GH-48593: [C++] C++20: use standard calendar / timezone APIs

.github/workflows/cpp.yml

@@ -367,8 +367,13 @@ jobs:
           export CMAKE_BUILD_PARALLEL_LEVEL=$NUMBER_OF_PROCESSORS
           ci/scripts/cpp_build.sh "$(pwd)" "$(pwd)/build"
       - name: Download Timezone Database
+        if: matrix.msystem_upper == 'CLANG64'
         shell: bash
-        run: ci/scripts/download_tz_database.sh
+        run: |
+          # TODO(GH-48743): msys2 clang64 uses libc++ and vendored date.h library
+          # which needs tzdata database to build Arrow with time zone support.
+          # https://github.com/apache/arrow/issues/48743
+          ci/scripts/download_tz_database.sh
       - name: Download MinIO
         shell: msys2 {0}
         run: |

.github/workflows/cpp_extra.yml

@@ -365,9 +365,6 @@ jobs:
         with:
           fetch-depth: 0
           submodules: recursive
-      - name: Download Timezone Database
-        shell: bash
-        run: ci/scripts/download_tz_database.sh
       - name: Install cmake
         shell: bash
         run: |

.github/workflows/cpp_windows.yml

@@ -81,9 +81,6 @@ jobs:
         with:
           fetch-depth: 0
           submodules: recursive
-      - name: Download Timezone Database
-        shell: bash
-        run: ci/scripts/download_tz_database.sh
       - name: Install msys2 (for tzdata for ORC tests)
         uses: msys2/setup-msys2@v2
         id: setup-msys2

.github/workflows/matlab.yml

@@ -147,9 +147,6 @@ jobs:
         uses: matlab-actions/setup-matlab@v2
         with:
           release: R2025b
-      - name: Download Timezone Database
-        shell: bash
-        run: ci/scripts/download_tz_database.sh
       - name: Install ccache
         shell: bash
         run: ci/scripts/install_ccache.sh 4.6.3 /usr

cpp/src/arrow/compute/function_test.cc

@@ -95,11 +95,9 @@ TEST(FunctionOptions, Equality) {
   options.emplace_back(new StrptimeOptions("%Y", TimeUnit::type::MILLI, true));
   options.emplace_back(new StrptimeOptions("%Y", TimeUnit::type::NANO));
   options.emplace_back(new StrftimeOptions("%Y-%m-%dT%H:%M:%SZ", "C"));
-#ifndef _WIN32
   options.emplace_back(new AssumeTimezoneOptions(
       "Europe/Amsterdam", AssumeTimezoneOptions::Ambiguous::AMBIGUOUS_RAISE,
       AssumeTimezoneOptions::Nonexistent::NONEXISTENT_RAISE));
-#endif
   options.emplace_back(new PadOptions(5, " "));
   options.emplace_back(new PadOptions(10, "A"));
   options.emplace_back(new PadOptions(10, "A", false));

cpp/src/arrow/compute/kernels/scalar_cast_test.cc

@@ -2358,15 +2358,7 @@ constexpr char kTimestampSecondsJson[] =
 constexpr char kTimestampExtremeJson[] =
     R"(["1677-09-20T00:00:59.123456", "2262-04-13T23:23:23.999999"])";
 
-class CastTimezone : public ::testing::Test {
- protected:
-  void SetUp() override {
-#ifdef _WIN32
-    // Initialize timezone database on Windows
-    ASSERT_OK(InitTestTimezoneDatabase());
-#endif
-  }
-};
+class CastTimezone : public ::testing::Test {};
 
 TEST(Cast, TimestampToDate) {
   // See scalar_temporal_test.cc
@@ -2595,6 +2587,11 @@ TEST(Cast, TimestampToTime) {
 }
 
