Merge pull request #115 from bxparks/develop

merge 2.3.0 into master

Author: bxparks

.github/workflows/validation.yml

@@ -61,19 +61,19 @@ jobs:
     - name: Build Compare tools
       run: |
         make -C ../acetimec/src
+        make -C ../AceTimeValidation/tools/compare_acetime
         make -C ../AceTimeValidation/tools/compare_acetimec
         make -C ../AceTimeValidation/tools/compare_libc
         make -C ../AceTimeValidation/tools/compare_hinnant
 
-    # Run the tests in AceTimeValidations (HinnantBasicTest,
-    # HinnantExtendedTest, AcetzBasicTest, AcetzExtendedTest). We don't run the
-    # others because when a new TZDB version comes out, the Python zoneinfo,
-    # Python pytz, Python dateutil, and Java validation tests will fail because
-    # they depend on the obsolete TZ database on the host Operating System. The
-    # 'make validations' will also compile the binary in
-    # AceTimeValidation/compare_cpp/ as necessary.
+    # Run the 'tests' target in AceTimeValidations which runs HinnantBasicTest,
+    # HinnantExtendedTest, AcetzBasicTest, and AcetzExtendedTest. We don't run
+    # the others because when a new TZDB version comes out, the Python
+    # zoneinfo, Python pytz, Python dateutil, Java tests, and others will fail
+    # because they depend on the obsolete TZ database on the host Operating
+    # System.
     - name: AceTimeValidation
       run: |
         cd ../AceTimeValidation
-        make -C tests tests
+        make -j2 -C tests tests
         make -C tests runtests

CHANGELOG.md

@@ -1,6 +1,79 @@
 # Changelog
 
 * Unreleased
+* 2.3.0 (2023-06-27, TZDB version 2023c)
+    * Create [CustomZoneRegistry](examples/CustomZoneRegistry/) example
+      to illustrate how to use a custom registry.
+    * Update `BasicZoneManager` to detect gaps.
+        * Allows `AceTimeValidation/validation` tests to pass.
+        * Still not able to distinguish between exact and overlap though.
+    * Allow multi-character `ZoneRule.letter` in `BasicZoneProcessor`.
+        * Regenerate `zonedb` which enables 2 more zones: `/Africa/Windhoek`,
+          and `America/Belize`
+        * Increases `zonedb` by 150-200 bytes.
+    * Rename `BrokerFactory` to `ZoneInfoStore` for better self-documentation.
+        * This is an internal implementation detail. Downstream clients should
+          not be affected.
+    * `ZoneContext`
+        * Move `ZoneContext`, `letters[]`, `fragments[]` into PROGMEM.
+        * Move `zonedbXxx::kTzDatabaseVersion` string into PROGMEM, and change
+          type to `const __FlashString*`.
+        * Create `ZoneContextBroker` around `ZoneContext`, for consistency
+          with all other zoneinfo data structures. Merge `ZoneContext.h` into
+          `ZoneInfo.h`.
+        * Add `startYearAccurate` and `untilYearAccurate` to `ZoneContext` which
+          define the interval of accurate transitions.
+        * Saves around 150-200 bytes of RAM on AVR processors,
+          200-350 bytes of RAM on ESP8266.
+    * Cleanly separate zoneinfo storage classes from their brokers classes.
+        * Three storage implementations instead of one: `zoneinfolow`,
+          `zoneinfomid`, `zoneinfohigh`, supporting low, middle, and high time
+          resolutions.
+        * `zoneinfohigh` was created to support `CompleteZoneProcessor`,
+          `CompleteZoneManager` and the `zonedbc` database.
+        * `zoneinfomid` initially used in `ExtendedZoneProcessor`, but now
+          unused.
+        * Merge `zoneinfo/ZonePolicy.h` into `zoneinfo/ZoneInfo.h`.
+        * Convert structs in `zoneinfo/ZoneInfo.inc` into templates, with
+          synthetic `typename S` selector. Merge `ZoneInfo.inc` into
+          `ZoneInfo.h`.
+    * Support timezones before 1972 using `CompleteZoneManager`,
+      `CompleteZoneProcessor`, and `zonedbc` database
+        * Before standardizing on UTC around 1970, many timezones had local
+          times offsets from UTC which require one-second resolution instead of
+          one-minute resolution. This requires changing many internal variables
+          from `int16_t` to `int32_t`.
+        * Change `TimeOffset` to support one-second resolution using `int32_t`.
+        * Rename `BasicZone::stdOffseMinutes()` to `BasicZone::stdOffset()`
+          which returns a `TimeOffset` object.
+        * Rename `ExtendedZone::stdOffseMinutes()` to
+          `ExtendedZone::stdOffset()` which returns a `TimeOffset` object.
+        * Leave `BasicZoneProcessor` using one-minute resolution because its
+          algorithm has an inherent limitations which cannot handle many
+          timezones before 1972, so no need to convert `int16_t` to `int32_t`
+          fields.
+        * Add `CompleteZoneManager`, `CompleteZoneProcessor`,
+          `CompleteZoneProcessorCache`, `CompleteZone`
+        * Add `zonedbc` database which is valid from `[1800,10000)`, which
+          includes all transitions in the TZDB since the first transition is
+          1844.
+        * Add `scope=complete` to access the `zonedbc` database.
+    * Zone Processor with graceful degradation
+        * Remove range checks against `ZoneContext.startYear()` and
+          `ZoneContext.untilYear()`
+        * Replace with `LocalDate::kMinYear` and `LocalDate::kMaxYear`, mostly
+          for formatting reasons (prevent negative
+          years, and years with more than 4 digits).
+        * The zone processors will always return something reasonble across the
+          entire `int16_t` range.
+        * Only the accuracy suffers outside of the `startYearAccurate()` and
+         `untilYearAccurate()` limits.
+        * Along with v2.2 which always generates anchor rules for all zone
+          policies, the `startYearAccureate()` and `untilYearAccurate()` allows
+          for graceful degradation of classes like `ZonedDateTime` for years
+          outside of this accuracy range.
+    * Rename `src/tzonedb*` directories
+        * to `src/zonedb*testing` for consistency with other acetime libraries
 * 2.2.3 (2023-05-31, TZDB version 2023c)
     * Update `ace_time/testing/*` classes to support splitting the test data
       into 2 separate lists: `transitions` and `samples`.