[refactor](mv) Refactor MV rewrite StructInfo lookup by relation ids

[foxtail463]

What problem does this PR solve?

Problem Summary:

Nested MV rewrite needs to distinguish two different identities during fuzzy
StructInfo collection:

-- Query starts from base tables and a view over dim_full.
SELECT ...
FROM fact_src t
LEFT JOIN dim_full d0 ...
LEFT JOIN v_dim_full_non_double d1 ...

-- The target MV is defined over child MVs.
SELECT ...
FROM mv_fact t
LEFT JOIN mv_dim_full d0 ...
LEFT JOIN mv_dim_full_view_non_double d1 ...

In this shape, child rewrite can first introduce MV scan relations into memo.
Then the parent group should be able to build a candidate plan from those MV
scan relations and match mv_target.

The old StructInfo candidate path used the table/common-table-id based cache key
in StructInfoMap's candidate map to organize memo candidates. That key only
describes the table family covered by one MV definition; it is a search-space
key, not the identity of a concrete candidate. The exact candidate identity is
relationIdSet, which describes the relations contained by one memo candidate plan
tree.

In the example above, the rewritten scan candidate for mv_dim_full and the
rewritten scan candidate for mv_dim_full_view_non_double can fall into the same
table/common-table-id cache key while representing different relationIdSet
values. If one candidate overwrites or is reused as the other, the parent
mv_target candidate is assembled with the wrong child relation, so the final
target MV rewrite becomes path-sensitive and may fail.

This refactor makes the identity boundary explicit:

• use table ids only to expand the relation search space for an MV
• use exact relationIdSet as the StructInfo candidate identity
• cache candidates by target relation search space, with exact relationIdSet as
  the inner key
• register tableId -> relationId when catalog relations enter memo, including
  nested MV scan relations
• clear StructInfoMap candidate caches when relation identity changes
• keep candidate plan materialization lazy until StructInfo is actually needed

This keeps base-table, view-derived, and rewritten MV-scan candidates coexisting
under the same coarse table family without overwriting each other.

Check List (For Author)

• Test

  • Regression test
  • Unit Test
  • Manual test (add detailed scripts or steps below)
  • No need to test or manual test. Explain why:
    • This is a refactor/code format and no logic has been changed.
    • Previous test can cover this change.
    • No code files have been changed.
    • Other reason

• Behavior changed:

  • No.
  • Yes.

• Does this need documentation?

  • No.
  • Yes.

Check List (For Reviewer who merge this PR)

• Confirm the release note
• Confirm test cases
• Confirm document
• Add branch pick label

[hello-stephen]

Thank you for your contribution to Apache Doris.
Don't know what should be done next? See How to process your PR.

Please clearly describe your PR:

1. What problem was fixed (it's best to include specific error reporting information). How it was fixed.
2. Which behaviors were modified. What was the previous behavior, what is it now, why was it modified, and what possible impacts might there be.
3. What features were added. Why was this function added?
4. Which code was refactored and why was this part of the code refactored?
5. Which functions were optimized and what is the difference before and after the optimization?

[foxtail463]

minimal reproducible case：

DROP DATABASE IF EXISTS mv_id_conflict_demo;
CREATE DATABASE mv_id_conflict_demo;
USE mv_id_conflict_demo;

SET enable_nereids_planner = true;
SET enable_fallback_to_original_planner = false;
SET enable_materialized_view_rewrite = true;
SET enable_materialized_view_nest_rewrite = true;
SET enable_nereids_timeout = false;
SET materialized_view_rewrite_duration_threshold_ms = 1800000;

CREATE TABLE fact_src (
dt DATE NOT NULL,
k VARCHAR(32) NOT NULL,
is_dyn VARCHAR(8),
sku_type VARCHAR(8)
) DUPLICATE KEY(dt, k)
PARTITION BY RANGE(dt) (PARTITION p1 VALUES LESS THAN ('2026-02-05'))
DISTRIBUTED BY HASH(k) BUCKETS 1
PROPERTIES ("replication_allocation" = "tag.location.default: 1");

CREATE TABLE dim_full (
dt DATE NOT NULL,
k VARCHAR(32) NOT NULL,
sku_type VARCHAR(8),
is_dyn VARCHAR(8),
bu VARCHAR(32),
mode_flag VARCHAR(8),
double_flag VARCHAR(8)
) UNIQUE KEY(dt, k, sku_type, is_dyn)
PARTITION BY RANGE(dt) (PARTITION p1 VALUES LESS THAN ('2026-02-05'))
DISTRIBUTED BY HASH(k) BUCKETS 1
PROPERTIES ("replication_allocation" = "tag.location.default: 1");

CREATE VIEW v_dim_full_non_double AS
SELECT dt, k, mode_flag, sku_type FROM dim_full WHERE double_flag = '0';

INSERT INTO fact_src VALUES ('2026-02-04', 'K1', '0', '1');
INSERT INTO dim_full VALUES ('2026-02-04', 'K1', '1', '0', 'D1', 'M', '0');

DROP MATERIALIZED VIEW IF EXISTS mv_fact;
CREATE MATERIALIZED VIEW mv_fact
BUILD IMMEDIATE REFRESH COMPLETE ON MANUAL
PARTITION BY (dt)
DISTRIBUTED BY HASH(k) BUCKETS 1
PROPERTIES ("replication_allocation" = "tag.location.default: 1")
AS
SELECT dt, k, is_dyn, sku_type
FROM fact_src
WHERE sku_type = '1';

DROP MATERIALIZED VIEW IF EXISTS mv_dim_full;
CREATE MATERIALIZED VIEW mv_dim_full
BUILD IMMEDIATE REFRESH COMPLETE ON MANUAL
PARTITION BY (dt)
DISTRIBUTED BY HASH(k) BUCKETS 1
PROPERTIES ("replication_allocation" = "tag.location.default: 1")
AS
SELECT dt, k, bu, is_dyn, sku_type
FROM dim_full;

