github-actions[bot] commented on code in PR #64440:
URL: https://github.com/apache/doris/pull/64440#discussion_r3425516584
##########
fe/fe-core/src/main/java/org/apache/doris/nereids/rules/rewrite/AccessPathExpressionCollector.java:
##########
@@ -375,25 +375,71 @@ public Void visitMapValues(MapValues mapValues,
CollectorContext context) {
}
private static boolean isFunctionNullCheckPath(List<String> suffixPath) {
- return suffixPath.size() == 1 &&
AccessPathInfo.ACCESS_NULL.equals(suffixPath.get(0));
+ return suffixPath.size() == 1 &&
AccessPathInfo.ACCESS_NULL.equals(suffixPath.get(0));
Review Comment:
This return statement lost the method-body indentation. FE checkstyle
enables `Indentation` with `basicOffset=4`, so `mvn validate` / `build.sh --fe`
should fail on this file. Please restore the 8-space indent.
```suggestion
return suffixPath.size() == 1 &&
AccessPathInfo.ACCESS_NULL.equals(suffixPath.get(0));
```
##########
regression-test/suites/nereids_rules_p0/column_pruning/map_contains_arg_pruning.groovy:
##########
@@ -0,0 +1,215 @@
+// Licensed to the Apache Software Foundation (ASF) under one
+// or more contributor license agreements. See the NOTICE file
+// distributed with this work for additional information
+// regarding copyright ownership. The ASF licenses this file
+// to you under the Apache License, Version 2.0 (the
+// "License"); you may not use this file except in compliance
+// with the License. You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing,
+// software distributed under the License is distributed on an
+// "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+// KIND, either express or implied. See the License for the
+// specific language governing permissions and limitations
+// under the License.
+
+// Regression tests for map_contains_key / map_contains_value /
map_contains_entry
+// when the key/value/entry argument references nested sub-columns.
+//
+// Bug: visitMapContainsKey/Value/Entry only visited the map argument and
skipped
+// the key/value/entry argument. When the key is a nested sub-column expression
+// (e.g. element_at(s, 'a')) whose data path was not registered, and the same
+// sub-column also appears in IS NULL, NestedColumnPruning would prune it to
+// null-only metadata access, causing wrong results.
+
+suite("map_contains_arg_pruning") {
+ sql """ DROP TABLE IF EXISTS map_contains_arg_pruning_tbl """
+ sql """
+ CREATE TABLE map_contains_arg_pruning_tbl (
+ id INT,
+ s STRUCT<a: STRING, b: INT> NULL,
+ m MAP<STRING, INT> NULL,
+ v VARIANT NULL
+ ) ENGINE = OLAP
+ DUPLICATE KEY(id)
+ DISTRIBUTED BY HASH(id) BUCKETS 1
+ PROPERTIES ("replication_allocation" = "tag.location.default: 1")
+ """
+ sql """
+ INSERT INTO map_contains_arg_pruning_tbl VALUES
+ (1, named_struct('a', 'hello', 'b', 100), {'hello': 1, 'world':
2}, '{"k": 42}'),
+ (2, named_struct('a', 'doris', 'b', 200), {'doris': 3},
NULL),
+ (3, named_struct('a', null, 'b', 300), NULL,
'{"x": 1}'),
+ (4, NULL, {},
'{}')
+ """
+
+ // ================================================================
+ // Case 1: map_contains_key + element_at IS NULL (original bug)
+ // map_contains_key(m, element_at(s, 'a')) needs full access to s.a
+ // as the key lookup value. Without fix, only [s.a.NULL] from
+ // element_at(s, 'a') IS NULL is registered.
+ // ================================================================
+ explain {
+ sql """
+ SELECT id,
+ element_at(s, 'a') IS NULL,
+ map_contains_key(m, element_at(s, 'a'))
+ FROM map_contains_arg_pruning_tbl ORDER BY id
+ """
+ contains "nested columns"
+ contains "s.a" // s.a should appear in access
paths
+ notContains "s.a.NULL" // should NOT be null-only
+ contains "m.KEYS" // map_contains_key needs KEYS
path
+ }
+
+ order_qt_case1 """
+ SELECT id,
+ element_at(s, 'a') IS NULL,
+ map_contains_key(m, element_at(s, 'a'))
+ FROM map_contains_arg_pruning_tbl ORDER BY id
+ """
+
+ // ================================================================
+ // Case 2: map_contains_value, value arg references sub-column
+ // map_contains_value(m, element_at(s, 'b')) needs s.b as value
+ // ================================================================
+ explain {
+ sql """
+ SELECT id,
+ element_at(s, 'b') IS NULL,
+ map_contains_value(m, element_at(s, 'b'))
+ FROM map_contains_arg_pruning_tbl ORDER BY id
+ """
+ contains "nested columns"
+ contains "s.b" // s.b should appear in access
paths
+ notContains "s.b.NULL" // should NOT be null-only
+ }
+
+ order_qt_case2 """
+ SELECT id,
+ element_at(s, 'b') IS NULL,
+ map_contains_value(m, element_at(s, 'b'))
+ FROM map_contains_arg_pruning_tbl ORDER BY id
+ """
+
+ // ================================================================
+ // Case 3: map_contains_entry(m, key, value) — ternary function.
