pyqt scatter widget

[fjorka]

Description

This PR introduces a complete Scatter Widget based on PyQtGraph, intended to replace the Matplotlib-based Scatter Widget.

Rationale

• Expected improved performance for large datasets.
• Smooth navigation for zooming and panning within the plot.
• Increased flexibility in customization.

Functionality

• It preserves all functionality of the previous widget.
• Documentation notebooks updated with the information and screenshots of the new widget.
• New functionality includes:
  • Separate color controls for continuous and discrete data:
    • Continuous color control offers:
      • Histogram of values chosen for color.
      • Ability to change LUT.
      • Control over contrast.
    • Discrete color control offers:
      • Scrollable legend of colors.
      • Ability to change the color for a selected class.
  • Pseudohistogram plot when only one axis is defined.
  • Highlight of nearby data points under cursor.
  • Dynamic status display of cursor position and highlighted data points.
  • Expanded ROI annotations:
    • Lasso or rectangle tool (useful for thresholding).
    • Ability to use several disjointed ROIs for annotation.
    • Modifiable ROIs.
  • Dialog to choose Annotation name (previously Export).
  • Saving options:
    • When started with AnnData, the Save button opens a path dialog to choose a location for an AnnData file (supports h5ad, zarr, and csv).
    • When opened with the Napari viewer, it mimics the behavior of the Annotation Widget:
      • With backed sdata, Save will modify sdata of the selected layer.
      • Without backed sdata, no saving action.
    • The widget displays the status and saving possibility when started.

Next Steps Beyond This PR

• Flip the Y-axis (scatterplot in spatial coordinates is flipped relative to the image).
• Signal point highlight to the main viewer for quick connection between visualizations in different dimensions.
• Highlight object in the main Napari window to trigger highlight in the scatterplot.
• Integrate Annotation Widget functionality into the Scatter Widget (the Annotation Widget facilitates annotating multiple classes within a single column).
• Performance improvements (1. How points belonging to polygons is checked.)

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[Czaki]

Once you have added widget to qtbot, then you should not use deleteLater.

[fjorka]

Hi @Czaki !

I do see the error:

E       AssertionError: Some instance of QtViewer is not properly cleaned in one of previous test. For easier debug one may use --save-leaked-object-graph flag for pytest to get graph of leaked objects. If you use qtbot (from pytest-qt) to clean Qt objects after test you may need to switch to manual clean using `deleteLater()` and `qtbot.wait(50)` later.
E       assert 1 == 0

when I run pytest locally (Python 3.10, Windows).

You are correct, the additional deleteLater is unnecessary. I added it following the error message as I was getting desperate. I will remove it to avoid confusion.

Now, the minimum example to reproduce the error is:

import pytest
from napari_spatialdata._model import DataModel
from napari_spatialdata._scatterwidgets import PlotWidget

@pytest.fixture
def plot_widget(qtbot):
    """Fixture for creating a PlotWidget instance."""
    model = DataModel()
    widget = PlotWidget(None, model)
    qtbot.addWidget(widget)
    yield widget

def test_initialization(plot_widget):
    """Test initialization of PlotWidget."""
    assert plot_widget is not None

def test_elementwidget(make_napari_viewer):
    _ = make_napari_viewer()
    assert 1 == 1

However, it's not the problem specific to this branch. If I remove all tests from test_scatterwidgets.py, the error persists:

Finally, the error is triggered only once. If I run the fake test (the code pasted in above) as test_aaa.py before any other test, the error appears in the fake test but is not triggered again by any of the original tests:

Any ideas what is going on here?

[Czaki]

It is not Idea. It is knowledge created in pain.
You may also see this error in tests executed on main.
Will try to fix this.
In general, the problem comes from previous tests that are spawning the Viewer.

in a general make_napari_viewer fixture has few weak spots (but writing a test without using it, may lead to even more spectacular failures)

I even have opened some PR to fix it, but it is really time-consuming.
for example napari/napari#6745

[fjorka]

• is it possible to remove the white border around the scatterplot points?

Replaced with black (invisible) border when color is defined - 1acdd85

• scatterwidget_annotation.ipynb is missing from docs/index.md

Included in 9191c7e

• Minor thing, can you please update the changelog? Thanks.

Suggestion in 973ebcc. Needs to be finalized.

[fjorka]

The notebooks start with the instruction:

There are two options to install napari-spatialdata:

(1) Run `pip install napari-spatialdata`
or,
(2) Clone this [repo](https://github.com/scverse/napari-spatialdata) and run `pip install -e .`

However, following this instruction and starting napari (or scatter widget) will give an error ImportError: No Qt bindings could be found. What about asking explicitly for napari installation first d40fb7c which will inform a user about Qt?

[fjorka]

To do:

• new screenshots for the notebooks (after approving the new look by @LucaMarconato )
• add explanations in scatterwidget.ipynb and scatterwidget_annotation.ipynb about starting with AnnData and options for saving
• (possibly) add a warning dialog about overwriting sdata object

[LucaMarconato]

TODO for me: dataset