GH-51215: [C++] Migrate remaining datetime consumers to the chrono shim by rok · Pull Request #51216 · apache/arrow

rok · 2026-09-07T14:48:46Z

Rationale for this change

Follow up on #48593 / #48601 - migrate remaining uses of arrow_vendored::date to arrow::internal::chrono, using std::chrono where supported.

What changes are included in this PR?

Route core calendar, parsing, formatting, and Gandiva callers through arrow::internal::chrono. Preserve backend selection and the vendored fallback, with a shared configuration header guarding the timezone implementation.

Are these changes tested?

By existing tests in CI.

Are there any user-facing changes?

Hopefully no.

AI disclosure: this was in large part generated by AI. However I am familiar with the codebase somewhat and have reviewed proposed changes manually.

GitHub Issue: [C++][Gandiva] Reduce dependency on vendored date.h library #51215

github-actions · 2026-09-07T14:49:12Z

⚠️ GitHub issue #51215 has been automatically assigned in GitHub to PR creator.

rok · 2026-09-07T16:12:22Z

@github-actions crossbow submit verify-rc-source-windows test-build-vcpkg-win

github-actions · 2026-09-07T16:15:02Z

Revision: d22fc96

Submitted crossbow builds: ursacomputing/crossbow @ actions-ef7574f790

Task	Status
test-build-vcpkg-win
verify-rc-source-windows

rok · 2026-09-07T16:19:25Z

After #51211 is merged we can completely move to the arrow::internal::chrono shim, which will make it easier to reason about which library is being used for temporal operations and eventually move off of the vendored date.h.

rok · 2026-09-07T17:38:35Z

@github-actions crossbow submit verify-rc-source-windows test-build-vcpkg-win

github-actions · 2026-09-07T17:40:50Z

Revision: ef7d5fe

Submitted crossbow builds: ursacomputing/crossbow @ actions-41c0a4fab7

Task	Status
test-build-vcpkg-win
verify-rc-source-windows

rok · 2026-09-07T18:58:32Z

@github-actions crossbow submit verify-rc-source-windows test-build-vcpkg-win

github-actions · 2026-09-07T19:00:43Z

Revision: a5d7ef9

Submitted crossbow builds: ursacomputing/crossbow @ actions-1c32742fcc

Task	Status
test-build-vcpkg-win
verify-rc-source-windows

rok · 2026-09-07T22:51:08Z

@github-actions crossbow submit verify-rc-source-windows test-build-vcpkg-win

rok · 2026-09-07T22:53:02Z

@pitrou does approach of this PR make sense? Especialy cpp/src/arrow/util/chrono_config_internal.h. If it does I'll polish it and push it to review.

github-actions · 2026-09-07T22:53:14Z

Revision: fdcaf7e

Submitted crossbow builds: ursacomputing/crossbow @ actions-6ed6db4ecf

Task	Status
test-build-vcpkg-win
verify-rc-source-windows

pitrou · 2026-09-08T08:33:32Z

@pitrou does approach of this PR make sense? Especialy cpp/src/arrow/util/chrono_config_internal.h. If it does I'll polish it and push it to review.

I think it does.

Route datetime consumers in Arrow and Gandiva through arrow::internal::chrono. Prefer std::chrono when the standard library has reliable C++20 timezone support, while retaining the vendored fallback and allowing an explicit backend override. Compile the vendored timezone sources, including the Apple helper, only when the fallback backend is selected.

rok · 2026-09-08T16:54:57Z

@github-actions crossbow submit test-r-fedora-clang test-r-linux-as-cran test-r-alpine-linux-cran test-r-macos-as-cran r-binary-packages verify-rc-source-windows test-build-vcpkg-win

github-actions · 2026-09-08T16:57:20Z

Revision: 09d3c61

Submitted crossbow builds: ursacomputing/crossbow @ actions-51d00219e4

Task	Status
r-binary-packages
test-build-vcpkg-win
test-r-alpine-linux-cran
test-r-fedora-clang
test-r-linux-as-cran
test-r-macos-as-cran
verify-rc-source-windows

github-actions · 2026-09-08T18:50:34Z

Revision: 6dbcec4

Submitted crossbow builds: ursacomputing/crossbow @ actions-33aa9ce534

Task	Status
r-binary-packages
test-build-vcpkg-win
test-r-alpine-linux-cran
test-r-fedora-clang
test-r-linux-as-cran
test-r-macos-as-cran
verify-rc-source-windows

rok · 2026-09-08T21:34:42Z

@github-actions crossbow submit test-debian-experimental-cpp-gcc-16

github-actions · 2026-09-08T21:37:34Z

Revision: 2a05f39

Submitted crossbow builds: ursacomputing/crossbow @ actions-6604537ea7

Task	Status
test-debian-experimental-cpp-gcc-16

lriggs · 2026-09-08T21:50:51Z

The gandiva changes look good.

rok · 2026-09-08T21:55:29Z

The gandiva changes look good.

