Empty cross-section curves, anti-aliasing kernel TypeError and About AttributeError #89

Describe the bugs

Three independent bugs, found while reviewing PlotPy before the 2.12.0 release.

1. Cross-section curve stays empty for an image added or shown again

With the X/Y cross-section panels enabled and a cross-section marker (or shape) already active, the cross-section curve of an image is empty (dataSize() == 0) when:

• the image is shown again after having been hidden (e.g. from the item list panel, or with plot.set_item_visible(item, True));
• the image is added to the plot.

The curve is only computed once the marker is moved again.

BaseCrossSectionPlot.items_changed (plotpy/panels/csection/csplot.py) creates the cross-section items of the new images after calling update_plot(refresh=False), and never computes their data. Before commit 9a221b7 ("only keep cross-section curves for visible images", not released yet), the curve of an image shown again was kept while hidden: it was not empty, but it showed the cross section at a former marker position. For an image added to the plot, the curve has always been empty.

Besides, the shade of the cross-section curves is SHADE / len(new_sources), where new_sources includes hidden images, so that the curves become almost transparent when many images are hidden.

2. TypeError: AA Mask data type must be ... with the anti-aliasing interpolation

The C++ scaling functions (_scale_rect, _scale_xy, _scale_tr) require the anti-aliasing kernel (item.interpolate[1]) to have the same data type as the array they scale. This is not the case:

• when exporting an image area without the LUT but with the interpolation (export_roi(..., apply_lut=False, apply_interpolation=True), e.g. with the rectangle snapshot tool, "Apply contrast settings" unchecked and "Apply interpolation algorithm" checked) while the Z-axis logarithmic scale is enabled: the kernel has the type of the logarithm of the data (float64), whereas the raw data (e.g. uint16) is exported;
• after set_data with data of another type (e.g. uint16 replaced by float32): the kernel keeps the former type. Both exporting the image and drawing it then fail. When drawing, the exception is raised in a Qt virtual method, which aborts the application with PyQt5 (exit code 0xC0000409 on Windows).

This affects ImageItem, XYImageItem, TrImageItem and their masked variants.

3. AttributeError in plotpy.widgets.about.get_python_libs_info and plotpy-benchmarks

plotpy.widgets.about.get_python_libs_info() and the benchmark banner (plotpy-benchmarks GUI script) call guidata.widgets.about.get_python_libs_info, which does not exist: the guidata function is get_python_libs_infos. Both raise:

AttributeError: module 'guidata.widgets.about' has no attribute 'get_python_libs_info'

This was introduced by the "info"/"infos" renaming of commit 6fc54bd (PlotPy 2.8.0), which also renamed the call to the guidata function. Pylint reports it as E1101.

To Reproduce

import numpy as np
from guidata.qthelpers import qt_app_context
from qtpy import QtCore as QC
from plotpy.builder import make
from plotpy.items import assemble_imageitems

with qt_app_context(exec_loop=False):
    # 1. Empty cross section
    win = make.dialog(type="image", show_xsection=True, show_ysection=True)
    win.show()
    plot = win.manager.get_plot()
    xcs = win.manager.get_xcs_panel().cs_plot
    img1 = make.image(np.arange(256, dtype=float).reshape(16, 16))
    img2 = make.image(np.full((16, 16), 42.0))
    plot.add_item(img1)
    plot.add_item(img2)
    marker = make.marker(position=(5, 5))
    plot.add_item(marker)
    xcs.register_shape(plot, marker)
    print(xcs.known_items[img1].dataSize())  # 16
    plot.set_item_visible(img1, False)
    plot.set_item_visible(img1, True)
    print(xcs.known_items[img1].dataSize())  # 0

    # 2. Anti-aliasing kernel data type
    rect = QC.QRectF(0, 0, 6, 4)
    item = make.image(np.arange(1, 25, dtype=np.uint16).reshape(4, 6))
    item.set_interpolation(2, 3)  # INTERP_AA
    item.set_zaxis_log_state(True)
    assemble_imageitems([item], rect, 6, 4, apply_lut=False, apply_interpolation=True)
    # TypeError: AA Mask data type must be 4

    # 3. About
    from plotpy.widgets.about import get_python_libs_info
    get_python_libs_info()  # AttributeError

Expected behavior

1. The cross-section curve of an image added or shown again is computed as soon as the image is shown (unless the auto-refresh of the cross-section panel is disabled), and the shade of the curves depends on the number of visible images only.
2. Exporting and drawing images with the anti-aliasing interpolation work whatever the Z-axis scale and the data type changes.
3. get_python_libs_info returns the Python and libraries information, and plotpy-benchmarks runs.