Re: Materialized views WIP patch

Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
> Kevin Grittner <kgrittn(at)ymail(dot)com> writes:
>> When I went to do this, I hit a shift/reduce conflict, because
>> with TABLE being optional it couldn't tell whether:
>
>> TRUNCATE MATERIALIZED VIEW x, y, z;
>
>> ... was looking for five relations or three.  That goes away
>> with MATERIALIZED escalated to TYPE_FUNC_NAME_KEYWORD.  Is that
>> OK?
>
> Not really.  I would much rather see us not bother with this
> pedantic syntax than introduce an even-partially-reserved word.

I'm not sure it's worth it either; but two people requested it and
I didn't forsee this shift/reduce conflict, so I took a shot at it.
 If we can't eliminate the conflict, I'm fine with leaving things
as they are in the latest posted patch.

> Having said that, I don't think I believe your analysis of why
> this doesn't work.  The presence or absence of commas ought to
> make the syntax non-ambiguous, I would think.  Maybe you just
> factored the grammar wrong.

Well, it wouldn't be the first time you've seen a better way to do
something in flex than I was able to see.  Taking just the gram.y
part of the change which implemented this, and omitting the change
in reservedness of MATERIALIZED, I have:

diff --git a/src/backend/parser/gram.y b/src/backend/parser/gram.y
index 820cb41..1d393c5 100644
--- a/src/backend/parser/gram.y
+++ b/src/backend/parser/gram.y
@@ -394,6 +394,7 @@ static Node *makeRecursiveViewSelect(char *relname, List *aliases, Node *query);

 %type <ival>   opt_column event cursor_options opt_hold opt_set_data
 %type <objtype>    reindex_type drop_type comment_type security_label_type
+               trunc_type

 %type <node>   fetch_args limit_clause select_limit_value
                offset_clause select_offset_value
@@ -5172,9 +5173,10 @@ attrs:       '.' attr_name
  *****************************************************************************/

 TruncateStmt:
-           TRUNCATE opt_table relation_expr_list opt_restart_seqs opt_drop_behavior
+           TRUNCATE trunc_type relation_expr_list opt_restart_seqs opt_drop_behavior
                {
                    TruncateStmt *n = makeNode(TruncateStmt);
+                   n->objtype = $2;
                    n->relations = $3;
                    n->restart_seqs = $4;
                    n->behavior = $5;
@@ -5182,6 +5184,12 @@ TruncateStmt:
                }
        ;

+trunc_type:
+           TABLE                       { $$ = OBJECT_TABLE; }
+           | MATERIALIZED VIEW         { $$ = OBJECT_MATVIEW; }
+           | /*EMPTY*/                 { $$ = OBJECT_UNSPECIFIED; }
+       ;
+
 opt_restart_seqs:
            CONTINUE_P IDENTITY_P       { $$ = false; }
            | RESTART IDENTITY_P        { $$ = true; }

I'm open to suggestions on a better way.

--
Kevin Grittner
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company