Avoid setting batched before the class is initialized

[maximlt]

@hoxbro feel free to reject this PR since it's not a fix per say.

What used to happen is showed in the example below, where the batched Parameter of GenericOverlayPlot was set (self.batched = batched) before Parameterized.__init__ was called. Setting a Parameter value before a Parameterized class is initialized (at the Param sense) has been supported somehow in Param for a long time. holoviz/param#766 broke that support temporarily, while holoviz/param#790 is an attempt at restoring it before Param 2.0 is released. Breaking that support temporarily allowed me to spot that this batched case was the only occurrence of a Parameter value being set early across a few code bases (Panel, Lumen, HoloViews). Because of that, and because it feels to me like an anti-pattern, I was tempted to open this PR to avoid this,

class GenericElementPlot(param.Parameterized):

    ....

    def __init__(self, batched=False, **params):
        self.batched = batched
        super().__init__(**params)

class GenericOverlayPlot(GenericElementPlot):

    ...

    batched = param.Boolean()

    def __init__(self, batched=True, **params):
        super().__init__(batched=batched, **params)

# When GenericOverlayPlot is instantiated `batched` is set before `Parameterized.__init__` is called.
GenericOverlayPlot()

[hoxbro]

Could we do something like this;

class GenericElementPlot(param.Parameterized):
    batched = param.Boolean(default=False)

    def __init__(self, **params):
        if param.get("batched"):
            ...
        super().__init__(**params)


class GenericOverlayPlot(GenericElementPlot):
    batched = param.Boolean(default=True)

    def __init__(self, **params):
        super().__init__(**params)

I also think it is an anti-pattern, and I will maybe go even further and say it is an anti-pattern that batched sometimes is a parameter, and sometimes it is not.

[codecov-commenter]

Codecov Report

Merging #5814 (1b5acf1) into main (003be41) will decrease coverage by 0.03%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##             main    #5814      +/-   ##
==========================================
- Coverage   88.17%   88.15%   -0.03%     
==========================================
  Files         307      307              
  Lines       62916    62916              
==========================================
- Hits        55479    55466      -13     
- Misses       7437     7450      +13     

Flag	Coverage Δ
ui-tests	?

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
holoviews/plotting/plot.py	93.51% <100.00%> (ø)

... and 2 files with indirect coverage changes

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[maximlt]

I also thought about adding batched as a Parameter to GenericElementPlot. Then I think I got confused by self.batched = batched being in between of other non Parameter attributes being set like zorder, cyclic_index or overlay_dims and wondered if it made sense for batched to be made more "officially" an input of GenericElementPlot while it seems to be dedicated to handle specifically NdOverlays?

[jbednar]

I also thought about adding batched as a Parameter to GenericElementPlot.

I agree that it's very strange to have GenericElementPlot accept and respect batched without batched being a Parameter on GenericElementPlot, only its subclass. Seems very fishy and it's hard to see any justification for doing that. I agree it's truly bizarre that generally GEP has only a batched attribute except that sometimes it has a batched Parameter (if it's also a GenericOverlayPlot).

[github-actions]

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