Merge pull request #107 from dh-tech/feature/convert-hijri

Add support for converting from Hijri calendar to undate and undate interval

.github/workflows/unit_tests.yml

@@ -8,6 +8,8 @@ on:
       - 'undate/**'
       - 'tests/**'
   pull_request:
+    branches:
+      - "**"
 
 env:
   # python version used to calculate and submit code coverage

README.md

@@ -140,7 +140,7 @@ An `UndateInterval` is a date range between two `Undate` objects. Intervals can
 ```
 
 You can initialize `Undate` or `UndateInterval` objects by parsing a date string with a specific converter, and you can also output an `Undate` object in those formats.
-Available converters are "ISO8601" and "EDTF" (but only)
+Currently available converters are "ISO8601" and "EDTF" and supported calendars.
 
 ```python
 >>> from undate import Undate
@@ -156,6 +156,33 @@ Available converters are "ISO8601" and "EDTF" (but only)
 <UndateInterval 1800/1900>
 ```
 
+### Calendars
+
+All `Undate` objects are calendar aware, and date converters include support for parsing and working with dates from other calendars. The Gregorian calendar is used by default; currently `undate` supports the Hijri Islamic calendar and the Anno Mundi Hebrew calendar based on calendar convertion logic implemented in the [convertdate](https://convertdate.readthedocs.io/en/latest/)package. 
+
+Dates are stored with the year, month, day and appropriate precision for the original calendar; internally, earliest and latest dates are calculated in Gregorian / Proleptic Gregorian calendar for standardized comparison across dates from different calendars.
+
+```python
+>>> from undate import Undate
+>>> tammuz4816 = Undate.parse("26 Tammuz 4816", "Hebrew")
+>>> tammuz4816
+<Undate '26 Tammuz 4816 Anno Mundi' 4816-04-26 (Hebrew)>
+>>> rajab495 = Undate.parse("Rajab 495", "Hijri")
+>>> rajab495
+<Undate 'Rajab 495 Hijrī' 0495-07 (Hijri)>
+>>> y2k = Undate.parse("2001", "EDTF")
+>>> y2k
+<Undate 2001 (Gregorian)>
+>>> [str(d.earliest) for d in [rajab495, tammuz4816, y2k]]
+['1102-04-28', '1056-07-17', '2001-01-01']
+>>> [str(d.precision) for d in [rajab495, tammuz4816, y2k]]
+['MONTH', 'DAY', 'YEAR']
+>>> sorted([rajab495, tammuz4816, y2k])
+[<Undate '26 Tammuz 4816 Anno Mundi' 4816-04-26 (Hebrew)>, <Undate 'Rajab 495 Hijrī' 0495-07 (Hijri)>, <Undate 2001 (Gregorian)>]
+```
+
+* * * 
+
 For more examples, refer to the [example notebooks](https://github.com/dh-tech/undate-python/tree/main/examples/notebooks/) included in this repository.
 
 ## Documentation

docs/undate/converters.rst

@@ -1,19 +1,25 @@
 Converters
 ==========
 
+Overview
+--------
+
 .. automodule:: undate.converters.base
    :members:
    :undoc-members:
 
+Formats
+--------
+
 ISO8601
--------
+^^^^^^^
 
 .. automodule:: undate.converters.iso8601
    :members:
    :undoc-members:
 
 Extended Date-Time Format (EDTF)
---------------------------------
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
 
 .. automodule:: undate.converters.edtf.converter
    :members:
@@ -23,8 +29,25 @@ Extended Date-Time Format (EDTF)
    :members:
    :undoc-members:
 
-.. transformer is more of an internal, probably doesn't make sense to include
-.. .. automodule:: undate.converters.edtf.transformer
-..    :members:
-..    :undoc-members:
+
+Calendars
+---------
+
+Gregorian
+^^^^^^^^^
+
+.. automodule:: undate.converters.calendars.gregorian
+   :members:
+
+Hijri (Islamic calendar)
+^^^^^^^^^^^^^^^^^^^^^^^^
+
+.. automodule:: undate.converters.calendars.hijri.converter
+   :members:
+
+Anno Mundi (Hebrew calendar)
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+.. automodule:: undate.converters.calendars.hebrew.converter
+   :members:
 

pyproject.toml

@@ -9,7 +9,7 @@ readme = "README.md"
 license = { text = "Apache-2" }
 requires-python = ">= 3.9"
 dynamic = ["version"]
-dependencies = ["lark", "numpy"]
+dependencies = ["lark[interegular]", "numpy", "convertdate", "strenum; python_version < '3.11'"]
 authors = [
     { name = "Rebecca Sutton Koeser" },
     { name = "Cole Crawford" },

src/undate/converters/base.py

@@ -1,10 +1,11 @@
 """
-:class:`undate.converters.BaseDateConverter` provides a base class for
+:class:`~undate.converters.BaseDateConverter` provides a base class for
 implementing date converters, which can provide support for
-parsing and generating dates in different formats and also converting
-dates between different calendars.
+parsing and generating dates in different formats.
+The converter subclass :class:`undate.converters.BaseCalendarConverter`
+provides additional functionaly needed for calendar conversion.
 
-To add support for a new date format or calendar conversion:
+To add support for a new date converter:
 
