Decorator for timelapse

[jo-mueller]

This PR adds a decorator that allows running functions on 4D points/surfaces/labels/etc.

Features

• It inspects the called functions and checks the argument type annotations.
• If it finds arguments of type napari.types.ImageData, LabelsData, PointsData or SurfaceData, it selects a suitable function to convert such 4D data to a list of 3D datasets which can then be processed frame by frame
• Depending on the return type annotation (e.g., ImageData, SurfaceData, etc), the resulting list of 3D data is converted to a single 4D data object.

Currently supported:

supported_data = [ImageData, PointsData, SurfaceData, LabelsData]

Open issues:

The conversion functions for 4D -> 3D should check whether the data actually is 4D (or 3D or 2D) and handle this somehow. For this data, however, it works for now.

Tests

Functions are tested to a certain degree, but only very weakly.

[jo-mueller]

@zoccoler if you have a few minutes to spare for this - the decorator running functions over multiple frames is the catchy part of this.

Does the current implementation make sense to you? What would be your preferred way to handle the "non" 4D-case?

[haesleinhuepf]

Hi Johannes @jo-mueller ,

I was commenting here instead of #11 because it's more convenient to comment on code in PRs than in issues.

So this looks all very cool! I could imagine though that making something like a general Converter class + some specific PointConverter, etc may make sense.

Best,
Robert

[jo-mueller]

To whom it may concern @haesleinhuepf @zoccoler @Cryaaa ,

I made quite a few changes to the code according to the suggestion above. A few notes to ease oversight:

• New Converter class that convert lists of frames to a single layer object (Converter.list_of_data_to_data(...) and Converter.data_to_list_of_data(...)). It does not yet support all layer types as input (LayerDataTuples cannot be handled as input and VectorData does not work).
• The Converter class is not yet able to differentiate between 3D and 4D data, which is crucial for further usage. I think this belongs in a separate PR, though.
• There are quite some functions here that I needed for processing but of which I think that they ultimately belong in napari-process-points-and-surfaces in which I will move upstream in the next PRs
• Most functions are not yet available as widgets in Napari but are only used from the notebooks. I think I will change this as soon as the time-slicing thing is resolved. For now, the functions serve their purpose primarily by being available from code.

[Cryaaa]

@jo-mueller, I don't know how feasible it is to use the time slicer functionality (or the concepts behind it) for points and surfaces. I really like the way 2D data is dealt with (t, z = 1, y, x), since this keeps things very uniform. Furthermore, the time slice will (soonish) have a version which directly generates a lazily loaded 4D dataset as a result. Maybe that would be something interesting? If you want to have a chat about 4D processing and approaches I'd be down :)

[jo-mueller]

Hi @Cryaaa ,

Thanks for the Feedback. The original idea of this improvement was to create a simple version of the time slicer that would work with 4D points/surface data because the "real" timeslicer is currently not able to do this and I really need this sort of functionality for this project.

I hope it's possible to merge this with the real timeslicer somewhere down the road.

I'll try to split this into several PRs so that it becomes clearer what is happening where.

[jo-mueller]

I'm closing this PR to be continued in a more concise thread PR #18