[zhang-jianq]

What changes are proposed in this pull request?

Adds the planner-side wiring so Flink's MD5(string) and SHA1(string) offload to the velox flinksql md5/sha1 functions (bigo-sg/velox#87 adds the functions themselves):

• RexCallConverterFactory: map the MD5 and SHA1 RexCalls to md5/sha1 via the default name converter. Both take exactly one VARCHAR argument and return VARCHAR, so a plain name mapping suffices.
• Tests: HashFunctionsTest covers end-to-end execution with 8 symmetric cases (4 per function) — basic / multibyte / NULL / long-input rows, literal inputs incl. the empty string, a nested call over LOWER, and non-string inputs (CAST(n AS STRING) plus the rejected bare MD5(n)/SHA1(n)). The MD5/SHA1 mappings are plain name mappings, so no separate converter unit test is added.

How was this patch tested?

• gluten-flink/ut: HashFunctionsTest (8 cases — basic / multibyte / NULL / long-input, literal incl. empty string, nested LOWER, and non-string CAST/rejected bare int) passes.

• End-to-end dual-run on Flink 1.19.2 (native vs gluten-offloaded, identical VALUES inputs, changelog multiset exact comparison): md5.sql and sha1.sql match native exactly (7 rows each). md5.sql differs only in projecting md5(s) into a md5_s column; the query is:

  CREATE TABLE snk (
    id     BIGINT,
    s      STRING,
    md5_s  STRING
  ) WITH ('connector' = 'print');
  
  INSERT INTO snk
  SELECT id, s, MD5(s)
  FROM (
    VALUES
    (1, CAST('test' AS STRING)),
    (2, CAST('abc' AS STRING)),
    (3, CAST('' AS STRING)),
    (4, CAST('你好' AS STRING)),
    (5, CAST('ABC' AS STRING)),
    (6, CAST(NULL AS STRING)),
    (7, CAST('aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa' AS STRING))
  ) AS src (id, s);

Related issue: #13416

[zhang-jianq]

@KevinyhZou @lgbo-ustc @ggjh-159 Would you mind reviewing this PR?

[ggjh-159]

Thanks for the minimal mapping and the end-to-end coverage.

1. The repo's spotless check (bound to the build's validate phase) FAILED — HashFunctionsTest.java is missing the trailing newline at EOF. Please run mvn spotless:apply (or add the final newline).
2. Please add the SQL details of the dual-run to the PR description — the queries in md5.sql/sha1.sql that were run against Flink 1.19.2.

[zhang-jianq]

@ggjh-159 Thanks for the review. I have fixed both:
• added the trailing newline to HashFunctionsTest.java (via mvn spotless:apply).
• added the dual-run queries to the PR description — the INSERT ... SELECT MD5(s) / SHA1(s)over the 7-row VALUES covering basic / empty / multibyte / case-variant / NULL / long inputs.

[ggjh-159]

One comment on test coverage — details in the line comment on HashFunctionsTest.java.

[ggjh-159]

Please extend HashFunctionsTest with: empty-string and long-input rows for both MD5 and SHA1 (as in the dual-run set), a constant-input case (e.g. MD5('abc')), a nested-call case (e.g. MD5(UPPER(s))), and non-string input cases — MD5(CAST(n AS STRING)) over an int column plus the rejected bare MD5(n) asserting the validation error.

[zhang-jianq]

Thanks for the review. Extended HashFunctionsTest to 8 symmetric cases (4 per function):

• testMd5Basic / testSha1Basic — plain / multibyte / NULL plus a 32-char long input.
• testMd5Constant / testSha1Constant — literal inputs, including the empty string (MD5('') / SHA1('')). The empty string is covered via a literal because the values connector's from_elements drops an empty-string row field, so an empty-string row can't be exercised through a table.
• testMd5Nested / testSha1Nested — MD5(LOWER(s)) / SHA1(LOWER(s)). I used LOWER rather than UPPER because UPPER is not registered in gluten yet.
• testMd5NonStringInput / testSha1NonStringInput — MD5(CAST(n AS STRING)) / SHA1(CAST(n AS STRING)) over an int column, plus assertThatThrownBy on the bare MD5(n) / SHA1(n) asserting the Flink ValidationException.

[ggjh-159]

@zhang-jianq please update velox4j version and Flink.md

[zhang-jianq]

@zhang-jianq please update velox4j version and Flink.md

@ggjh-159 The velox4j-side PR (#68) is merged — flink.yml and Flink.md now pin bigo-sg/velox4j gluten-20260829 at 69485bb. Please take another look.

[ggjh-159]

Thanks for expanding the tests to the constant, nested, and non-string forms.

The PR description's "What changes" section still lists two changes that are no longer in the diff: "CI routing: .github/workflows/flink.yml clones the pinned velox4j branch" and "Docs: gluten-flink/docs/Flink.md build notes synced" — after the rebase onto main, the velox4j ref is managed by main's central gluten-flink/velox4j-ref.txt and the diff no longer touches either file. Please remove these two bullets.

[?] (no anchor) 一般——body——The PR description's "What changes" section still lists two changes that are no longer in the diff: "CI routing: .github/workflows/flink.yml clones the pinned velox4j branch" and "Docs: gluten-flink/docs/Flink.md build notes synced" — after the rebase onto main, the velox4j ref is managed by main's central gluten-flink/velox4j-ref.txt and the diff no longer touches either file. Please remove these two bullets.

[zhang-jianq]

Thanks for expanding the tests to the constant, nested, and non-string forms.

The PR description's "What changes" section still lists two changes that are no longer in the diff: "CI routing: .github/workflows/flink.yml clones the pinned velox4j branch" and "Docs: gluten-flink/docs/Flink.md build notes synced" — after the rebase onto main, the velox4j ref is managed by main's central gluten-flink/velox4j-ref.txt and the diff no longer touches either file. Please remove these two bullets.

[?] (no anchor) 一般——body——The PR description's "What changes" section still lists two changes that are no longer in the diff: "CI routing: .github/workflows/flink.yml clones the pinned velox4j branch" and "Docs: gluten-flink/docs/Flink.md build notes synced" — after the rebase onto main, the velox4j ref is managed by main's central gluten-flink/velox4j-ref.txt and the diff no longer touches either file. Please remove these two bullets.

@ggjh-159 Thanks for your review. It's fixed now. please review it again.

[ggjh-159]

@KevinyhZou @lgbo-ustc @philo-he cc

[ggjh-159]

LGTM