aboutsummaryrefslogtreecommitdiff
path: root/src/backend/optimizer/util/pathnode.c
diff options
context:
space:
mode:
authorTom Lane <tgl@sss.pgh.pa.us>2019-02-07 13:10:46 -0500
committerTom Lane <tgl@sss.pgh.pa.us>2019-02-07 13:11:17 -0500
commit1eeee69099e0402722457534df4744d8679bbe23 (patch)
tree52242659368ee22ead08b507999fd4b1d5e818cc /src/backend/optimizer/util/pathnode.c
parent7b8c6ef1e05fede7123bed5a3af4b780da241b4e (diff)
downloadpostgresql-1eeee69099e0402722457534df4744d8679bbe23.tar.gz
postgresql-1eeee69099e0402722457534df4744d8679bbe23.zip
Ensure that foreign scans with lateral refs are planned correctly.
As reported in bug #15613 from Srinivasan S A, file_fdw and postgres_fdw neglected to mark plain baserel foreign paths as parameterized when the relation has lateral_relids. Other FDWs have surely copied this mistake, so rather than just patching those two modules, install a band-aid fix in create_foreignscan_path to rectify the mistake centrally. Although the band-aid is enough to fix the visible symptom, correct the calls in file_fdw and postgres_fdw anyway, so that they are valid examples for external FDWs. Also, since the band-aid isn't enough to make this work for parameterized foreign joins, throw an elog(ERROR) if such a case is passed to create_foreignscan_path. This shouldn't pose much of a problem for existing external FDWs, since it's likely they aren't trying to make such paths anyway (though some of them may need a defense against joins with lateral_relids, similar to the one this patch installs into postgres_fdw). Add some assertions in relnode.c to catch future occurrences of the same error --- in particular, as backstop against core-code mistakes like the one fixed by commit bdd9a99aa. Discussion: https://postgr.es/m/15613-092be1be9576c728@postgresql.org
Diffstat (limited to 'src/backend/optimizer/util/pathnode.c')
-rw-r--r--src/backend/optimizer/util/pathnode.c23
1 files changed, 23 insertions, 0 deletions
diff --git a/src/backend/optimizer/util/pathnode.c b/src/backend/optimizer/util/pathnode.c
index e74aa094c02..af0c35876dc 100644
--- a/src/backend/optimizer/util/pathnode.c
+++ b/src/backend/optimizer/util/pathnode.c
@@ -1467,6 +1467,29 @@ create_foreignscan_path(PlannerInfo *root, RelOptInfo *rel,
{
ForeignPath *pathnode = makeNode(ForeignPath);
+ /*
+ * Since the path's required_outer should always include all the rel's
+ * lateral_relids, forcibly add those if necessary. This is a bit of a
+ * hack, but up till early 2019 the contrib FDWs failed to ensure that,
+ * and it's likely that the same error has propagated into many external
+ * FDWs. Don't risk modifying the passed-in relid set here.
+ */
+ if (rel->lateral_relids && !bms_is_subset(rel->lateral_relids,
+ required_outer))
+ required_outer = bms_union(required_outer, rel->lateral_relids);
+
+ /*
+ * Although this function is only designed to be used for scans of
+ * baserels, before v12 postgres_fdw abused it to make paths for join and
+ * upper rels. It will work for such cases as long as required_outer is
+ * empty (otherwise get_baserel_parampathinfo does the wrong thing), which
+ * fortunately is the expected case for now.
+ */
+ if (!bms_is_empty(required_outer) &&
+ !(rel->reloptkind == RELOPT_BASEREL ||
+ rel->reloptkind == RELOPT_OTHER_MEMBER_REL))
+ elog(ERROR, "parameterized foreign joins are not supported yet");
+
pathnode->path.pathtype = T_ForeignScan;
pathnode->path.parent = rel;
pathnode->path.param_info = get_baserel_parampathinfo(root, rel,