> > I have gram.y patches to do this. No visible effect on the regression execution speed,
> > which thanks to Bruce is on my machine down to a record 2:23 elapsed execution time;
> > the last speeds were in the 2:31 range.
> >
> > I'm planning on bumping the gram.y cvs version to 2.0 for this change, due to the
> > _large_ number of accumulated improvements. Any objections?
>
> Go ahead. Are there security problems?
Security problems with the patch? No, I don't think so. The only thing I'm doing is
implementing xlateSqlFunc() similarly to the existing xlateSqlType(), and defining a
"func_name" state in place of just "name" which was used before.
The only downside is that backend functions implemented with names "char_length" and
"character_length" will be masked by the parser translation so cannot be executed. In this
case I think that is OK, but I should put something in the new docs about that :)
- Tom