`BoolItem`: `numpy.bool_` compatibility fix

[PierreRaybaut]

BoolItem: numpy.bool_ Compatibility Fix

Issue Title

Fix BoolItem to convert numpy.bool_ to Python bool for Qt compatibility

Labels

• bug
• enhancement
• compatibility

Description

Problem

BoolItem values could be stored as numpy.bool_ type instead of Python bool, causing compatibility issues with Qt APIs that strictly require Python bool type.

Error encountered:

TypeError: setChecked(self, a0: bool): argument 1 has unexpected type 'numpy.bool'

This occurred in applications like DataLab when:

1. Image parameters with boolean values are saved to HDF5
2. Values are deserialized and later used with Qt widgets
3. Qt's QAction.setChecked() (and similar methods) reject numpy.bool_ values

This bug was reported in downstream project DataLab Issue #265.

Root Cause

When boolean values are read from HDF5 files via h5py, they can be stored as numpy.bool_ scalars. While numpy.bool_ is technically a subclass of Python bool, Qt's type system performs strict type checking and requires the native Python bool type.

Solution

Added a __set__ override to the BoolItem class that automatically converts all assigned values to Python bool:

def __set__(self, instance: DataSet, value: bool | None) -> None:
    """Set data item's value, ensuring it's a Python bool

    This override ensures that numpy.bool_ values are converted to Python bool,
    which is necessary for compatibility with Qt APIs that strictly require
    Python bool type (e.g., QAction.setChecked()).
    """
    if value is not None:
        value = bool(value)
    super().__set__(instance, value)

Benefits

• Generic fix: Handles the issue at the data storage layer, protecting all BoolItem users
• Prevents propagation: Stops numpy.bool_ from spreading throughout the application
• Backward compatible: bool(bool) is a no-op, so existing code is unaffected
• Works for all scenarios: Direct assignment, HDF5 deserialization, etc.

Testing

• Added comprehensive unit tests in guidata/tests/unit/test_boolitem_numpy.py
• Tests cover numpy.bool_ assignment, Python bool assignment, HDF5 serialization, and mixed scenarios
• All existing guidata tests pass (58 tests)

Files Changed

• guidata/dataset/dataitems.py: Added __set__ method to BoolItem class
• guidata/tests/unit/test_boolitem_numpy.py: New test file (5 tests)

Impact

This fix resolves Qt compatibility issues in downstream projects (PlotPy, DataLab, Sigima) without requiring changes to those codebases.