WIP: Refactor PV slicing tool

[astrofrog]

Still very early days... Remaining things to do in this PR:

• Rebase once Fix compatibility with PySide2 and drop support for PySide and PyQt4 #1662 is merged
• Extract slice for all attributes from current reference data
• Automatically open an image viewer if needed
• Add status bar documentation
• Rename tool to not be PV-specific
• Add tests
• Add changelog entry
• Fix the fact that rectangular selections are somehow propagated to PV slices
• Make subsets work properly
• When closing window with a slice and then updating the slice, do we re-open a new window?
• How do we deal with multiple datasets being present? do we extract all datasets as attributes in the new PV slice at resolution of reference data?
• Seeing C/C++ reference error when closing a PV slice window and then re-openng it with data and using the mouse over path
• Use APE 14 WCS to get nice labels on sliced data
• Don't show PV sliced data in link editor? (since links are implicit). Or show them with a link.
• Test that things work fine in a multi-attribute dataset

Things that could be done in future:

• Extract slice for all visible layers (although only pixel-aligned ones)
• Extract slice for reprojected data
• Preserve derived components
• Make PVSliceData components be dynamically evaluated
• Allow thick slices
• Have a way of selecting whether a new slice is created or modifying the old one (for now we can just have one path per viewer and have that always update)

Other thoughts

• For now, for simplicity, don't use links between PV sliced data and parent data - allow subsets on main data to be viewer in slice but selections in slice shouldn't propagate back (except maybe via the non-slice axis)

[codecov]

Codecov Report

Merging #1663 (e1a520b) into main (2d64631) will decrease coverage by 0.42%.
The diff coverage is 69.69%.

❗ Current head e1a520b differs from pull request most recent head a61a71c. Consider uploading reports for the commit a61a71c to get more accurate results

@@            Coverage Diff             @@
##             main    #1663      +/-   ##
==========================================
- Coverage   88.07%   87.64%   -0.43%     
==========================================
  Files         247      247              
  Lines       23279    23272       -7     
==========================================
- Hits        20502    20397     -105     
- Misses       2777     2875      +98     

Impacted Files	Coverage Δ
glue/core/data.py	91.66% <ø> (ø)
glue/viewers/image/qt/__init__.py	100.00% <ø> (ø)
glue/app/qt/application.py	78.38% <50.00%> (-0.19%)	⬇️
glue/plugins/tools/pv_slicer/pv_sliced_data.py	64.95% <64.95%> (ø)
glue/plugins/tools/pv_slicer/qt/pv_slicer.py	75.80% <70.96%> (+2.47%)	⬆️
glue/viewers/matplotlib/toolbar_mode.py	95.69% <87.50%> (-0.42%)	⬇️
glue/plugins/tools/pv_slicer/qt/__init__.py	100.00% <100.00%> (ø)
glue/viewers/common/qt/toolbar.py	95.07% <100.00%> (-2.81%)	⬇️
glue/viewers/image/qt/data_viewer.py	100.00% <100.00%> (ø)
glue/viewers/image/qt/profile_viewer_tool.py	33.33% <100.00%> (+9.92%)	⬆️
... and 16 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 2d64631...a61a71c. Read the comment docs.

[astrofrog]

I'm currently stuck on an issue with Matplotlib getting stuck in an infinite loop in the following case:

• Set up an RGB image with three cubes
• Extract a path
• Update the path

When the path is updated, the second viewer then gets stuck in a drawing loop and never renders fully. Some things that seem to matter - if in the state class for the image viewer I return a random array instead of the fixed resolution buffer after the initial call, then there is no infinite loop. If I do the same in the PV SlicedData, it sometimes helps and sometimes not (randomly). This makes me think there is a race condition of some kind.