Thanks for reviewing @lriggs!

rok · 2026-09-09T13:10:22Z

@github-actions crossbow submit test-debian-experimental-cpp-gcc-16

github-actions · 2026-09-09T13:12:59Z

Revision: fe99b79

Submitted crossbow builds: ursacomputing/crossbow @ actions-b091cd8d4e

Task	Status
test-debian-experimental-cpp-gcc-16

rok · 2026-09-09T14:10:00Z

@github-actions crossbow submit test-r-fedora-clang test-r-linux-as-cran test-r-alpine-linux-cran test-r-macos-as-cran r-binary-packages verify-rc-source-windows test-build-vcpkg-win

github-actions · 2026-09-09T14:12:53Z

Revision: fe99b79

Submitted crossbow builds: ursacomputing/crossbow @ actions-9a9adb0e6a

Task	Status
r-binary-packages
test-build-vcpkg-win
test-r-alpine-linux-cran
test-r-fedora-clang
test-r-linux-as-cran
test-r-macos-as-cran
verify-rc-source-windows

rok · 2026-09-09T14:20:32Z

@pitrou this is ready for review. I am pretty confident in parts other than string handling in cpp/src/arrow/util/chrono_internal.h, where I'm not sure we want to keep arrow-specific behaviour or switch over to std::chrono behavior.

Copilot

🔵 Needs a closer look

It changes core datetime formatting/parsing and timezone-backend selection across Arrow and Gandiva, so it needs careful human validation across supported platforms/stdlib combinations.

Pull request overview

This PR continues Arrow C++’s migration away from arrow_vendored::date call sites by routing remaining parsing/formatting/calendar/timezone consumers (including Gandiva) through the arrow::internal::chrono shim, while preserving backend selection (std::chrono where supported, vendored fallback otherwise) behind a shared configuration header.

Changes:

Introduces chrono_config_internal.h and refactors chrono_internal.h to centralize backend selection and align timezone implementation compilation with that choice.
Updates remaining core and Gandiva datetime users to use arrow::internal::chrono types/APIs (calendar types, sys/local time, tz helpers).
Adjusts tests and runtime configuration behavior to reflect “OS/stdlib-managed tzdb” vs Arrow-managed tzdb, and adds CI coverage targeting newer libstdc++ behavior.

File summaries

File	Description
dev/tasks/tasks.yml	Adds a Debian experimental GCC 16 CI task to exercise newer libstdc++ chrono/timezone behavior.
cpp/src/gandiva/to_date_holder.cc	Removes direct vendored datetime include (callers now rely on chrono shim transitively).
cpp/src/gandiva/precompiled/timestamp_arithmetic.cc	Switches `year_month_day` usage to `arrow::internal::chrono` types.
cpp/src/gandiva/precompiled/time.cc	Switches precompiled date parsing helpers to `arrow::internal::chrono` calendar types.
cpp/src/gandiva/precompiled/epoch_time_point.h	Replaces relative vendored include with `arrow/util/chrono_internal.h` and updates calendar/time-of-day utilities accordingly.
cpp/src/gandiva/gdv_function_stubs.cc	Uses chrono shim timezone APIs (`locate_zone`, `zoned_time`, `sys_time`) for UTC/local conversions.
cpp/src/gandiva/cast_time.cc	Switches timezone/date computations from vendored date to chrono shim.
cpp/src/arrow/vendored/datetime.cpp	Avoids compiling a second vendored timezone implementation when std-chrono backend is selected.
cpp/src/arrow/vendored/datetime_ios.mm	Adds an iOS wrapper to conditionally compile vendored timezone implementation depending on backend selection.
cpp/src/arrow/util/value_parsing.h	Migrates date/timestamp parsing helpers from vendored date types to chrono shim types.
cpp/src/arrow/util/meson.build	Installs the new internal chrono headers needed by installed public headers.
cpp/src/arrow/util/logger_test.cc	Changes test logging macro expansion to avoid Windows `ERROR` macro collisions.
cpp/src/arrow/util/formatting.h	Migrates formatting helpers from vendored date to chrono shim types and adjusts comments accordingly.
cpp/src/arrow/util/CMakeLists.txt	Explicitly installs internal chrono headers as dependencies of installed headers.
cpp/src/arrow/util/chrono_internal.h	Refactors chrono shim: backend selection moved out, and std-backend formatting now preserves Arrow strftime compatibility (incl. literal braces/unsupported directives).
cpp/src/arrow/util/chrono_config_internal.h	New shared compile-time backend selection header (includes libstdc++ version gating for known tz bugs).
cpp/src/arrow/testing/util.cc	Skips vendored tzdb initialization when runtime reports OS/stdlib-managed tzdb.
cpp/src/arrow/public_api_test.cc	Extends timezone configuration tests for OS/stdlib tzdb builds and validates rejection of runtime tzdb path.
cpp/src/arrow/pretty_print.cc	Removes unused vendored datetime include.
cpp/src/arrow/config.h	Clarifies `RuntimeInfo::using_os_timezone_db` semantics to include stdlib-managed tzdb.
cpp/src/arrow/config.cc	Routes runtime tzdb behavior through chrono shim configuration macros and functions.
cpp/src/arrow/compute/kernels/scalar_temporal_test.cc	Adds strftime syntax compatibility tests (braces, unsupported directives, %Q/%q behavior).
cpp/src/arrow/compute/kernels/scalar_cast_temporal.cc	Migrates Date32 duration type from vendored `days` to chrono shim `days`.
cpp/src/arrow/CMakeLists.txt	Ensures iOS vendored timezone compilation is wrapped to respect std-chrono backend selection.
cpp/src/arrow/array/diff.cc	Migrates date/time formatting in array diff output to chrono shim formatting.
cpp/src/arrow/acero/test_util_internal.cc	Makes `dummy_schema()` static (small test utility refactor).