DROP MATERIALIZED VIEW IF EXISTS mv_dim_full_view_non_double;
CREATE MATERIALIZED VIEW mv_dim_full_view_non_double
BUILD IMMEDIATE REFRESH COMPLETE ON MANUAL
PARTITION BY (dt)
DISTRIBUTED BY HASH(k) BUCKETS 1
PROPERTIES ("replication_allocation" = "tag.location.default: 1")
AS
SELECT dt, k, mode_flag, sku_type
FROM v_dim_full_non_double;

DROP MATERIALIZED VIEW IF EXISTS mv_target;
CREATE MATERIALIZED VIEW mv_target
BUILD IMMEDIATE REFRESH COMPLETE ON MANUAL
PARTITION BY (dt)
DISTRIBUTED BY HASH(k) BUCKETS 1
PROPERTIES ("replication_allocation" = "tag.location.default: 1")
AS
SELECT
t.dt,
t.k,
d0.bu AS out_bu,
d1.mode_flag AS out_mode
FROM mv_fact t
LEFT JOIN mv_dim_full d0
ON t.dt = d0.dt
AND t.k = d0.k
AND t.sku_type = d0.sku_type
AND t.is_dyn = d0.is_dyn
LEFT JOIN mv_dim_full_view_non_double d1
ON t.dt = d1.dt
AND t.k = d1.k
AND t.sku_type = d1.sku_type;

REFRESH MATERIALIZED VIEW mv_fact COMPLETE;
REFRESH MATERIALIZED VIEW mv_dim_full COMPLETE;
REFRESH MATERIALIZED VIEW mv_dim_full_view_non_double COMPLETE;
--等几秒
REFRESH MATERIALIZED VIEW mv_target COMPLETE;

EXPLAIN
SELECT
t.dt,
t.k,
d0.bu AS out_bu,
d1.mode_flag AS out_mode
FROM fact_src t
LEFT JOIN dim_full d0
ON t.dt = d0.dt
AND t.k = d0.k
AND t.sku_type = d0.sku_type
AND t.is_dyn = d0.is_dyn
LEFT JOIN v_dim_full_non_double d1
ON t.dt = d1.dt
AND t.k = d1.k
AND t.sku_type = d1.sku_type
WHERE t.dt = '2026-02-04'
AND t.sku_type = '1'
ORDER BY t.k;

[foxtail463]

/review

[foxtail463]

Performance Evaluation

This benchmark checks whether the new StructInfo candidate lookup adds visible MV rewrite overhead. It compares the current patch 95e91304f729ef6d446973bc0f0d95d923aada51 with doris/master 4e81acee0d83e1db3100a0eb8fe820d05c833c31.

The benchmark SQL models a nested MV rewrite case with base tables, a view, child MVs, and a parent MV. The query starts from base tables and a view, while the target MV is defined over child MVs. This shape exercises the nested MV rewrite path and the StructInfo candidate lookup changed by this PR.

Example:

-- Child MVs.
CREATE MATERIALIZED VIEW mv_fact AS
SELECT dt, k, is_dyn, sku_type
FROM fact_src
WHERE sku_type = '1';
CREATE MATERIALIZED VIEW mv_dim_full AS

SELECT dt, k, bu, is_dyn, sku_type

FROM dim_full;
CREATE MATERIALIZED VIEW mv_dim_full_view_non_double AS

SELECT dt, k, mode_flag, sku_type

FROM v_dim_full_non_double;
-- Parent MV built from child MVs.

CREATE MATERIALIZED VIEW mv_target AS

SELECT

t.dt,

t.k,

d0.bu AS out_bu,

d1.mode_flag AS out_mode

FROM mv_fact t

LEFT JOIN mv_dim_full d0

ON t.dt = d0.dt

AND t.k = d0.k

AND t.sku_type = d0.sku_type

AND t.is_dyn = d0.is_dyn

LEFT JOIN mv_dim_full_view_non_double d1

ON t.dt = d1.dt

AND t.k = d1.k

AND t.sku_type = d1.sku_type;
-- Query starts from the original base table and view.

EXPLAIN

SELECT

t.dt,

t.k,

d0.bu AS out_bu,

d1.mode_flag AS out_mode

FROM fact_src t

LEFT JOIN dim_full d0

ON t.dt = d0.dt

AND t.k = d0.k

AND t.sku_type = d0.sku_type

AND t.is_dyn = d0.is_dyn

LEFT JOIN v_dim_full_non_double d1

ON t.dt = d1.dt

AND t.k = d1.k

AND t.sku_type = d1.sku_type

WHERE t.dt = '2026-02-04'

AND t.sku_type = '1';

The benchmark has three scales:

Scale	Shape
ordinary	2 physical tables (fact_src, dim_full), 1 view, 4 MVs; query and parent MV each contain 3 scan refs and 2 joins; current planChars=2176
large	3 physical tables, 3 views, 5 MVs; includes fact-detail/fact-agg/dim-view/partial-join/target MV paths; current planChars=3829
super	11 physical tables, 6 views, 15 MVs in the generated benchmark schema; combines the minimal conflict, alias, multi-stage, nested, and large-MV shapes; current planChars=5525

All three scales assert that the target MV is chosen.

End-to-end EXPLAIN benchmark

scale	case	baseline steadyAvgMs	current steadyAvgMs	change
ordinary	ordinary_multi_alias_table_join_target_mv_hit	142.256	140.238	-1.4%
large	large_multi_alias_table_join_target_mv_hit	285.378	253.697	-11.1%
super	super_multi_alias_table_join_target_mv_hit	527.796	533.037	+1.0%