 TEST_F(CastTimezone, ZonedTimestampToTime) {
+  // TODO(GH-48743): GCC libstdc++ has a bug with DST transitions
+  // https://gcc.gnu.org/bugzilla/show_bug.cgi?id=116110
+#if defined(_WIN32) && defined(__GNUC__) && !defined(__clang__)
+  GTEST_SKIP() << "Test triggers GCC libstdc++ bug (GH-48743).";
+#endif
   CheckCast(ArrayFromJSON(timestamp(TimeUnit::NANO, "Pacific/Marquesas"), kTimestampJson),
             ArrayFromJSON(time64(TimeUnit::NANO), R"([
           52259123456789, 50003999999999, 56480001001001, 65000000000000,

cpp/src/arrow/compute/kernels/scalar_temporal_binary.cc

@@ -27,7 +27,6 @@
 #include "arrow/util/checked_cast.h"
 #include "arrow/util/logging_internal.h"
 #include "arrow/util/time.h"
-#include "arrow/vendored/datetime.h"
 
 namespace arrow {
 
@@ -37,28 +36,30 @@ using internal::checked_pointer_cast;
 namespace compute {
 namespace internal {
 
+namespace chrono = arrow::internal::chrono;
+
 namespace {
 
-using arrow_vendored::date::days;
-using arrow_vendored::date::floor;
-using arrow_vendored::date::hh_mm_ss;
-using arrow_vendored::date::local_days;
-using arrow_vendored::date::local_time;
-using arrow_vendored::date::sys_days;
-using arrow_vendored::date::sys_time;
-using arrow_vendored::date::trunc;
-using arrow_vendored::date::weekday;
-using arrow_vendored::date::weeks;
-using arrow_vendored::date::year_month_day;
-using arrow_vendored::date::year_month_weekday;
-using arrow_vendored::date::years;
-using arrow_vendored::date::literals::dec;
-using arrow_vendored::date::literals::jan;
-using arrow_vendored::date::literals::last;
-using arrow_vendored::date::literals::mon;
-using arrow_vendored::date::literals::sun;
-using arrow_vendored::date::literals::thu;
-using arrow_vendored::date::literals::wed;
+using chrono::days;
+using chrono::dec;
+using chrono::floor;
+using chrono::hh_mm_ss;
+using chrono::jan;
+using chrono::last;
+using chrono::local_days;
+using chrono::local_time;
+using chrono::mon;
+using chrono::sun;
+using chrono::sys_days;
+using chrono::sys_time;
+using chrono::thu;
+using chrono::trunc;
+using chrono::wed;
+using chrono::weekday;
+using chrono::weeks;
+using chrono::year_month_day;
+using chrono::year_month_weekday;
+using chrono::years;
 using internal::applicator::ScalarBinaryNotNullStatefulEqualTypes;
 
 using DayOfWeekState = OptionsWrapper<DayOfWeekOptions>;

cpp/src/arrow/compute/kernels/scalar_temporal_test.cc

@@ -29,6 +29,7 @@
 #include "arrow/type_fwd.h"
 #include "arrow/type_traits.h"
 #include "arrow/util/checked_cast.h"
+#include "arrow/util/chrono_internal.h"  // for ARROW_USE_STD_CHRONO
 #include "arrow/util/formatting.h"
 #include "arrow/util/logging_internal.h"
 
@@ -411,14 +412,6 @@ class ScalarTemporalTest : public ::testing::Test {
   RoundTemporalOptions round_to_15_quarters =
       RoundTemporalOptions(15, CalendarUnit::QUARTER);
   RoundTemporalOptions round_to_15_years = RoundTemporalOptions(15, CalendarUnit::YEAR);
-
- protected:
-  void SetUp() override {
-#ifdef _WIN32
-    // Initialize timezone database on Windows
-    ASSERT_OK(InitTestTimezoneDatabase());
-#endif
-  }
 };
 
 class ScalarTemporalTestStrictCeil : public ScalarTemporalTest {
@@ -716,6 +709,11 @@ TEST_F(ScalarTemporalTest, TestIsLeapYear) {
 }
 
 TEST_F(ScalarTemporalTest, TestZoned1) {
+  // TODO(GH-48743): GCC libstdc++ has a bug with DST transitions
+  // https://gcc.gnu.org/bugzilla/show_bug.cgi?id=116110
+#if defined(_WIN32) && defined(__GNUC__) && !defined(__clang__)
+  GTEST_SKIP() << "Test triggers GCC libstdc++ bug (GH-48743).";
+#endif
   std::vector<std::string> timezones = {"Pacific/Marquesas", "-09:30"};
   for (const auto& timezone : timezones) {
     auto unit = timestamp(TimeUnit::NANO, timezone);
@@ -814,6 +812,11 @@ TEST_F(ScalarTemporalTest, TestZoned1) {
 }
 
