Re: Boolean partitions syntax

Hello. Thank you for the comment.

the attached v6 patch differs only in gram.y since v5.

At Fri, 13 Apr 2018 18:55:30 +0900, Amit Langote <Langote_Amit_f8(at)lab(dot)ntt(dot)co(dot)jp> wrote in <ca37d679-014e-1895-e4bc-89b7129ce58b(at)lab(dot)ntt(dot)co(dot)jp>
> Horiguchi-san,
>
> Thanks for the latest patch.
>
> On 2018/04/12 13:12, Kyotaro HORIGUCHI wrote:
> > Thank you for verification and the revised patch. The format is
> > fine and the fix is correct but I noticed that I forgot to remove
> > plural S's from error messages. The attached is the version with
> > the fix.
>
> Looking at the gram.y changes in the latest patch, I think there needs to
> be some explanatory comments about about the new productions -- u_expr,
> b0_expr, and c0_expr.

I think I did that. And refactord the rules.

It was a bother that some rules used c_expr directly but I
managed to replace all of them with a_expr by lowering precedence
of some ordinary keywords (PASSING, BY, COLUMNS and ROW). c_expr
is no loger used elsewhere so we can just remove columnref from
c_expr. Finally [abc]0_expr was eliminated and we have only
a_expr, b_expr, u_expr and c_expr. This seems simple enough.

The relationship among the rules after this change is as follows.

a_expr --+-- columnref
+-- u_expr ---+-- c_expr -- (all old c_expr stuff except columnref)
+-- (all old a_expr stuff)

b_expr --+-- columnref
+-- c_expr -- (ditto)
+-- (all old b_expr stuff)

On the way fixing this, I noticed that the precedence of some
keywords (PRESERVE, STRIP_P) that was more than necessary and
corrected it.

regards,

--
Kyotaro Horiguchi
NTT Open Source Software Center