feat: add date_trunc function for time-bucketing rollups by VioletQwQ-0 · Pull Request #24627 · matrixorigin/matrixone

VioletQwQ-0 · 2026-05-27T07:42:57Z

What type of PR is this?

Which issue(s) this PR fixes:

What this PR does / why we need it:

Add date_trunc(unit, temporal_value) to truncate a DATE, DATETIME, or TIMESTAMP value to a specified precision for time-series rollups and bucketing.

Supported units:

year - truncate to January 1st
quarter - truncate to first day of the quarter
month - truncate to first day of the month
week - truncate to Monday of the ISO week
day - truncate to midnight
hour - truncate to start of the hour
minute - truncate to start of the minute
second - truncate to the second

Supported input types:

first argument: string unit (VARCHAR/string literal)
second argument: DATE, DATETIME, or TIMESTAMP

Intentional non-goals / boundaries:

String timestamp values are not accepted as the second argument; callers should use an explicit CAST(... AS DATE/DATETIME/TIMESTAMP).
DATE inputs reject sub-day units (hour, minute, second) because the result type remains DATE.
TIMESTAMP inputs preserve TIMESTAMP result semantics and use the session time zone for truncation.
Week truncation uses ISO week semantics (Monday as first day of week).

Validation:

CGO_CFLAGS="-I/Users/violet/code/matrixone/cgo -I/Users/violet/code/matrixone/thirdparties/install/include" CGO_LDFLAGS="-L/Users/violet/code/matrixone/cgo -lmo -L/Users/violet/code/matrixone/thirdparties/install/lib" go test ./pkg/sql/plan/function -run 'TestDateTrunc|Test_GetFunctionByName' -count=1
git diff --check -- test/distributed/cases/function/date_trunc.sql test/distributed/cases/function/date_trunc.result

qodo-code-review · 2026-05-27T07:43:01Z

Qodo reviews are paused for this user.

Troubleshooting steps vary by plan Learn more →

On a Teams plan?
Reviews resume once this user has a paid seat and their Git account is linked in Qodo.
Link Git account →

Using GitHub Enterprise Server, GitLab Self-Managed, or Bitbucket Data Center?
These require an Enterprise plan - Contact us
Contact us →

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

qodo-code-review · 2026-06-03T07:16:25Z

Qodo reviews are paused for this user.

Troubleshooting steps vary by plan Learn more →

On a Teams plan?
Reviews resume once this user has a paid seat and their Git account is linked in Qodo.
Link Git account →

Using GitHub Enterprise Server, GitLab Self-Managed, or Bitbucket Data Center?
These require an Enterprise plan - Contact us
Contact us →

XuPeng-SH

I’m requesting changes because this introduces observable type/semantic regressions in date_trunc.

The date_trunc(varchar, timestamp) overload is registered to return DATETIME, and the implementation appends types.Datetime after converting the timestamp into session-local wall-clock time. That drops TIMESTAMP semantics for an absolute instant. Existing timestamp-preserving temporal operators in this package return TIMESTAMP for timestamp inputs (for example date_add/date_sub on timestamp), so this new overload changes downstream comparison/cast behavior in a surprising way. Please keep the overload return type as TIMESTAMP and convert the truncated local datetime back to timestamp before appending.
The date_trunc(varchar, date) overload accepts sub-day units such as hour, minute, and second, but it always converts the truncated value back to DATE. Those units therefore collapse to the original date / day truncation and become effectively misleading no-ops for DATE inputs. Please either reject sub-day units for DATE inputs or change the overload/result type so those units are representable.

qodo-code-review · 2026-06-04T02:24:18Z

Qodo reviews are paused for this user.

Troubleshooting steps vary by plan Learn more →

On a Teams plan?
Reviews resume once this user has a paid seat and their Git account is linked in Qodo.
Link Git account →

Using GitHub Enterprise Server, GitLab Self-Managed, or Bitbucket Data Center?
These require an Enterprise plan - Contact us
Contact us →

…day units on DATE - date_trunc(varchar, timestamp) now returns TIMESTAMP instead of DATETIME, matching the behavior of date_add/date_sub which preserve timestamp semantics. - date_trunc(varchar, date) now rejects hour/minute/second units since DATE has no time component, making the operation a misleading no-op. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

qodo-code-review · 2026-06-05T08:06:06Z

Qodo reviews are paused for this user.

