Exposing some options as enums?

[maximlt]

There are a number of keywords that accept a set of values (usually strings). In HoloViews, they are usually declared in a Selector Parameter, so they are more organized and more exposed than in hvPlot which doesn't use Param.

Here's a non-exhaustive list:

Customization:

• xaxis/yaxis: e.g. top, bottom, right, left
• legend: top, bottom, right, left
• tools: e.g. tap, box_select
• hover: vline, hline
• widget_location

Downsampling:

• aggregator: e.g. mean, count, min, max,
• downsample: e.g. lttb, minmax, m4

Geographic:

• coastline: e.g. 10m, 50m
• features: e.g. borders, lakes

I would like to discuss whether having them available as Enums in hvPlot would help improve the user experience. As suggested by @ahuang11, if we were to rely on Enums they should probably be first added to HoloViews. Here's the sort of change we could make in HoloViews, adding an Enum for legend_position and updating the Parameter in a way that doesn't change it at all.

# holoviews/plotting/enums.py
from enum import Enum
import param

class LEGEND_POSITION(str, Enum):
    TOP_RIGHT = 'top_right'
    TOP_LEFT = 'top_left'
    BOTTOM_LEFT = 'bottom_left'
    RIGHT = 'right'
    LEFT = 'left'
    TOP = 'top'
    BOTTOM = 'bottom'

# HoloViews plotting modules
class Element(param.Parameterized):
    legend_position = param.Selector(objects=[member.value for member in LEGEND_POSITION])

hvPlot could expose the Enums in its top-level module.

# hvplot.__init__.py
from hv.plotting.enums import LEGEND_POSITION as LEGEND

So users can leverage them easily:

# USER CODE
import hvplot.pandas
from bokeh.sampledata.penguins import data as df

df.hvplot.scatter(x='bill_length_mm', y='bill_depth_mm', by='species', legend=hvplot.LEGEND.TOP_RIGHT)