 TEST_F(ScalarTemporalTest, TestZoned2) {
+  // TODO(GH-48743): GCC libstdc++ has a bug with DST transitions
+  // https://gcc.gnu.org/bugzilla/show_bug.cgi?id=116110
+#if defined(_WIN32) && defined(__GNUC__) && !defined(__clang__)
+  GTEST_SKIP() << "Test triggers GCC libstdc++ bug (GH-48743).";
+#endif
   for (auto u : TimeUnit::values()) {
     auto unit = timestamp(u, "Australia/Broken_Hill");
     auto month = "[1, 3, 1, 5, 1, 12, 12, 12, 1, 1, 1, 1, 12, 12, 12, 1, null]";
@@ -2775,6 +2778,11 @@ TEST_F(ScalarTemporalTestMultipleSinceGreaterUnit, CeilUTC) {
 }
 
 TEST_F(ScalarTemporalTestMultipleSinceGreaterUnit, CeilZoned) {
+  // TODO(GH-48743): GCC libstdc++ has a bug with DST transitions
+  // https://gcc.gnu.org/bugzilla/show_bug.cgi?id=116110
+#if defined(_WIN32) && defined(__GNUC__) && !defined(__clang__)
+  GTEST_SKIP() << "Test triggers GCC libstdc++ bug (GH-48743).";
+#endif
   std::string op = "ceil_temporal";
 
   // Data for tests below was generated via lubridate with the exception
@@ -3165,6 +3173,11 @@ TEST_F(ScalarTemporalTestMultipleSinceGreaterUnit, FloorUTC) {
 }
 
 TEST_F(ScalarTemporalTestMultipleSinceGreaterUnit, FloorZoned) {
+  // TODO(GH-48743): GCC libstdc++ has a bug with DST transitions
+  // https://gcc.gnu.org/bugzilla/show_bug.cgi?id=116110
+#if defined(_WIN32) && defined(__GNUC__) && !defined(__clang__)
+  GTEST_SKIP() << "Test triggers GCC libstdc++ bug (GH-48743).";
+#endif
   std::string op = "floor_temporal";
 
   // Data for tests below was generated via lubridate with the exception
@@ -3598,6 +3611,11 @@ TEST_F(ScalarTemporalTestMultipleSinceGreaterUnit, RoundUTC) {
 }
 
 TEST_F(ScalarTemporalTestMultipleSinceGreaterUnit, RoundZoned) {
+  // TODO(GH-48743): GCC libstdc++ has a bug with DST transitions
+  // https://gcc.gnu.org/bugzilla/show_bug.cgi?id=116110
+#if defined(_WIN32) && defined(__GNUC__) && !defined(__clang__)
+  GTEST_SKIP() << "Test triggers GCC libstdc++ bug (GH-48743).";
+#endif
   std::string op = "round_temporal";
 
   // Data for tests below was generated via lubridate with the exception

cpp/src/arrow/compute/kernels/scalar_temporal_unary.cc

@@ -29,7 +29,6 @@
 #include "arrow/util/logging_internal.h"
 #include "arrow/util/time.h"
 #include "arrow/util/value_parsing.h"
-#include "arrow/vendored/datetime.h"
 
 namespace arrow {
 
@@ -38,34 +37,36 @@ using internal::checked_pointer_cast;
 
