Support .collate for DynamicMaps?

[jbednar]

The example on the holoviews.org home page shows how to use .collate() to turn the illegal "HoloMap of Layouts" construction into the displayable "Layout of HoloMaps", which is very convenient in practice. However, .collate() does not seem to be supported for DynamicMaps.

Here's an example of what I'd want to use that for:

import numpy as np
import holoviews as hv
import imagen as ig
import imagen.transferfn as tf

from imagen.image import FileImage
from imagen.transferfn.sheet_tf import Convolve

hv.notebook_extension()

%opts Image (cmap="gray") {+axiswise}

FileImage.filename = "/Users/jbednar/imagen/doc/images/ellen_arthur.pgm"

kdims = [hv.Dimension('size',                   range=(0.1, 2)),
         hv.Dimension('positive_size',          range=(0.1, 2)),
         hv.Dimension('negative_size',          range=(0.3, 4)),
         hv.Dimension('positive_aspect_ratio',  range=(1.0,10)),
         hv.Dimension('negative_aspect_ratio',  range=(1.0,10))]

Right now, I can easily view one static image along with a slider-controlled dynamic map, but I have to pick a single HoloViews object to return from the callback:

def filtered_images(size,positive_size,negative_size,positive_aspect_ratio,negative_aspect_ratio):
    convolver = Convolve(kernel_pattern=ig.DifferenceOfGaussians(**locals()))
    return FileImage(output_fns=[convolver])[:]

FileImage()[:] + hv.DynamicMap(filtered_images, kdims=kdims)

def filtered_images(size,positive_size,negative_size,positive_aspect_ratio,negative_aspect_ratio):
    convolver = Convolve(kernel_pattern=ig.DifferenceOfGaussians(**locals()))
    return convolver.kernel_pattern[:]

FileImage()[:] + hv.DynamicMap(filtered_images, kdims=kdims)

But what I want is to make one single callback that returns a Layout of both the two dynamic plots above:

def filtered_images(size,positive_size,negative_size,positive_aspect_ratio,negative_aspect_ratio):
    convolver = Convolve(kernel_pattern=ig.DifferenceOfGaussians(**locals()))
    return convolver.kernel_pattern[:] + FileImage(output_fns=[convolver])[:]

FileImage()[:] + hv.DynamicMap(filtered_images, kdims=kdims).collate()

Unfortunately, that doesn't work, whether I call .collate() (see output below) or not. Of course, I can construct multiple DynamicMaps, using multiple callbacks, but they both need to share the same kernel_pattern state, so I'd also need to make a class to hold that state and change the callbacks to be methods on the object with the shared state. That's ugly, a lot of code, and a pain to set up. Can we just make .collate() work on DynamicMaps, and then people can write one callback to handle all the dynamic stuff in a given final Layout?

---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
<ipython-input-4-3c23b390e441> in <module>()
      3     return convolver.kernel_pattern[:] + FileImage(output_fns=[convolver])[:]
      4 
----> 5 FileImage()[:] + hv.DynamicMap(filtered_images, kdims=kdims).collate()

/Users/jbednar/holoviews_git/holoviews/core/spaces.pyc in collate(self, merge_type, drop, drop_constant)
    255         merge_type=merge_type if merge_type else self.__class__
    256         return Collator(self, merge_type=merge_type, drop=drop,
--> 257                         drop_constant=drop_constant)()
    258 
    259 

/Users/jbednar/holoviews_git/holoviews/core/element.pyc in __call__(self)
    592 
    593         components = ndmapping.values()
--> 594         accumulator = ndmapping.last.clone(components[0].data)
    595         for component in components:
    596             accumulator.update(component)

AttributeError: 'NoneType' object has no attribute 'clone'

[jlstevens]

A dynamic collate should be something we try to get ready for 1.7. I agree it is important and we can shift the assigned milestone if it turns out to be too tricky.

[philippjfr]

Now implemented, assigning to @jlstevens to document.

[jlstevens]

Now implemented, assigning to @jlstevens to document.

Will do!

[philippjfr]

I'll try to come up with an example.

[jbednar]

As a start, can you show how to make the above code work? I just tried with the latest master and the same code that didn't work above with or without .collate() still doesn't work with or without .collate(), and it still doesn't give a very useful error message. Are there now arguments that I need to pass to determine how the streams line up, and are there ways the error messages could guide me towards adding those?

[philippjfr]

This has now been implemented and I fixed a bug related to adding more plots to a collated Layout. Going to close this and will be adding some examples to our list of stream examples.

[github-actions]

This issue has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue for related bugs.