 - Create a new file under ``undate/converters/``
     - For converters with sufficient complexity, you may want to create a submodule;
@@ -18,6 +19,26 @@
 The new subclass should be loaded automatically and included in the converters
 returned by :meth:`BaseDateConverter.available_converters`
 
+To add support for a new calendar converter:
+
+- Create a new file under ``undate/converters/calendars/``
+    - For converters with sufficient complexity, you may want to create a submodule;
+      see ``undate.converters.calendars.hijri`` for an example.
+- Extend ``BaseCalendarConverter`` and implement ``parse`` and ``to_string``
+  formatter methods as desired/appropriate for your converter as well as the
+  additional methods for ``max_month``, ``max_day``, and convertion ``to_gregorian``
+  calendar.
+- Import your calendar in ``undate/converters/calendars/__init__.py`` and include in `__all__``
+- Add unit tests for the new calendar logic under ``tests/test_converters/calendars/``
+- Add the new calendar to the ``Calendar`` enum of supported calendars in
+  ``undate/undate.py`` and confirm that the `get_converter` method loads your
+  calendar converter correctly (an existing unit test should cover this).
+- Consider creating a notebook to demonstrate the use of the calendar
+  converter.
+
+Calendar converter subclasses are also automatically loaded and included
+in the list of available converters.
+
 -------------------
 """
 
@@ -90,6 +111,54 @@ def available_converters(cls) -> Dict[str, Type["BaseDateConverter"]]:
         """
         Dictionary of available converters keyed on name.
         """
+        return {c.name: c for c in cls.subclasses()}  # type: ignore
+
+    @classmethod
+    def subclasses(cls) -> list[Type["BaseDateConverter"]]:
+        """
+        List of available converters classes. Includes calendar convert
+        subclasses.
+        """
         # ensure undate converters are imported
         cls.import_converters()
-        return {c.name: c for c in cls.__subclasses__()}  # type: ignore
+
+        # find all direct subclasses, excluding base calendar converter
+        subclasses = cls.__subclasses__()
+        subclasses.remove(BaseCalendarConverter)
+        # add all subclasses of calendar converter base class
+        subclasses.extend(BaseCalendarConverter.__subclasses__())
+        return subclasses
+
+
+class BaseCalendarConverter(BaseDateConverter):
+    """Base class for calendar converters, with additional methods required
+    for calendars."""
+
+    #: Converter name. Subclasses must define a unique name.
+    name: str = "Base Calendar Converter"
+
+    def min_month(self) -> int:
+        """Smallest numeric month for this calendar."""
+        raise NotImplementedError
+
+    def max_month(self, year: int) -> int:
+        """Maximum numeric month for this calendar"""
+        raise NotImplementedError
+
+    def first_month(self) -> int:
+        """first month in this calendar; by default, returns :meth:`min_month`."""
+        return self.min_month()
+
+    def last_month(self, year: int) -> int:
+        """last month in this calendar; by default, returns :meth:`max_month`."""
+        return self.max_month(year)
+
+    def max_day(self, year: int, month: int) -> int:
+        """maximum numeric day for the specified year and month in this calendar"""
+        raise NotImplementedError
+
+    def to_gregorian(self, year, month, day) -> tuple[int, int, int]:
+        """Convert a date for this calendar specified by numeric year, month, and day,
+        into the Gregorian equivalent date. Should return a tuple of year, month, day.
+        """
+        raise NotImplementedError

src/undate/converters/calendars/__init__.py

@@ -0,0 +1,5 @@
+from undate.converters.calendars.gregorian import GregorianDateConverter
+from undate.converters.calendars.hijri import HijriDateConverter
+from undate.converters.calendars.hebrew import HebrewDateConverter
+
+__all__ = ["HijriDateConverter", "GregorianDateConverter", "HebrewDateConverter"]

src/undate/converters/calendars/gregorian.py

@@ -0,0 +1,51 @@
+from calendar import monthrange
+
+from undate.converters.base import BaseCalendarConverter
+
+
+class GregorianDateConverter(BaseCalendarConverter):
+    """
+    Calendar converter class for Gregorian calendar.
+    """
+
+    #: converter name: Gregorian
+    name: str = "Gregorian"
+    #: calendar
+    calendar_name: str = "Gregorian"
+
+    #: known non-leap year
+    NON_LEAP_YEAR: int = 2022
+
+    def min_month(self) -> int:
+        """First month for the Gregorian calendar."""
+        return 1
+
+    def max_month(self, year: int) -> int:
+        """maximum numeric month for the specified year in the Gregorian calendar"""