Troubleshooting steps vary by plan Learn more →

On a Teams plan?
Reviews resume once this user has a paid seat and their Git account is linked in Qodo.
Link Git account →

Using GitHub Enterprise Server, GitLab Self-Managed, or Bitbucket Data Center?
These require an Enterprise plan - Contact us
Contact us →

qodo-code-review · 2026-06-05T10:45:52Z

Qodo reviews are paused for this user.

Troubleshooting steps vary by plan Learn more →

On a Teams plan?
Reviews resume once this user has a paid seat and their Git account is linked in Qodo.
Link Git account →

Using GitHub Enterprise Server, GitLab Self-Managed, or Bitbucket Data Center?
These require an Enterprise plan - Contact us
Contact us →

XuPeng-SH

Blocking correctness issues:

date_trunc(..., timestamp) can return the wrong instant in repeated DST hours.

In pkg/sql/plan/function/func_binary.go, the timestamp path does:

ts.ToDatetime(loc)
truncate the wall-clock Datetime
truncated.ToTimestamp(loc)

That round-trip loses which DST fold the original instant belonged to. In a fallback transition (for example, the second 01:30 in a zone with DST rollback), truncating to hour should yield the second 01:00, but reconstructing only from the local civil fields makes the result ambiguous and can map to the wrong instant.

The explicit (varchar, varchar) overload hijacks ordinary string-literal calls and forces canonical DATETIME(6) text output.

Because date_trunc uses fixedTypeMatch, a call like date_trunc('day', '2024-01-02') matches the exact (varchar, varchar) overload instead of a temporal overload. That implementation always parses as Datetime and returns truncated.String2(6), so string-literal calls are forced into YYYY-MM-DD HH:MM:SS.ffffff text semantics instead of behaving like temporal truncation on typed values.

So there are still real compatibility/correctness issues in both the timestamp semantics and the varchar overload design.

…runc - Use FromClockZone instead of ToTimestamp to correctly preserve the DST fold when converting truncated local datetime back to timestamp. In a fallback transition, the original round-trip through ToTimestamp could map to the wrong instant. - Remove the (varchar, varchar) overload so string-literal calls are routed through the temporal overloads via implicit casting, instead of being forced into Datetime(6) text semantics. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

XuPeng-SH

这版实现我没有挖到硬 correctness blocker，但按高标准看，当前 SQL/BVT 覆盖还不够支撑一个新的用户可见 builtin 直接合入。

最主要的缺口有两类：

TIMESTAMP 的正向 SQL 覆盖太弱
现在 BVT 里只有一条混合查询顺手带到了 date_trunc(hour, ts)，但 DATE_TRUNC 的 TIMESTAMP 路径本身有独立的时区/DST 语义，应该至少在 SQL 层明确锁住一组直接结果：year / quarter / month / week / day / hour / minute / second。现在这部分主要靠 Go unit test，不够。\n\n2. DATE 的正向 SQL 覆盖不完整\n 当前只测了 month 成功和 hour 失败，但像 year / quarter / week / day 这些合法 DATE 路径并没有直接在 BVT 里钉住。对于新函数，这会留下很明显的回归窗口。\n\n另外还有几条很便宜但值回票价的 SQL case：\n- NULL 传播（date_trunc(hour, null) / date_trunc(null, dt)）\n- unit 大小写混用（YEAR, Hour）\n- week 的年边界 case（比如 1 月 1 日附近）\n\n所以我这边的结论是：代码可以，但产品级回归面还没打够。建议把至少 TIMESTAMP + DATE 的正向 SQL coverage 补齐，再继续。

VioletQwQ-0 · 2026-06-08T10:03:24Z

Updated according to the latest CHANGES_REQUESTED.

Changes:

Added direct SQL/BVT coverage for DATE_TRUNC on TIMESTAMP for year / quarter / month / week / day / hour / minute / second.
Added DATE positive coverage for year / quarter / week / day, keeping the existing month and sub-day rejection cases.
Added NULL propagation coverage for timestamp input and NULL unit.
Added mixed/uppercase unit coverage and a week year-boundary case.

