Spotfire

NumPy 2.0 Support
* Update requirement to support NumPy 2.0. (54)
* Update other requirements to NumPy 2.0 compatible versions, including Pandas and Cython (47).
* Due to NumPy's requirements, update the minimum Python requirement to 3.9 (from 3.7).

SBDF Improvements
* `GeoDataFrame` objects (from the `geopandas` package) are no longer automatically exported as if they were Spotfire geocoding tables. The `XMin`, `XMax`, and `XCenter` columns (and the corresponding `Y*` versions) are not recomputed (nor are they created if they do not exist in the original `GeoDataFrame`), and the metadata properties are not set (the biggest issue is with the `MapChart.GeometryType` property, which can prevent data sets with mixed types of geometry (e.g., LineStrings and MultiLineStrings) from being exported). For users who require these columns and metadata properties for their Spotfire analyses, the export logic has been moved to a new public function `set_geocoding_table` (in the `spotfire` module). (44)
* Correctly export NumPy `datetime64` values for all date/time resolutions (i.e., other than `datetime64[ns]`). (65)

Data Function Improvements
* Expose information about data function inputs and outputs to the data function via new dunder-objects (`__spotfire_inputs__` and `__spotfire_outputs__`, sequences of `AnalyticInput` and `AnalyticOutput` objects) in the global namespace of the data function. (64)
* Make it possible to set filenames in data function backtraces (instead of `<data_function>`. Intended for use during data function development; this feature is not used by Spotfire. (67)
* Fix a unit test that could fail under race conditions based on when the test was run. (52)

Spotfire SPK Package Building Improvements
* When building SPK packages, now include the requirements and constraints files used to construct the SPK in the package. (56)
* Use more Pythonic methods for reading package metadata and identifying package files. Removes usages of the deprecated `pkg_resources` module from the `setuptools` package (removed in version 67.5.0). (59)

Other Improvements
* Add a new `spotfire.support` module for creating troubleshooting bundles with information about your Python environment for use in diagnosing problems together with Spotfire support. (57)

Packaging Improvements
* Migrate the project to use the `pyproject.toml` file as the primary project metadata location.
* Expose optional packages (like `geopandas` for geographic data or the various supported plotting packages) as `spotfire` package extras. (43)
* Improved the buildability of the package from the 'sdist' distribution. Mainly affects platforms that are not using the pre-compiled wheels. (55)
* Added Python 3.12 to the set of pre-compiled wheels. (66)
* Fixed issues with `Py_UNICODE`'s removal from Python 3.12.
* Improved the static analysis of all code.
* Updated the version of PyLint used for development. (38, 68)
* Added static typing analysis with `mypy` (51), general analysis of Cython code with `cython-lint` (49), and general analysis of C code with `cpplint` (50).

* Support the code signing of built SPK packages with certificates in Windows system certificate stores. (58)

* Update the NumPy requirement to exclude 2.0 releases based on the expected C ABI break as part of that release. (53)

* Do not fail with an "unknown dtype" when exporting [timezone aware](https://docs.python.org/3/library/datetime.html#determining-if-an-object-is-aware-or-naive) datetime objects to SBDF. (30)
* Reduce the logging of table and column metadata to the data function flow's debug logs to avoid overly-large log output. (48)

* Correctly export PIL images, Matplotlib figures, and Seaborn grids. (42)

Performance Improvements

* The `spotfire.sbdf` module has been rewritten as a native C extension using [Cython](https://cython.org), which allows using the [C Library for TIBCO Spotfire Binary Data Format](https://github.com/TIBCOSoftware/spotfire-sbdf-c) (vendored into the `vendor/sbdf-c` directory) instead of the original pure-Python implementation. This change improves performance and allows the reading and writing of SBDF files that are written using RLE encoding (which was not implemented for the pure-Python implementation). (#36)
* With the introduction of Cython support in `setup.py`, the `spotfire.codesign` module has been rewritten using Cython.
* To work around the deprecation of `msilib` in Python 3.11, the `spotfire.cabfile` module has been rewritten using Cython to directly call the Windows API to create cabinet files. Non-Windows platforms will raise `OSError` exceptions on object creation instead of the current behavior of conditionally defining the class. (29)

As a result of these architectural changes, the version number of this package has been incremented to 2.0.0.

Creating SPK Packages

* [User installed](https://pip.pypa.io/en/latest/user_guide/#user-installs) packages (installed with `pip install --user`) could override the `spotfire` package that is deployed via the Spotfire server, potentially installing an older version that does not contain support for required features. Now, when you build an interpreter SPK package (using `python -m spotfire.spk python ...`), the `spotfire` package is installed into a separate directory, which is then automatically inserted into the Python `sys.path`, before any "site-packages" directory, on interpreter startup. (39)
* When you build a packages SPK package (using the command `python -m spotfire.spk packages ...`), wheels with data directories (such as `xgboost` 1.7.5 on Windows) no longer result in SPK packages missing Python files from the wheel. (40)