Review details

Files reviewed: 26/26 changed files
Comments generated: 0
Review effort level: Lite

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

Copilot

🔵 Needs a closer look

Cross-platform timezone backend selection and formatting compatibility warrant final human review despite the targeted coverage.

Review details

Files reviewed: 26/26 changed files
Comments generated: 0 new
Review effort level: Balanced

rok · 2026-09-09T15:12:43Z

+  test-debian-experimental-cpp-gcc-16:
+    ci: github
+    template: docker-tests/github.linux.yml
+    params:
+      env:
+        ARCH: "amd64"
+        DEBIAN: "experimental"
+        GCC: "16"
+        LLVM: "22"
+      image: debian-cpp


Note: I'm adding a newer debian image here to get GCC with newer std::chrono so that the std backend gets test tested on linux as well. I'm ok if we exclude this form the final change.

pitrou · 2026-09-09T15:14:16Z

+// https://gcc.gnu.org/bugzilla/show_bug.cgi?id=116110 (fully fixed in GCC 16.2).
+// Check library macros, not __GNUC__, so Clang using libstdc++ agrees with GCC.
+// The datestamp distinguishes 16.2 (2026-08-07) from 16.1 and early snapshots.
+// Keep the existing Windows backend selection unchanged.


"Unchanged" from what?

pitrou · 2026-09-09T15:14:33Z

+// Share backend selection with the vendored implementation without including its
+// headers. datetime.h undefines macros needed when compiling the implementation.


What is "the implementation"? Which one?

Did you generate this using AI?

pitrou · 2026-09-09T15:15:08Z

+// Use the standard backend by default. Builds may explicitly define
+// ARROW_USE_STD_CHRONO to 0 or 1 when they need to select a backend.


Should this be moved up?

pitrou · 2026-09-09T15:15:24Z

+// Use the standard backend by default. Builds may explicitly define
+// ARROW_USE_STD_CHRONO to 0 or 1 when they need to select a backend.
+//
+// Automatically disable the default for libraries without the C++20 timezone APIs.


"Disable the default" is a bit cryptic. Please be explciit.

pitrou · 2026-09-09T15:15:47Z

+// Automatically disable the default for libraries without the C++20 timezone APIs.
+// On non-Windows, older libstdc++ versions also need the fallback because of
+// https://gcc.gnu.org/bugzilla/show_bug.cgi?id=116110 (fully fixed in GCC 16.2).
+// Check library macros, not __GNUC__, so Clang using libstdc++ agrees with GCC.


This sentence is of little value. Obviously we use libstdc++ symbols when checking for libstdc++...

pitrou · 2026-09-09T15:29:46Z

+// Format a duration or time point using strftime-like format specifiers.
+// Converts "%H%M" style to C++20's "{:L%H%M}" style and uses std::vformat.
+template <typename Temporal>
+std::string format(const char* fmt, const Temporal& value) {


Does this implicitly call the to_stream above? If not, why doesn't it do the same `ToChronoFormat call?

pitrou · 2026-09-09T15:30:36Z

+# The automatic rule excludes internal headers, but these are dependencies of
+# the installed formatting.h and value_parsing.h headers.


Then why make them internal?

pitrou · 2026-09-09T15:32:28Z

+// Emit log via the default logger. Token-paste here to prevent Windows' ERROR
+// macro from expanding before the logger macro is selected.