Validation:

CGO_CFLAGS="-I/Users/violet/code/matrixone/cgo -I/Users/violet/code/matrixone/thirdparties/install/include" CGO_LDFLAGS="-L/Users/violet/code/matrixone/cgo -lmo -L/Users/violet/code/matrixone/thirdparties/install/lib" go test ./pkg/sql/plan/function -run 'TestDateTrunc|Test_GetFunctionByName' -count=1\n- git diff --check -- test/distributed/cases/function/date_trunc.sql test/distributed/cases/function/date_trunc.result\n- Local temporary MO smoke on the new SQL cases with time_zone = +00:00.

VioletQwQ-0 · 2026-06-08T10:25:15Z

Updated the PR according to the remaining review-risk sweep.

Changes:

Clarified the PR body semantic boundary: the second argument supports DATE/DATETIME/TIMESTAMP only; string timestamp values must be explicitly cast and are an intentional non-goal.
Added BVT coverage for bare NULL second argument: date_trunc('hour', null).
Added BVT coverage for DATE sub-day rejection on minute and second, in addition to the existing hour case.
Fixed the SCA failure reported by the current CI run: removed the ineffectual initial assignment in dateTruncCheck.
Synced the PR branch with latest upstream/main.

Validation:

CGO_CFLAGS="-I/Users/violet/code/matrixone/cgo -I/Users/violet/code/matrixone/thirdparties/install/include" CGO_LDFLAGS="-L/Users/violet/code/matrixone/cgo -lmo -L/Users/violet/code/matrixone/thirdparties/install/lib" go test ./pkg/sql/plan/function -run 'TestDateTrunc|Test_GetFunctionByName' -count=1
git diff --check -- pkg/sql/plan/function/func_binary.go test/distributed/cases/function/date_trunc.sql test/distributed/cases/function/date_trunc.result
ineffassign ./pkg/sql/plan/function

# Conflicts: # pkg/sql/plan/function/function_id.go # pkg/sql/plan/function/function_id_test.go

qodo-code-review · 2026-06-09T06:53:59Z

Qodo reviews are paused for this user.

Troubleshooting steps vary by plan Learn more →

On a Teams plan?
Reviews resume once this user has a paid seat and their Git account is linked in Qodo.
Link Git account →

Using GitHub Enterprise Server, GitLab Self-Managed, or Bitbucket Data Center?
These require an Enterprise plan - Contact us
Contact us →

# Conflicts: # pkg/sql/plan/function/function_id.go # pkg/sql/plan/function/function_id_test.go

qodo-code-review · 2026-06-16T02:45:36Z

Qodo reviews are paused for this user.

Troubleshooting steps vary by plan Learn more →

On a Teams plan?
Reviews resume once this user has a paid seat and their Git account is linked in Qodo.
Link Git account →

Using GitHub Enterprise Server, GitLab Self-Managed, or Bitbucket Data Center?
These require an Enterprise plan - Contact us
Contact us →

qodo-code-review · 2026-06-24T07:01:39Z

Qodo reviews are paused for this user.

Troubleshooting steps vary by plan Learn more →

On a Teams plan?
Reviews resume once this user has a paid seat and their Git account is linked in Qodo.
Link Git account →

Using GitHub Enterprise Server, GitLab Self-Managed, or Bitbucket Data Center?
These require an Enterprise plan - Contact us
Contact us →

VioletQwQ-0 requested review from aunjgr and ouyuanning as code owners May 27, 2026 07:42

VioletQwQ-0 had a problem deploying to ci May 27, 2026 07:43 — with GitHub Actions Error

VioletQwQ-0 temporarily deployed to ci May 27, 2026 07:43 — with GitHub Actions Inactive

VioletQwQ-0 had a problem deploying to ci May 27, 2026 07:43 — with GitHub Actions Error

VioletQwQ-0 had a problem deploying to ci May 27, 2026 07:43 — with GitHub Actions Failure

VioletQwQ-0 had a problem deploying to ci May 27, 2026 07:43 — with GitHub Actions Error

VioletQwQ-0 temporarily deployed to ci May 27, 2026 07:43 — with GitHub Actions Inactive

mergify Bot added the kind/feature label May 27, 2026

matrix-meow added the size/M Denotes a PR that changes [100,499] lines label May 27, 2026

VioletQwQ-0 temporarily deployed to ci May 27, 2026 07:59 — with GitHub Actions Inactive

