[HACKERS] Interesting parsing problem ...

[Hans-Jürgen Schönig]

I am willing to add NOWAIT to a couple of commands and I have tried to 
resolve a bison problem for quite some time now:

As a first step I wanted to add NOWAIT to DELETE:
DELETE FROM ... WHERE ... NOWAIT;
Therefore I used:
/*****************************************************************************
 *
 *              QUERY:
 *                              DELETE STATEMENTS
 *
*****************************************************************************/
DeleteStmt: DELETE_P FROM relation_expr where_clause opt_nowait
opt_nowait is already defined and used by Tatsuo's LOCK TABLE NOWAIT.
I got ...
[EMAIL PROTECTED] parser]$ bison -y -d -v gram.y
conflicts: 6 shift/reduce
... 6 errors:
State 1197 conflicts: 1 shift/reduce
State 1198 conflicts: 1 shift/reduce
State 1201 conflicts: 1 shift/reduce
State 1852 conflicts: 1 shift/reduce
State 1853 conflicts: 1 shift/reduce
State 1855 conflicts: 1 shift/reduce
...
  1196           | BCONST
  1197           | XCONST
...
  1201           | TRUE_P
...
state 1852
  1024 b_expr: b_expr '%' .
  1030       | b_expr '%' . b_expr
    ABORT_P            shift, and go to state 146
    ABSOLUTE_P         shift, and go to state 147
    ACCESS             shift, and go to state 148
    ACTION             shift, and go to state 149

The interesting thing here is - if I change opt_nowait to
DeleteStmt: DELETE_P FROM relation_expr where_clause opt_lock (just for 
tesing), I still get an error ...

[EMAIL PROTECTED] parser]$ bison -y -d -v gram.y
conflicts: 1 shift/reduce
In my understanding of bison DELETE ... WHERE ... IN some_mode MODE;
should work ...
Can anybody provide me a fix?
Basically all I wanted to do was SELECT FOR UPDATE NOWAIT, DELETE NOWAIT 
and UPDATE ... NOWAIT.
The rest of the patch seems to be fairly simple but can anybody lead me 
out of "parser's hell"?

        Best regards,
                Hans

---------------------------(end of broadcast)---------------------------
TIP 9: the planner will ignore your desire to choose an index scan if your
     joining column's datatypes do not match