 700 or more long, even after exclusion of formatted
> transput. Some of this bloat comes from the metanotions MONAD and
> NOMAD. Even some modes in metanotion STANDARD, don't appear elsewhere
> in the grammar.
>
> But before I tackle these things, I must know whether it is worth the
> trouble. So, the next step after the metanotion LL(1)-problem is the
> application of a small subset of hypernotions excluding some of the
> peculiar features like UNITED mode or FLEX rows, composed operator
> tags and so on.
>
> BTW. I found one other guy working with the Fisher parser:
> https://github.com/vonbecmann/fisher-parser
> I don't know enough about smalltalk to appreciate his work, and I did
> not find out, on which grammars he intended to apply.