Re: Row pattern recognition

On Sat, 09 Dec 2023 07:22:58 +0900 (JST)
Tatsuo Ishii <ishii@sraoss.co.jp> wrote:

> > On 04.12.23 12:40, Tatsuo Ishii wrote:
> >> diff --git a/src/backend/parser/gram.y b/src/backend/parser/gram.y
> >> index d631ac89a9..5a77fca17f 100644
> >> --- a/src/backend/parser/gram.y
> >> +++ b/src/backend/parser/gram.y
> >> @@ -251,6 +251,8 @@ static Node *makeRecursiveViewSelect(char
> >> *relname, List *aliases, Node *query);
> >>   	DefElem	   *defelt;
> >>   	SortBy	   *sortby;
> >>   	WindowDef  *windef;
> >> +	RPCommonSyntax	*rpcom;
> >> +	RPSubsetItem	*rpsubset;
> >>   	JoinExpr   *jexpr;
> >>   	IndexElem  *ielem;
> >>   	StatsElem  *selem;
> >> @@ -278,6 +280,7 @@ static Node *makeRecursiveViewSelect(char
> >> *relname, List *aliases, Node *query);
> >>   	MergeWhenClause *mergewhen;
> >>   	struct KeyActions *keyactions;
> >>   	struct KeyAction *keyaction;
> >> +	RPSkipTo	skipto;
> >>   }
> >>     %type <node>	stmt toplevel_stmt schema_stmt routine_body_stmt
> > 
> > It is usually not the style to add an entry for every node type to the
> > %union.  Otherwise, we'd have hundreds of entries in there.
> 
> Ok, I have removed the node types and used existing node types.  Also
> I have moved RPR related %types to same place to make it easier to know
> what are added by RPR.
> 
> >> @@ -866,6 +878,7 @@ static Node *makeRecursiveViewSelect(char
> >> *relname, List *aliases, Node *query);
> >>   %nonassoc UNBOUNDED /* ideally would have same precedence as IDENT */
> >>   %nonassoc IDENT PARTITION RANGE ROWS GROUPS PRECEDING FOLLOWING CUBE
> >>   %ROLLUP
> >>   			SET KEYS OBJECT_P SCALAR VALUE_P WITH WITHOUT
> >> +%nonassoc	MEASURES AFTER INITIAL SEEK PATTERN_P
> >>   %left Op OPERATOR /* multi-character ops and user-defined operators */
> >>   %left		'+' '-'
> >>   %left		'*' '/' '%'
> > 
> > It was recently discussed that these %nonassoc should ideally all have
> > the same precedence.  Did you consider that here?
> 
> No, I didn't realize it. Thanks for pointing it out. I have removed
> %nonassoc so that MEASURES etc. have the same precedence as IDENT etc.
> 
> Attached is the new diff of gram.y against master branch.

Thank you very much for providing the patch for the RPR implementation.

After applying the v12-patches, I noticed an issue that
the rpr related parts in window clauses were not displayed in the
view definitions (the definition column of pg_views).

To address this, I have taken the liberty of adding an additional patch
that modifies the relevant rewriter source code.
I have attached the rewriter patch for your review and would greatly appreciate your feedback.

Thank you for your time and consideration.

-- 
SRA OSS LLC
Ningwei Chen <chen@sraoss.co.jp>
TEL: 03-5979-2701 FAX: 03-5979-2702