Movatterモバイル変換

[0]ホーム
URL:

Skip to content

Navigation Menu

Sign in
Appearance settings

Search code, repositories, users, issues, pull requests...

Provide feedback

We read every piece of feedback, and take your input very seriously.

Saved searches

Use saved searches to filter your results more quickly

 Sign up
Appearance settings

Commitd359f71

Browse files
committed
Fix non-equivalence of VARIADIC and non-VARIADIC function call formats.
For variadic functions (other than VARIADIC ANY), the syntaxes foo(x,y,...)and foo(VARIADIC ARRAY[x,y,...]) should be considered equivalent, since theformer is converted to the latter at parse time. They have indeed beenequivalent, in all releases before 9.3. However, commit75b39e7 made anill-considered decision to record which syntax had been used in FuncExprnodes, and then to make equal() test that in checking node equality ---which caused the syntaxes to not be seen as equivalent by the planner.This is the underlying cause of bug #9817 from Dmitry Ryabov.It might seem that a quick fix would be to make equal() disregardFuncExpr.funcvariadic, but the same commit made that untenable, becausethe field actually *is* semantically significant for some VARIADIC ANYfunctions. This patch instead adopts the approach of redefiningfuncvariadic (and aggvariadic, in HEAD) as meaning that the last argumentis a variadic array, whether it got that way by parser intervention or wassupplied explicitly by the user. Therefore the value will always be truefor non-ANY variadic functions, restoring the principle of equivalence.(However, the planner will continue to consider use of VARIADIC as ameaningful difference for VARIADIC ANY functions, even though some suchfunctions might disregard it.)In HEAD, this change lets us simplify the decompilation logic inruleutils.c, since the funcvariadic/aggvariadic flag tells directly whetherto print VARIADIC. However, in 9.3 we have to continue to cope withexisting stored rules/views that might contain the previous definition.Fortunately, this just means no change in ruleutils.c, since its existingbehavior effectively ignores funcvariadic for all cases other than VARIADICANY functions.In HEAD, bump catversion to reflect the fact that FuncExpr.funcvariadicchanged meanings; this is sort of pro forma, since I don't believe anybuilt-in views are affected.Unfortunately, this patch doesn't magically fix everything for affected9.3 users. After installing 9.3.5, they might need to recreate theirrules/views/indexes containing variadic function calls in order to geteverything consistent with the new definition. As in the cited bug,the symptom of a problem would be failure to use a nominally matchingindex that has a variadic function call in its definition. We'll needto mention this in the 9.3.5 release notes.
1 parent2186533 commitd359f71

File tree

5 files changed

+24
-13
lines changed

5 files changed

+24
-13
lines changed

‎doc/src/sgml/xfunc.sgml

Lines changed: 4 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -3154,9 +3154,10 @@ CREATE OR REPLACE FUNCTION retcomposite(IN integer, IN integer,
31543154
is zero based. <function>get_call_result_type</> can also be used
31553155
as an alternative to <function>get_fn_expr_rettype</>.
31563156
There is also <function>get_fn_expr_variadic</>, which can be used to
3157-
find out whether the call contained an explicit <literal>VARIADIC</>
3158-
keyword. This is primarily useful for <literal>VARIADIC "any"</>
3159-
functions, as described below.
3157+
find out whether variadic arguments have been merged into an array.
3158+
This is primarily useful for <literal>VARIADIC "any"</> functions,
3159+
since such merging will always have occurred for variadic functions
3160+
taking ordinary array types.
31603161
</para>
31613162

31623163
<para>

‎src/backend/parser/parse_func.c

Lines changed: 5 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -374,6 +374,11 @@ ParseFuncOrColumn(ParseState *pstate, List *funcname, List *fargs,
374374
newa->location=exprLocation((Node*)vargs);
375375

376376
fargs=lappend(fargs,newa);
377+
378+
/* We could not have had VARIADIC marking before ... */
379+
Assert(!func_variadic);
380+
/* ... but now, it's a VARIADIC call */
381+
func_variadic= true;
377382
}
378383

379384
/* build the appropriate output structure */

‎src/backend/utils/adt/ruleutils.c

Lines changed: 11 additions & 9 deletions
Original file line numberDiff line numberDiff line change
@@ -8604,7 +8604,7 @@ generate_relation_name(Oid relid, List *namespaces)
86048604
*
86058605
* If we're dealing with a potentially variadic function (in practice, this
86068606
* means a FuncExpr and not some other way of calling the function), then
8607-
* was_variadicmust specify whetherVARIADIC appeared in the original call,
8607+
* was_variadicshould specify whethervariadic arguments have been merged,
86088608
* and *use_variadic_p will be set to indicate whether to print VARIADIC in
86098609
* the output.For non-FuncExpr cases, was_variadic should be FALSE and
86108610
* use_variadic_p can be NULL.
@@ -8639,14 +8639,16 @@ generate_function_name(Oid funcid, int nargs, List *argnames, Oid *argtypes,
86398639
* since it affects the lookup rules in func_get_detail().
86408640
*
86418641
* Currently, we always print VARIADIC if the function is variadic and
8642-
* takes a variadic type other than ANY. (In principle, if VARIADIC
8643-
* wasn't originally specified and the array actual argument is
8644-
* deconstructable, we could print the array elements separately and not
8645-
* print VARIADIC, thus more nearly reproducing the original input. For
8646-
* the moment that seems like too much complication for the benefit.)
8647-
* However, if the function takes VARIADIC ANY, then the parser didn't
8648-
* fold the arguments together into an array, so we must print VARIADIC if
8649-
* and only if it was used originally.
8642+
* takes a variadic type other than ANY. However, if the function takes
8643+
* VARIADIC ANY, then the parser didn't fold the arguments together into
8644+
* an array, so we must print VARIADIC if and only if it was used
8645+
* originally.
8646+
*
8647+
* Note: with the current definition of funcvariadic, we could just set
8648+
* use_variadic = was_variadic, which indeed is the solution adopted in
8649+
* 9.4. However, in rules/views stored before 9.3.5, funcvariadic will
8650+
* reflect the previous definition (was VARIADIC written in the call?).
8651+
* So in 9.3 we cannot trust it unless the function is VARIADIC ANY.
86508652
*/
86518653
if (use_variadic_p)
86528654
{

‎src/backend/utils/fmgr/fmgr.c

Lines changed: 2 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -2452,6 +2452,8 @@ get_call_expr_arg_stable(Node *expr, int argnum)
24522452
* Get the VARIADIC flag from the function invocation
24532453
*
24542454
* Returns false (the default assumption) if information is not available
2455+
*
2456+
* Note this is generally only of interest to VARIADIC ANY functions
24552457
*/
24562458
bool
24572459
get_fn_expr_variadic(FmgrInfo*flinfo)

‎src/include/nodes/primnodes.h

Lines changed: 2 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -346,7 +346,8 @@ typedef struct FuncExpr
346346
Oidfuncid;/* PG_PROC OID of the function */
347347
Oidfuncresulttype;/* PG_TYPE OID of result value */
348348
boolfuncretset;/* true if function returns set */
349-
boolfuncvariadic;/* true if VARIADIC was used in call */
349+
boolfuncvariadic;/* true if variadic arguments have been
350+
* combined into an array last argument */
350351
CoercionFormfuncformat;/* how to display this function call */
351352
Oidfunccollid;/* OID of collation of result */
352353
Oidinputcollid;/* OID of collation that function should use */

0 commit comments

Comments
 (0)
[8]ページ先頭
©2009-2025 Movatter.jp