[eldenmoon]

What problem does this PR solve?

Related PR: #70359

Problem Summary: a heavy schema change that changes a NOT NULL DATETIMEV2 column's scale can be cancelled by BE with a result type mismatch, for example:

CAST required_dt(DateTimeV2(3)) TO DateTimeV2(6) return column type mismatch:
declared=DateTimeV2(6), actual=Nullable(DATETIMEV2)

SchemaChangeJobV2 directly creates a legacy CastExpr without the nullable metadata normally supplied by Nereids translation. The legacy fallback does not recognize DATETIMEV2 scale conversion, while the existing BE converter returns a nullable physical column.

Fix

At the schema-change CAST construction site, compute nullability using the existing Cast.castNullable function, the already resolved source/target types, and the source slot nullability. Store it through the existing setNullableFromNereids metadata, which takes precedence when serializing the task expression.

This changes one production file. It adds no per-type special case and leaves CastExpr.isNullable, normal query translation, and runtime-filter construction unchanged. The skipped string-to-string and DECIMALV2-to-DECIMALV2 paths remain unchanged.

Validation

• Before the fix: the new task-serialization UT fails because the CAST node is non-nullable. The baseline SQL schema change was cancelled with the declared/actual nullable mismatch above, using the same ASAN BE.
• After the fix: 56 FE unit tests passed (CastExprTest, CastTest, TryCastTest, CastExprThriftSimplificationTest, SchemaChangeJobV2Test including three new task cases, and CloudSchemaChangeJobV2Test).
• build.sh --fe passed with zero Checkstyle violations. The unchanged BE was built with ASAN via build.sh --be.
• SQL goldens were generated by run-regression-test.sh, then normal comparison passed: test_schema_change_cast_nullable and existing test_alter_table_column_nullable (2 suites). Both scale directions (3 -> 6 -> 3), NOT NULL and nullable columns, maximum datetime, historical NULL, and microsecond writes are covered. Downscale results match an ordinary CAST before the alter.

Compatibility and performance

BE CAST conversion semantics and storage formats are unchanged. No new Thrift fields, public APIs, or dependencies are introduced. The task declares the nullable result already returned by the converter; required-column constraints remain in effect.

The inference runs while constructing each changed-column expression, reusing types that are already resolved. There is no new per-row work. Other direct CastExpr entry points were inspected; their legacy fallback is preserved to avoid changing runtime-filter behavior. No performance benchmark is claimed.

Release note

Fix heavy schema change failures when modifying DATETIMEV2 scale.

Check List (For Author)

• Test: 56 FE unit tests passed; FE Checkstyle and ASAN BE builds passed; 2 SQL regression suites passed
• Behavior changed: Yes. DATETIMEV2 scale schema changes no longer fail because of an incorrect CAST result nullability declaration; conversion semantics are unchanged.
• Does this need documentation: No

[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?

[eldenmoon]

Local pipeline review — ✅ PASS

schema: doris-repo-review/v1
status: PASS
pr: apache/doris#70362
commit: d9dc16fea95eb5a8fd51e9587ae78e87792f4c4a
base: 66122209cac78805a96a42f44d2a4a20b428f038
reviewed_at: 2026-11-07T13:33+08:00
reviewer: eldenmoon
model: gpt-6-astra
effort: xhigh
findings: {blocker: 0, major: 0, minor: 0, nit: 0}
rounds: 1
converged: true

Notes for maintainers

• Static source review only; no builds or tests were run by the reviewers. Existing author logs were inspected: 56 FE tests, FE build/Checkstyle, generated goldens and two normal schema-change regression suites passed. CI status was not verified.

_{Reviewed locally with the doris-repo-review pipeline. Repository policy may accept this receipt for the matching commit; it is not a human Apache approval.}

[eldenmoon]

/run buildall

[hello-stephen]

FE UT Coverage Report

Increment line coverage 100.00% (2/2) 🎉
Increment coverage report
Complete coverage report

[eldenmoon]

/run buildall

[hello-stephen]

FE UT Coverage Report

Increment line coverage 100.00% (2/2) 🎉
Increment coverage report
Complete coverage report

[eldenmoon]

run cloud_p0

[eldenmoon]

run external

[eldenmoon]

run cloud_p0

[eldenmoon]

run external

[eldenmoon]

run feut

[eldenmoon]

run nonConcurrent

[eldenmoon]

run cloud_p0

[eldenmoon]

run external

[eldenmoon]

run feut

[eldenmoon]

run nonConcurrent

[eldenmoon]

run p0

[eldenmoon]

run cloud_p0

[eldenmoon]

run external

[eldenmoon]

run feut

[eldenmoon]

run nonConcurrent

[eldenmoon]

run p0

[eldenmoon]

run buildall

[eldenmoon]

run external

[eldenmoon]

run p0

[eldenmoon]

run external

[eldenmoon]

run nonConcurrent

[eldenmoon]

run external

[eldenmoon]

run nonConcurrent

[eldenmoon]

run p0

[eldenmoon]

run compile

[eldenmoon]

run feut

[eldenmoon]

run external

[eldenmoon]

run nonConcurrent

[eldenmoon]

run external

[eldenmoon]

run nonConcurrent

[eldenmoon]

run external

[eldenmoon]

run external

[eldenmoon]

run external

[eldenmoon]

run external

[eldenmoon]

run external

[eldenmoon]

run nonConcurrent