+ // visitMapContainsEntry must visit BOTH search arguments (arg1=key,
+ // arg2=value) with fresh contexts. Without the fix, only arg1 was
+ // collected and arg2 was silently skipped, leaving s.b unregistered
+ // so that element_at(s, 'b') IS NULL would prune s.b to null-only.
+ // ================================================================
+ explain {
+ sql """
+ SELECT id,
+ element_at(s, 'a') IS NULL,
+ element_at(s, 'b') IS NULL,
+ map_contains_entry(m, element_at(s, 'a'), element_at(s,
'b'))
+ FROM map_contains_arg_pruning_tbl ORDER BY id
+ """
+ contains "nested columns"
+ contains "s.a"
+ contains "s.b"
+ notContains "s.a.NULL"
+ notContains "s.b.NULL"
+ }
+
+ order_qt_case3 """
+ SELECT id,
+ element_at(s, 'a') IS NULL,
+ element_at(s, 'b') IS NULL,
+ map_contains_entry(m, element_at(s, 'a'), element_at(s, 'b'))
+ FROM map_contains_arg_pruning_tbl ORDER BY id
+ """
+
+ // ================================================================
+ // Case 4: map_contains_key(...) IS NULL — the key argument must NOT
+ // be collected as a full-data path because map_contains_key returns
+ // NULL only when the map itself is NULL. The search key does not
+ // affect nullability.
+ //
+ // When map_contains_key(...) IS NULL is the only expression, the
+ // NULL-only context causes the key argument to be skipped entirely,
+ // so s.a is never registered and no nested column pruning is emitted.
+ // ================================================================
+ explain {
+ sql """
+ SELECT id,
+ map_contains_key(m, element_at(s, 'a')) IS NULL
+ FROM map_contains_arg_pruning_tbl ORDER BY id
+ """
+ // s.a must NOT appear in access paths: the key argument inside
+ // map_contains_key(...) IS NULL is skipped in NULL-only context.
+ notContains "[s.a]"
+ }
+
+ order_qt_case4 """
+ SELECT id,
+ map_contains_key(m, element_at(s, 'a')) IS NULL
+ FROM map_contains_arg_pruning_tbl ORDER BY id
+ """
+
+ // ================================================================
+ // Case 5: map_contains_value(...) IS NULL — same behaviour as
+ // map_contains_key: the value argument does not affect nullability.
+ // ================================================================
+ explain {
+ sql """
+ SELECT id,
+ map_contains_value(m, element_at(s, 'b')) IS NULL
+ FROM map_contains_arg_pruning_tbl ORDER BY id
+ """
+ notContains "[s.b]"
+ }
+
+ order_qt_case5 """
+ SELECT id,
+ map_contains_value(m, element_at(s, 'b')) IS NULL
+ FROM map_contains_arg_pruning_tbl ORDER BY id
+ """
+
+ // ================================================================
+ // Case 6: map_contains_entry(...) IS NULL — both key and value
+ // arguments must be skipped. No nested column pruning is expected
+ // because only the map argument is visited with NULL-only path.
+ // ================================================================
+ explain {
+ sql """
+ SELECT id,
+ map_contains_entry(m, element_at(s, 'a'), element_at(s,
'b')) IS NULL
+ FROM map_contains_arg_pruning_tbl ORDER BY id
+ """
+ notContains "[s.a]"
Review Comment:
This assertion is too narrow for the ternary NULL-only regression. If the
skip logic regresses and both search arguments are collected, `PlanNode`
formats the slot's paths as a list such as `all access paths: [s.a, s.b]`; that
string contains neither `[s.a]` nor `[s.b]`, so both checks can still pass.
This is distinct from the existing null-routing implementation thread: the
implementation may be correct, but this new test would not catch the
two-argument failure it is meant to guard. Please match `s.a` / `s.b` directly
or assert that the slot `s` access-path block is absent.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
