rdelaage
/
postgresql
mirror of https://git.postgresql.org/git/postgresql.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Fix get_expr_result_type() to find field names for RECORD Consts. 

This is a back-patch of commit d57534740 ("Fix EXPLAIN of SEARCH
BREADTH FIRST with a constant initial value") into pre-v14 branches.
At the time I'd thought it was not needed in branches that lack the
SEARCH/CYCLE feature, but that was just a failure of imagination.
It's possible to demonstrate "record type has not been registered"
failures in older branches too, during deparsing of views that contain
references to fields of composite constants.

Back-patch only the code changes, as the test cases added by d57534740
all require SEARCH/CYCLE syntax.  A suitable test case will be added
in the upcoming fix for bug #18077.

Discussion: https://postgr.es/m/17644-3bd1f3036d6d7a16@postgresql.org
Discussion: https://postgr.es/m/3607145.1694803130@sss.pgh.pa.us
This commit is contained in:
Tom Lane 2023-09-15 16:39:27 -04:00
parent ece1154f4c
commit 479b99125d
2 changed files with 36 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
src/backend/utils/adt/ruleutils.c
View File

@ -6944,7 +6944,8 @@ get_name_for_var_field(Var *var, int fieldno,
/*
* If it's a RowExpr that was expanded from a whole-row Var, use the
* column names attached to it.
* column names attached to it. (We could let get_expr_result_tupdesc()
* handle this, but it's much cheaper to just pull out the name we need.)
*/
if (IsA(var, RowExpr))
{

34
src/backend/utils/fmgr/funcapi.c
View File

@ -283,6 +283,40 @@ get_expr_result_type(Node *expr,
*resultTupleDesc = BlessTupleDesc(tupdesc);
return TYPEFUNC_COMPOSITE;
}
else if (expr && IsA(expr, Const) &&
((Const *) expr)->consttype == RECORDOID &&
!((Const *) expr)->constisnull)
{
/*
* When EXPLAIN'ing some queries with SEARCH/CYCLE clauses, we may
* need to resolve field names of a RECORD-type Const. The datum
* should contain a typmod that will tell us that.
*/
HeapTupleHeader rec;
Oid tupType;
int32 tupTypmod;
rec = DatumGetHeapTupleHeader(((Const *) expr)->constvalue);
tupType = HeapTupleHeaderGetTypeId(rec);
tupTypmod = HeapTupleHeaderGetTypMod(rec);
if (resultTypeId)
*resultTypeId = tupType;
if (tupType != RECORDOID || tupTypmod >= 0)
{
/* Should be able to look it up */
if (resultTupleDesc)
*resultTupleDesc = lookup_rowtype_tupdesc_copy(tupType,
tupTypmod);
return TYPEFUNC_COMPOSITE;
}
else
{
/* This shouldn't really happen ... */
if (resultTupleDesc)
*resultTupleDesc = NULL;
return TYPEFUNC_RECORD;
}
}
else
{
/* handle as a generic expression; no chance to resolve RECORD */