[astrofrog]

Actually I think I found where the issue comes from - if I change the get_image_data on the layer artist to comment out the attribute disabling (which triggers a clear):

    def get_image_data(self, bounds=None):

        if self.uuid is None:
            return None

        # try:
        image = self.state.get_sliced_data(bounds=bounds)
        # except (IncompatibleAttribute, IndexError):
        #     # The following includes a call to self.clear()
        #     # self.disable_invalid_attributes(self.state.attribute)
        #     return None
        # else:
        #     self.enable()

then I get the following error which shouldn't be happening:

Traceback (most recent call last):
  File "/home/tom/Code/matplotlib/lib/matplotlib/backends/backend_qt.py", line 465, in _draw_idle
    self.draw()
  File "/home/tom/Code/glue/glue/viewers/matplotlib/qt/widget.py", line 124, in draw
    return super(MplCanvas, self).draw(*args, **kwargs)
  File "/home/tom/Code/matplotlib/lib/matplotlib/backends/backend_agg.py", line 436, in draw
    self.figure.draw(self.renderer)
  File "/home/tom/Code/matplotlib/lib/matplotlib/artist.py", line 73, in draw_wrapper
    result = draw(artist, renderer, *args, **kwargs)
  File "/home/tom/Code/matplotlib/lib/matplotlib/artist.py", line 50, in draw_wrapper
    return draw(artist, renderer)
  File "/home/tom/Code/matplotlib/lib/matplotlib/figure.py", line 2859, in draw
    mimage._draw_list_compositing_images(
  File "/home/tom/Code/matplotlib/lib/matplotlib/image.py", line 133, in _draw_list_compositing_images
    a.draw(renderer)
  File "/home/tom/python/dev/lib/python3.9/site-packages/astropy/visualization/wcsaxes/core.py", line 464, in draw
    super().draw(renderer, **kwargs)
  File "/home/tom/Code/matplotlib/lib/matplotlib/artist.py", line 50, in draw_wrapper
    return draw(artist, renderer)
  File "/home/tom/Code/matplotlib/lib/matplotlib/axes/_base.py", line 3105, in draw
    mimage._draw_list_compositing_images(
  File "/home/tom/Code/matplotlib/lib/matplotlib/image.py", line 133, in _draw_list_compositing_images
    a.draw(renderer)
  File "/home/tom/Code/matplotlib/lib/matplotlib/artist.py", line 50, in draw_wrapper
    return draw(artist, renderer)
  File "/home/tom/Code/matplotlib/lib/matplotlib/image.py", line 636, in draw
    im, l, b, trans = self.make_image(
  File "/home/tom/Code/mpl-scatter-density/mpl_scatter_density/base_image_artist.py", line 186, in make_image
    array = self._array_func(bins=bins, range=((ymin, ymax), (xmin, xmax)))
  File "/home/tom/Code/glue/glue/viewers/image/frb_artist.py", line 27, in array_func_wrapper
    array = self._array_maker(bounds)
  File "/home/tom/Code/glue/glue/viewers/image/composite_array.py", line 94, in __call__
    array = layer['array'](bounds=bounds)
  File "/home/tom/Code/glue/glue/viewers/image/layer_artist.py", line 138, in get_image_data
    image = self.state.get_sliced_data(bounds=bounds)
  File "/home/tom/Code/glue/glue/viewers/image/state.py", line 452, in get_sliced_data
    image = self.layer.compute_fixed_resolution_buffer(full_view, target_data=self.viewer_state.reference_data,
  File "/home/tom/Code/glue/glue/plugins/tools/pv_slicer/pv_sliced_data.py", line 211, in compute_fixed_resolution_buffer
    result = compute_fixed_resolution_buffer(self.original_data, new_bounds,
  File "/home/tom/Code/glue/glue/core/fixed_resolution_buffer.py", line 244, in compute_fixed_resolution_buffer
    if target_cid is not None and isinstance(data, Data) and isinstance(data.get_component(target_cid), DaskComponent):
  File "/home/tom/Code/glue/glue/core/data.py", line 1434, in get_component
    raise IncompatibleAttribute(component_id)
glue.core.exceptions.IncompatibleAttribute: PRIMARY