Make it easy to set global opts defaults

[MarcSkovMadsen]

I wanted to document to the LLM in holoviz-mcp how to change the defaults to create publication ready plots.

I couldn't find a description anywhere for changing global defaults. What I could find is descriptions for specific plot types like opts.Bars.

You can also see a user asking in https://discourse.holoviz.org/t/set-default-plot-size-for-all-plot-types/5722. And even core developers struggle providing a uniform, non-hacky answer.

You can see the ways I found below. The most obvious one (method_1) is not working and raising an error though.

Please

• Explain to me here what the recommended way of updating global defaults is.
• Make it easier for users by either
  • Adding documentation describing how-to update global defaults
  • Optionally implementing functionality to make this easier.

"""Reference publication quality hvplot code following best practices."""

import hvsampledata
import hvplot.pandas  # noqa: F401
import panel as pn
from holoviews.plotting.bokeh import ElementPlot
from bokeh.models.formatters import NumeralTickFormatter
import holoviews as hv

def method_1():
    """Will raise exception: Exception: opts.defaults only accepts "backend" keyword argument"""
    hv.opts.defaults(
        autohide_toolbar=True,
        yformatter = NumeralTickFormatter(format='0a')
    )

def method_2():
    """Works. No autocompletion or tooltips. No good explanation why have to split across multiple calls."""
    hv.opts.defaults(autohide_toolbar=True,)
    hv.opts.defaults(yformatter = NumeralTickFormatter(format='0a'))

def method_3():
    """Works. Autocompletion or tooltips."""
    ElementPlot.autohide_toolbar = True
    ElementPlot.yformatter = NumeralTickFormatter(format='0a')

def method_4():
    """Works. No Autocompletion or tooltips."""
    ElementPlot.param.autohide_toolbar.default = True
    ElementPlot.param.yformatter.default = NumeralTickFormatter(format='0a')
    

def method_5():
    """Works. No autocompletion or tooltips."""
    ElementPlot.param.update(
        autohide_toolbar=True,
        yformatter = NumeralTickFormatter(format='0a')
    )

method_3() # Makes most sense to me.

# Extract
data = hvsampledata.earthquakes("pandas")

# Transform: Group by mag_class and count occurrences
mag_class_counts = data.groupby('mag_class').size().reset_index(name='counts')

# Plot: Use hvplot
plot = mag_class_counts.hvplot.bar(
    # DO define what to show on x axes
    x='mag_class',
    # DO define what to show on y axes. Can be single element or list for multiple
    y='counts',
    # DO add an explicit title
)
if pn.state.served: # Only run when served with `panel serve`
    pn.extension()

    pn.panel(plot, sizing_mode="stretch_both").servable()

[MarcSkovMadsen]

Below we have another example where I struggle setting backend_opts globally. I would like to have horizontal gridlines and some extra space outside of the axis titles because by default the axis titles are very close to the border of the plot.

I cannot figure out how to change this globally (method_1, method_2). But I can for Bars specifically (method_3).

"""Reference publication quality hvplot code following best practices."""

import hvsampledata
import hvplot.pandas  # noqa: F401
import panel as pn
from holoviews.plotting.bokeh import ElementPlot
from bokeh.models.formatters import NumeralTickFormatter
import holoviews as hv
from bokeh.themes import Theme

def method_1():
    """Does not work. The Bar plot does not take the backend_opts."""
    hv.opts.defaults(
        backend_opts={
        'plot.ygrid.grid_line_color': "black",
        'plot.ygrid.grid_line_alpha': 0.1,
        'plot.min_border_left': 80,            # Add padding on left (for y-axis label)
        'plot.min_border_bottom': 80,          # Add padding on bottom (for x-axis label)
        'plot.min_border_right': 30,           # Add padding on right
        'plot.min_border_top': 80,             # Add padding on top
    }
    )

def method_2():
    """Does not work. The theme of the Bar plot does not change. Don't know if this is a bokeh problem?"""
    hv.renderer('bokeh').theme = Theme(json={
        'attrs': {
            'Plot': {
                'ygrid': { 
                    'grid_line_color': "black",
                    'grid_line_alpha': 0.1,
                },
                'min_border_left': 80,            # Add padding on left (for y-axis label)
                'min_border_bottom': 80,          # Add padding on bottom (for x-axis label)
                'min_border_right': 30,           # Add padding on right
                'min_border_top': 80,             # Add padding on top
            },
        }
    })

def method_3():
    """Works. The Bar plot does take the backend_opts."""
    ElementPlot.backend_opts={
        'plot.ygrid.grid_line_color': "black",
        'plot.ygrid.grid_line_alpha': 0.1,
        'plot.min_border_left': 80,            # Add padding on left (for y-axis label)
        'plot.min_border_bottom': 80,          # Add padding on bottom (for x-axis label)
        'plot.min_border_right': 30,           # Add padding on right
        'plot.min_border_top': 80,             # Add padding on top
    }

method_3() # Only one that works

# Extract
data = hvsampledata.earthquakes("pandas")

# Transform: Group by mag_class and count occurrences
mag_class_counts = data.groupby('mag_class').size().reset_index(name='counts')

# Plot: Use hvplot
plot = mag_class_counts.hvplot.bar(
    # DO define what to show on x axes
    x='mag_class',
    # DO define what to show on y axes. Can be single element or list for multiple
    y='counts',
    # DO add an explicit title
)
if pn.state.served: # Only run when served with `panel serve`
    pn.extension()

    pn.panel(plot, sizing_mode="stretch_both").servable()