VioletQwQ-0 had a problem deploying to ci May 27, 2026 07:59 — with GitHub Actions Failure

VioletQwQ-0 temporarily deployed to ci May 27, 2026 07:59 — with GitHub Actions Inactive

VioletQwQ-0 temporarily deployed to ci May 27, 2026 09:09 — with GitHub Actions Inactive

VioletQwQ-0 had a problem deploying to ci May 27, 2026 09:09 — with GitHub Actions Error

VioletQwQ-0 temporarily deployed to ci May 27, 2026 09:09 — with GitHub Actions Inactive

VioletQwQ-0 had a problem deploying to ci May 27, 2026 09:09 — with GitHub Actions Failure

VioletQwQ-0 had a problem deploying to ci May 27, 2026 09:09 — with GitHub Actions Error

VioletQwQ-0 temporarily deployed to ci May 27, 2026 09:09 — with GitHub Actions Inactive

VioletQwQ-0 and others added 3 commits June 2, 2026 15:45

Merge upstream/main into violet/issue-24551-date-trunc

2d10ecc

chore: gofmt function_id.go

ac7ad78

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

Merge branch 'main' into violet/issue-24551-date-trunc

b8dc6d4

XuPeng-SH requested changes Jun 3, 2026

View reviewed changes

Merge branch 'main' into violet/issue-24551-date-trunc

633383e

Merge upstream/main into violet/issue-24551-date-trunc

e2a7bff

XuPeng-SH requested changes Jun 5, 2026

View reviewed changes

VioletQwQ-0 and others added 3 commits June 8, 2026 10:54

fix(plan): harden date_trunc semantics

0b1bb86

test(plan): cover date_trunc invalid arguments

04688b4

XuPeng-SH requested changes Jun 8, 2026

View reviewed changes

test: expand date_trunc bvt coverage

4b4e974

VioletQwQ-0 added 2 commits June 8, 2026 18:23

test: cover date_trunc edge cases

a0ef83a

Merge upstream/main into violet/issue-24551-date-trunc

da0e20f

Merge upstream/main into violet/issue-24551-date-trunc

9f2d5af

# Conflicts: # pkg/sql/plan/function/function_id.go # pkg/sql/plan/function/function_id_test.go

heni02 approved these changes Jun 9, 2026

View reviewed changes

VioletQwQ-0 added 2 commits June 9, 2026 18:18

Merge upstream/main into violet/issue-24551-date-trunc

95301eb

# Conflicts: # pkg/sql/plan/function/function_id.go # pkg/sql/plan/function/function_id_test.go

Merge upstream/main into violet/issue-24551-date-trunc

6078a88

Merge upstream/main into violet/issue-24551-date-trunc

6f9380e

Merge upstream/main into violet/issue-24551-date-trunc

52d9a5c

Uh oh!

Conversation

VioletQwQ-0 commented May 27, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

What type of PR is this?

Which issue(s) this PR fixes:

What this PR does / why we need it:

Uh oh!

qodo-code-review Bot commented May 27, 2026

Qodo reviews are paused for this user.

Uh oh!

qodo-code-review Bot commented Jun 3, 2026

Qodo reviews are paused for this user.

Uh oh!

XuPeng-SH left a comment

Choose a reason for hiding this comment

Uh oh!

qodo-code-review Bot commented Jun 4, 2026

Qodo reviews are paused for this user.

Uh oh!

qodo-code-review Bot commented Jun 5, 2026

Qodo reviews are paused for this user.

Uh oh!

qodo-code-review Bot commented Jun 5, 2026

Qodo reviews are paused for this user.

Uh oh!

XuPeng-SH left a comment

Choose a reason for hiding this comment

Uh oh!

XuPeng-SH left a comment

Choose a reason for hiding this comment

Uh oh!

VioletQwQ-0 commented Jun 8, 2026

Uh oh!

VioletQwQ-0 commented Jun 8, 2026

Uh oh!

qodo-code-review Bot commented Jun 9, 2026

Qodo reviews are paused for this user.

Uh oh!

qodo-code-review Bot commented Jun 16, 2026

Qodo reviews are paused for this user.

Uh oh!

qodo-code-review Bot commented Jun 24, 2026

Qodo reviews are paused for this user.

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

8 participants

VioletQwQ-0 commented May 27, 2026 •

edited

Loading