the unspecified date character (formerly lower case ‘u’) is superseded by the character (upper case) 'X';

Brings some updates to comply with final standard

…ad of 'u')

… the exponential indicator 'e', both previously lowercase, are now 'Y' and 'E' (uppercase);

the significant digit indicator 'p' is now 'S' (uppercase) --> significat digit is still NOT implemented yet

…n' and 'open' have been replaced with null and the double-dot notation ['..'] respectively;

general overwork of intervals and consecutives to deal with open beginnings and open ends

…eason; year overflow for northern winter/southern sommer still not considered

…pecification expression

…tion of the new spec but keep the old way with brackets

…tion of the new spec but keep the old way with brackets

…into develop

…g way with brackets

Partial merge of fork from https://github.com/saw-leipzig/python-edtf.git

- Tests are failing due to testInterval(), but are being run correctly. On Python 3.12, `nose` is failing. We should upgrade to `nose2`, or more likely, switch from unittest + nose to pytest.
- Move tox config to setup.cfg for now
- TODO move to pyproject.toml instead of setup.cfg
- TODO remove .travis.yml

Remove setup.cfg and setup.py in favor of pyproject.toml

Nose and Tox are next to go. Nose seems to be failing on Python 3.10, no need for it any more as pytest should cover its features. Likely no need for tox either as we are using the matrix strategy in Github Actions, though it could be useful for local dev

Github Actions CI and modern config

- Use parameterization in pytest to simplify many test cases by wrapping the input and expected values in tuples
- No need for unittest or class wrappers
- There is a legitimately failing test in parser/tests.py - code change needed in parser_classes to handle this I think
- natlang/tests.py will need to be updated to match the new spec, as will the whole text_to_edtf() function

The max and min functions now use a generator expression to filter out infinite values unless they are directly relevant to the calculation; if inf or -inf are found, they are returned instead of doing a comparison.

Use v5, which uses node v20 rather than deprecated v16

In Django 3.0, "support for the context argument of Field.from_db_value() and Expression.convert_value() is removed": https://github.com/django/django/blob/91a4b9a8ec2237434f06866f39c7977e889aeae6/docs/releases/3.0.txt#L641-L642

Previously, the Django field could directly take an EDTF string, only a natural language string that was then parsed and turned into EDTF.

Django 4 test project for now

Ignore SQLlite local database

This basic Django app shows how a user could create a model using the EDTFField and store data in it. The integration tests check that the EDTFField and associated fields (date_edtf_direct and date_display, in this case) work correctly.

There is a weird issue in test_date_display() where if we use an instance variable (self.event1, self.event2) the event.date_display property is available, but if we retrieve the object from the database it is not. I tried using TestEvent.objects.create() as well as the current method (make and then save an instance to no effect).

CI is set up to run the Django integration tests after Pytest. We could move to using pytest/django-pytest for these tests as well

In the parser:
- Update regular expressions for SHORT_YEAR_RE and LONG_YEAR_RE to use X instead of x and u and Y instead of y
- Replaced`unknown` with null as per the 2018 spec. It does not look like python-edtf currently has open intervals (`open` before, `..` now)?
- Replaced `?~` with `%`

In the tests:
- eliminate masked precision - no u/x just X for unknown regardless of why the data is missing
- replace unknown with null
- replace ~? with %

Reformat tests to use Pytest style; fix infinite comparison for OneOfASet

Ensure that `EDTFField` properly updates related fields whenever it changes inspired by ImageField.

- Use EDTFFieldDescriptorClass as a descriptor for EDTFField. This inherits from DeferredAttribute and handles getting, setting, and updating values. Whenever the field value is set, additional logic is processed to potentially update the field again based on other fields.
- update_values() replaces pre_save() to better handle updates/dependencies when EDTFField value changes
- contribute_to_class() attaches update_values() to the `post_init` signal

These changes should make the field updates more stable and (not reliant on definition order in models using EDTFField).

Thanks for the suggestion @aweakley #47 (comment)

Co-Authored-By: aweakley <[email protected]>

Django integration tests

Use immediate modules within the package, and import from the package level for integration tests.

For now, ignore `pickle` security warning.

Split requirements into primary and dev if still installing from requirements.txt

Ruff config and module import updates

Apply it to the entire date when a date is parsed as UncertainOrApproximate (L1 qualified)

Update natural language parser

The ci.yml updates add a commit to PRs. I tested this locally using `act` as best I could, but ran into an issue that I think will resolve when running on the real runner (undefined head). We'll see how it works when the workflows actually run ...

The new workflow adds a badge to the readme based on coverage for Python 3.12.

(for now at least..)

- Switch to using pytest-django to run the Django tests, as that has JUnit support. Add Django settings as a flag rather than in pyproject.toml because defining it there makes the normal pytest run fail since it can't find the module.
- Adds a simple script using junitparser to combine the two JUnit XML files.

```
File read successfully "/home/runner/work/python-edtf/python-edtf/./combined_junit_pytest.xml"
Warning: Your comment is too long (maximum is 65536 characters), coverage report will not be added.
Warning: Try add: "--cov-report=term-missing:skip-covered", or add "hide-report: true", or add "report-only-changed-files: true", or switch to "multiple-files" mode
```

Significant digits should work on a year in any format: "four-digit, 'Y' prefix, or exponential." These correspond to the python-edtf classes of Date, LongYear, and ExponentialYear.

- Add significant digit parsing for Date (year) and LongYear (y-prefixed)
- Standardize grammar for significant digits
- Use significant_digits rather than precision. Precision is used throughout for other functionality.
- Add estimated() public functions for the above EDTF classes

Two tests were failing:
```
FAILED edtf/parser/tests.py::test_edtf_examples[[1667, 1668, 1670..1672]-expected_tuple62] - AttributeError: 'list' object has no attribute 'expandtabs'
FAILED edtf/parser/tests.py::test_edtf_examples[{1667,1668, 1670..1672}-expected_tuple67] - AttributeError: 'list' object has no attribute 'expandtabs'
```

pyparsing.parse_string() was being passed a list by year somehow. Added year_basic for this use case (4 digit year without significant digits). If we need to support Consecutives with significant digits then this isn't a sufficient workaround.

- Adds functionality for significant digits to Date, LongYear, and ExponentialYear
- Updates the tests for significant digits
- Updates the docs for significant digits and a few other references to old syntax (lowercase e, grouping)
- ExponentialYear inherits from LongYear so only need to add it there; LongYear does not inherit from Date, so a bit of code duplication in the _fuzzy() overrides

Significant Digits

No CI yet. #50

50 add performance tests

Unspecified dates previously could not handle qualification. Unspecified dates also couldn't handle dates with 3 unspecified digits ("1XXX"). This commit adds both those features and tests for those use cases.

Requires quite a few overrides of lower_ and upper_ range methods to properly handle dates due to padding working in the opposite direction for negative dates, esp when combined with month/day padding.

If not in debug mode, use a simpler EDTFParseException rather than returning the full pyparsing error

Handle unspecified + qualified EDTF strings