+        return 12
+
+    def max_day(self, year: int, month: int) -> int:
+        """maximum numeric day for the specified year and month in this calendar"""
+        # if month is known, use that to calculate
+        if month:
+            # if year is known, use it; otherwise use a known non-leap year
+            # (only matters for February)
+            year = year or self.NON_LEAP_YEAR
+
+            # Use monthrange from python builtin calendar module.
+            # returns first day of the month and number of days in the month
+            # for the specified year and month.
+            _, max_day = monthrange(year, month)
+        else:
+            # if year and month are unknown, return maximum possible
+            max_day = 31
+
+        return max_day
+
+    def to_gregorian(self, year, month, day) -> tuple[int, int, int]:
+        """Convert to Gregorian date. This returns the specified by year, month,
+        and day unchanged, but is provided for consistency since all calendar
+        converters need to support conversion to Gregorian calendar for
+        a common point of comparison.
+        """
+        return (year, month, day)

src/undate/converters/calendars/hebrew/__init__.py

@@ -0,0 +1,3 @@
+from undate.converters.calendars.hebrew.converter import HebrewDateConverter
+
+__all__ = ["HebrewDateConverter"]

src/undate/converters/calendars/hebrew/converter.py

@@ -0,0 +1,78 @@
+from typing import Union
+
+from convertdate import hebrew  # type: ignore
+from lark.exceptions import UnexpectedCharacters
+
+from undate.converters.base import BaseCalendarConverter
+from undate.converters.calendars.hebrew.parser import hebrew_parser
+from undate.converters.calendars.hebrew.transformer import HebrewDateTransformer
+from undate.undate import Undate, UndateInterval
+
+
+class HebrewDateConverter(BaseCalendarConverter):
+    """
+    Converter for Hebrew Anno Mundicalendar.
+
+    Support for parsing Anno Mundi dates and converting to Undate and UndateInterval
+    objects in the Gregorian calendar.
+    """
+
+    #: converter name: Hebrew
+    name: str = "Hebrew"
+    calendar_name: str = "Anno Mundi"
+
+    def __init__(self):
+        self.transformer = HebrewDateTransformer()
+
+    def min_month(self) -> int:
+        """Smallest numeric month for this calendar."""
+        return 1
+
+    def max_month(self, year: int) -> int:
+        """Maximum numeric month for this calendar. In Hebrew calendar, this is 12 or 13
+        depending on whether it is a leap year."""
+        return hebrew.year_months(year)
+
+    def first_month(self) -> int:
+        """First month in this calendar. The Hebrew civil year starts in Tishri."""
+        return hebrew.TISHRI
+
+    def last_month(self, year: int) -> int:
+        """Last month in this calendar. Hebrew civil year starts in Tishri,
+        Elul is the month before Tishri."""
+        return hebrew.ELUL
+
+    def max_day(self, year: int, month: int) -> int:
+        """maximum numeric day for the specified year and month in this calendar"""
+        # NOTE: unreleased v2.4.1 of convertdate standardizes month_days to month_length
+        return hebrew.month_days(year, month)
+
+    def to_gregorian(self, year: int, month: int, day: int) -> tuple[int, int, int]:
+        """Convert a Hebrew date, specified by year, month, and day,
+        to the Gregorian equivalent date. Returns a tuple of year, month, day.
+        """
+        return hebrew.to_gregorian(year, month, day)
+
+    def parse(self, value: str) -> Union[Undate, UndateInterval]:
+        """
+        Parse a Hebrew date string and return an :class:`~undate.undate.Undate` or
+        :class:`~undate.undate.UndateInterval`.
+        The Hebrew date string is preserved in the undate label.
+        """
+        if not value:
+            raise ValueError("Parsing empty string is not supported")
+
+        # parse the input string, then transform to undate object
+        try:
+            # parse the string with our Hebrew date parser
+            parsetree = hebrew_parser.parse(value)
+            # transform the parse tree into an undate or undate interval
+            undate_obj = self.transformer.transform(parsetree)
+            # set the original date as a label, with the calendar name
+            undate_obj.label = f"{value} {self.calendar_name}"
+            return undate_obj
+        except UnexpectedCharacters as err:
+            raise ValueError(f"Could not parse '{value}' as a Hebrew date") from err
+
+    # do we need to support conversion the other direction?
+    # i.e., generate a Hebrew date from an abitrary undate or undate interval?