 namespace compute::internal {
 
+namespace chrono = arrow::internal::chrono;
+
 namespace {
 
-using arrow_vendored::date::ceil;
-using arrow_vendored::date::days;
-using arrow_vendored::date::floor;
-using arrow_vendored::date::hh_mm_ss;
-using arrow_vendored::date::local_days;
-using arrow_vendored::date::local_time;
-using arrow_vendored::date::locate_zone;
-using arrow_vendored::date::Monday;
-using arrow_vendored::date::months;
-using arrow_vendored::date::round;
-using arrow_vendored::date::Sunday;
-using arrow_vendored::date::sys_time;
-using arrow_vendored::date::trunc;
-using arrow_vendored::date::weekday;
-using arrow_vendored::date::weeks;
-using arrow_vendored::date::year;
-using arrow_vendored::date::year_month_day;
-using arrow_vendored::date::year_month_weekday;
-using arrow_vendored::date::years;
-using arrow_vendored::date::literals::dec;
-using arrow_vendored::date::literals::jan;
-using arrow_vendored::date::literals::last;
-using arrow_vendored::date::literals::mon;
-using arrow_vendored::date::literals::sun;
-using arrow_vendored::date::literals::thu;
-using arrow_vendored::date::literals::wed;
+using chrono::ceil;
+using chrono::days;
+using chrono::dec;
+using chrono::floor;
+using chrono::hh_mm_ss;
+using chrono::jan;
+using chrono::last;
+using chrono::local_days;
+using chrono::local_time;
+using chrono::locate_zone;
+using chrono::mon;
+using chrono::Monday;
+using chrono::months;
+using chrono::round;
+using chrono::sun;
+using chrono::Sunday;
+using chrono::sys_time;
+using chrono::thu;
+using chrono::trunc;
+using chrono::wed;
+using chrono::weekday;
+using chrono::weeks;
+using chrono::year;
+using chrono::year_month_day;
+using chrono::year_month_weekday;
+using chrono::years;
 using std::chrono::duration_cast;
 using std::chrono::hours;
 using std::chrono::minutes;
@@ -525,8 +526,8 @@ struct Week {
   }
 
   Localizer localizer_;
-  arrow_vendored::date::weekday wd_;
-  arrow_vendored::date::days days_offset_;
+  chrono::weekday wd_;
+  chrono::days days_offset_;
   const bool count_from_zero_;
   const bool first_week_is_fully_in_year_;
 };
@@ -1379,35 +1380,32 @@ struct AssumeTimezone {
   T Call(KernelContext*, Arg0 arg, Status* st) const {
     try {
       return get_local_time<T, Arg0>(arg, &tz_);
-    } catch (const arrow_vendored::date::nonexistent_local_time& e) {
+    } catch (const chrono::nonexistent_local_time& e) {
       switch (options.nonexistent) {
         case AssumeTimezoneOptions::Nonexistent::NONEXISTENT_RAISE: {
           *st = Status::Invalid("Timestamp doesn't exist in timezone '", options.timezone,
                                 "': ", e.what());
           return arg;
         }
         case AssumeTimezoneOptions::Nonexistent::NONEXISTENT_EARLIEST: {
-          return get_local_time<T, Arg0>(arg, arrow_vendored::date::choose::latest,
-                                         &tz_) -
-                 1;
+          return get_local_time<T, Arg0>(arg, chrono::choose::latest, &tz_) - 1;
         }
         case AssumeTimezoneOptions::Nonexistent::NONEXISTENT_LATEST: {
-          return get_local_time<T, Arg0>(arg, arrow_vendored::date::choose::latest, &tz_);
+          return get_local_time<T, Arg0>(arg, chrono::choose::latest, &tz_);
         }
       }
-    } catch (const arrow_vendored::date::ambiguous_local_time& e) {
+    } catch (const chrono::ambiguous_local_time& e) {
       switch (options.ambiguous) {
         case AssumeTimezoneOptions::Ambiguous::AMBIGUOUS_RAISE: {
           *st = Status::Invalid("Timestamp is ambiguous in timezone '", options.timezone,
                                 "': ", e.what());
           return arg;
         }
         case AssumeTimezoneOptions::Ambiguous::AMBIGUOUS_EARLIEST: {
-          return get_local_time<T, Arg0>(arg, arrow_vendored::date::choose::earliest,
-                                         &tz_);
+          return get_local_time<T, Arg0>(arg, chrono::choose::earliest, &tz_);
         }
         case AssumeTimezoneOptions::Ambiguous::AMBIGUOUS_LATEST: {
-          return get_local_time<T, Arg0>(arg, arrow_vendored::date::choose::latest, &tz_);
+          return get_local_time<T, Arg0>(arg, chrono::choose::latest, &tz_);
         }
       }
     }