Why did this suddenly start to occur?

pitrou · 2026-09-09T15:33:41Z

+// Standard-library builds must not compile a second timezone implementation.
+#if !ARROW_USE_STD_CHRONO
+#  include "datetime/visibility.h"
+


Remove empty line?

pitrou · 2026-09-09T15:34:40Z

+// Evaluate automatic backend selection only when the build has not selected one.
+#ifndef ARROW_USE_STD_CHRONO


Why the guard? Just include the file and it won't do anything if ARROW_USE_STD_CHRONO is defined, right?

github-actions Bot added Component: C++ Component: Gandiva awaiting committer review Awaiting committer review labels Sep 7, 2026

rok force-pushed the gh-51215-chrono-shim branch 2 times, most recently from 3e23a57 to d22fc96 Compare September 7, 2026 15:50

rok force-pushed the gh-51215-chrono-shim branch from ef7d5fe to a5d7ef9 Compare September 7, 2026 18:58

github-actions Bot added Component: Python Component: Documentation labels Sep 7, 2026

rok force-pushed the gh-51215-chrono-shim branch 3 times, most recently from b69f711 to fdcaf7e Compare September 7, 2026 22:48

rok force-pushed the gh-51215-chrono-shim branch 2 times, most recently from ba0f5f3 to 27b9bd4 Compare September 8, 2026 15:48

rok force-pushed the gh-51215-chrono-shim branch 2 times, most recently from dd9b10c to 09d3c61 Compare September 8, 2026 16:52

apacheGH-51215: [CI] Add GCC 16 experimental C++ task

2a05f39

apacheGH-51215: [C++] Fix GCC 16 build

fe99b79

rok marked this pull request as ready for review September 9, 2026 14:20

rok requested a review from pitrou as a code owner September 9, 2026 14:20

Copilot AI lite review requested due to automatic review settings September 9, 2026 14:20

rok requested review from assignUser, jonkeane, kou, raulcd and zanmato1984 as code owners September 9, 2026 14:20

Copilot started reviewing on behalf of rok September 9, 2026 14:24 View session

Copilot AI reviewed Sep 9, 2026

View reviewed changes

rok requested a balanced review from Copilot September 9, 2026 14:39

Copilot started reviewing on behalf of rok September 9, 2026 14:40 View session

Copilot AI reviewed Sep 9, 2026

View reviewed changes

rok commented Sep 9, 2026

View reviewed changes

github-actions Bot added awaiting changes Awaiting changes and removed awaiting committer review Awaiting committer review labels Sep 9, 2026

rok mentioned this pull request Sep 9, 2026

[C++] Remove the vendored datetime fallback when toolchain support permits #51267

Open

pitrou reviewed Sep 9, 2026

View reviewed changes

		// Share backend selection with the vendored implementation without including its
		// headers. datetime.h undefines macros needed when compiling the implementation.

		// Use the standard backend by default. Builds may explicitly define
		// ARROW_USE_STD_CHRONO to 0 or 1 when they need to select a backend.

		# The automatic rule excludes internal headers, but these are dependencies of
		# the installed formatting.h and value_parsing.h headers.

		// Emit log via the default logger. Token-paste here to prevent Windows' ERROR
		// macro from expanding before the logger macro is selected.

		// Evaluate automatic backend selection only when the build has not selected one.
		#ifndef ARROW_USE_STD_CHRONO

Uh oh!

Conversation

rok commented Sep 7, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Rationale for this change

What changes are included in this PR?

Are these changes tested?

Are there any user-facing changes?

Uh oh!

github-actions Bot commented Sep 7, 2026

Uh oh!

rok commented Sep 7, 2026

Uh oh!

github-actions Bot commented Sep 7, 2026

Uh oh!

rok commented Sep 7, 2026

Uh oh!

rok commented Sep 7, 2026

Uh oh!

github-actions Bot commented Sep 7, 2026

Uh oh!

rok commented Sep 7, 2026

Uh oh!

github-actions Bot commented Sep 7, 2026

Uh oh!

rok commented Sep 7, 2026

Uh oh!

rok commented Sep 7, 2026

Uh oh!

github-actions Bot commented Sep 7, 2026

Uh oh!

pitrou commented Sep 8, 2026

Uh oh!

rok commented Sep 8, 2026

Uh oh!

github-actions Bot commented Sep 8, 2026

Uh oh!

github-actions Bot commented Sep 8, 2026

Uh oh!

rok commented Sep 8, 2026

Uh oh!

github-actions Bot commented Sep 8, 2026

Uh oh!

lriggs commented Sep 8, 2026

Uh oh!

rok commented Sep 8, 2026

Uh oh!

rok commented Sep 9, 2026

Uh oh!

github-actions Bot commented Sep 9, 2026

Uh oh!

rok commented Sep 9, 2026

Uh oh!

github-actions Bot commented Sep 9, 2026

Uh oh!

rok commented Sep 9, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

🔵 Needs a closer look

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

🔵 Needs a closer look

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

rok commented Sep 7, 2026 •

edited

Loading