We could do something like this. Is this good?
I tried to merge WITH and WITHOUT with the precedence class immediately
above, but that failed: the main grammar compiles fine and no tests
fail, but ECPG does fail to compile the sqljson.pgc test, so there's
some problem there. Now, the ecpg grammar stuff *is* absolute black
magic to me, so I have no idea what to do about that.
(TBH I don't think the added comments really explain the problems fully.
That's most likely because I don't actually understand what the problems
are.)
--
Álvaro Herrera 48°01'N 7°57'E — https://www.EnterpriseDB.com/
Thou shalt study thy libraries and strive not to reinvent them without
cause, that thy code may be short and readable and thy days pleasant
and productive. (7th Commandment for C Programmers)
>From 3499494625a59e2b5bda85c207b2d0e1181d703d Mon Sep 17 00:00:00 2001
From: Alvaro Herrera <alvhe...@alvh.no-ip.org>
Date: Mon, 27 Nov 2023 18:01:44 +0100
Subject: [PATCH v1] Merge JSON-related grammar keyword precedence classes
Some of these classes were unnecessary. Grouping them this way also
helps us document the reason for these to exist at all.
---
src/backend/parser/gram.y | 20 ++++++++++++++------
1 file changed, 14 insertions(+), 6 deletions(-)
diff --git a/src/backend/parser/gram.y b/src/backend/parser/gram.y
index 8c00b119ec..3060dc88f8 100644
--- a/src/backend/parser/gram.y
+++ b/src/backend/parser/gram.y
@@ -821,11 +821,6 @@ static Node *makeRecursiveViewSelect(char *relname, List *aliases, Node *query);
%nonassoc BETWEEN IN_P LIKE ILIKE SIMILAR NOT_LA
%nonassoc ESCAPE /* ESCAPE must be just above LIKE/ILIKE/SIMILAR */
-/* SQL/JSON related keywords */
-%nonassoc UNIQUE JSON
-%nonassoc KEYS OBJECT_P SCALAR VALUE_P
-%nonassoc WITH WITHOUT
-
/*
* To support target_el without AS, it used to be necessary to assign IDENT an
* explicit precedence just less than Op. While that's not really necessary
@@ -850,9 +845,22 @@ static Node *makeRecursiveViewSelect(char *relname, List *aliases, Node *query);
* an explicit priority lower than '(', so that a rule with CUBE '(' will shift
* rather than reducing a conflicting rule that takes CUBE as a function name.
* Using the same precedence as IDENT seems right for the reasons given above.
+ *
+ * We also need to support WITH UNIQUE KEYS and WITHOUT UNIQUE KEYS, where
+ * KEYS is optional and the whole rule is also optional. We can do this by
+ * putting UNIQUE above KEYS, and that one above WITH and WITHOUT. We also
+ * have to declare the precedence of the empty production to be that of KEYS.
+ * We arbitrarily choose to put UNIQUE together with UNBOUNDED, and KEYS
+ * together with IDENT, then WITH and WITHOUT in a class of their own.
+ *
+ * For IS JSON {VALUE,OBJECT,SCALAR}, we need to split out the latter three
+ * from JSON; we put JSON with UNBOUNDED and the other three keywords in the
+ * group below.
*/
-%nonassoc UNBOUNDED /* ideally would have same precedence as IDENT */
+%nonassoc UNBOUNDED UNIQUE JSON /* ideally would have same precedence as IDENT */
%nonassoc IDENT PARTITION RANGE ROWS GROUPS PRECEDING FOLLOWING CUBE ROLLUP
+ KEYS OBJECT_P SCALAR VALUE_P
+%nonassoc WITH WITHOUT
%left Op OPERATOR /* multi-character ops and user-defined operators */
%left '+' '-'
%left '*' '/' '%'
--
2.39.2
