diff options
author | Alvaro Herrera <alvherre@alvh.no-ip.org> | 2018-07-10 15:07:28 -0400 |
---|---|---|
committer | Alvaro Herrera <alvherre@alvh.no-ip.org> | 2018-07-10 15:16:27 -0400 |
commit | e7df94f317c979ad176adc0155a3e8af63e68ac9 (patch) | |
tree | 3bffd42ebdcac79f2c1e277e556c85a7528c6567 /src/backend/partitioning/partbounds.c | |
parent | 4cf30f6fdeed4977dfe52bf1d4fda74c883d3760 (diff) |
Better handle pseudotypes as partition keys
We fail to handle polymorphic types properly when they are used as
partition keys: we were unnecessarily adding a RelabelType node on top,
which confuses code examining the nodes. In particular, this makes
predtest.c-based partition pruning not to work, and ruleutils.c to emit
expressions that are uglier than needed. Fix it by not adding RelabelType
when not needed.
In master/11 the new pruning code is separate so it doesn't suffer from
this problem, since we already fixed it (in essentially the same way) in
e5dcbb88a15d, which also added a few tests; back-patch those tests to
pg10 also. But since UPDATE/DELETE still uses predtest.c in pg11, this
change improves partitioning for those cases too. Add tests for this.
The ruleutils.c behavior change is relevant in pg11/master too.
Co-authored-by: Amit Langote <Langote_Amit_f8@lab.ntt.co.jp>
Co-authored-by: Álvaro Herrera <alvherre@alvh.no-ip.org>
Reviewed-by: Álvaro Herrera <alvherre@alvh.no-ip.org>
Reviewed-by: Robert Haas <robertmhaas@gmail.com>
Discussion: https://postgr.es/m/54745d13-7ed4-54ac-97d8-ea1eec95ae25@lab.ntt.co.jp
Diffstat (limited to 'src/backend/partitioning/partbounds.c')
-rw-r--r-- | src/backend/partitioning/partbounds.c | 50 |
1 files changed, 20 insertions, 30 deletions
diff --git a/src/backend/partitioning/partbounds.c b/src/backend/partitioning/partbounds.c index b19c76acc8a..9015a05d323 100644 --- a/src/backend/partitioning/partbounds.c +++ b/src/backend/partitioning/partbounds.c @@ -1144,8 +1144,12 @@ get_partition_bound_num_indexes(PartitionBoundInfo bound) /* * get_partition_operator * - * Return oid of the operator of given strategy for a given partition key - * column. + * Return oid of the operator of the given strategy for the given partition + * key column. It is assumed that the partitioning key is of the same type as + * the chosen partitioning opclass, or at least binary-compatible. In the + * latter case, *need_relabel is set to true if the opclass is not of a + * polymorphic type (indicating a RelabelType node needed on top), otherwise + * false. */ static Oid get_partition_operator(PartitionKey key, int col, StrategyNumber strategy, @@ -1154,40 +1158,26 @@ get_partition_operator(PartitionKey key, int col, StrategyNumber strategy, Oid operoid; /* - * First check if there exists an operator of the given strategy, with - * this column's type as both its lefttype and righttype, in the - * partitioning operator family specified for the column. + * Get the operator in the partitioning opfamily using the opclass' + * declared input type as both left- and righttype. */ operoid = get_opfamily_member(key->partopfamily[col], - key->parttypid[col], - key->parttypid[col], + key->partopcintype[col], + key->partopcintype[col], strategy); + if (!OidIsValid(operoid)) + elog(ERROR, "missing operator %d(%u,%u) in partition opfamily %u", + strategy, key->partopcintype[col], key->partopcintype[col], + key->partopfamily[col]); /* - * If one doesn't exist, we must resort to using an operator in the same - * operator family but with the operator class declared input type. It is - * OK to do so, because the column's type is known to be binary-coercible - * with the operator class input type (otherwise, the operator class in - * question would not have been accepted as the partitioning operator - * class). We must however inform the caller to wrap the non-Const - * expression with a RelabelType node to denote the implicit coercion. It - * ensures that the resulting expression structurally matches similarly - * processed expressions within the optimizer. + * If the partition key column is not of the same type as the operator + * class and not polymorphic, tell caller to wrap the non-Const expression + * in a RelabelType. This matches what parse_coerce.c does. */ - if (!OidIsValid(operoid)) - { - operoid = get_opfamily_member(key->partopfamily[col], - key->partopcintype[col], - key->partopcintype[col], - strategy); - if (!OidIsValid(operoid)) - elog(ERROR, "missing operator %d(%u,%u) in opfamily %u", - strategy, key->partopcintype[col], key->partopcintype[col], - key->partopfamily[col]); - *need_relabel = true; - } - else - *need_relabel = false; + *need_relabel = (key->parttypid[col] != key->partopcintype[col] && + key->partopcintype[col] != RECORDOID && + !IsPolymorphicType(key->partopcintype[col])); return operoid; } |