Parameter Management

[jinglinpeng]

Is your feature request related to a problem? Please describe.
Currently the parameters are added casually in each function. The goal of this issue is to provide a unified approach to manage the parameters.

Describe the solution you'd like

We can have a configuration for each type of plot, and have a class named FigConfig to include all the configs. We have a default value for each config, and user could customize it by setter functions or a dict. For example,

# built by setter
fig_config = FigConfig()
                .set_hist(bins = 10, kde = True)
                .set_bar(show = False)

It means the 1) for hist: the bin number is 10, and the kde curve will be displayed with hist plot. 2) for bar chart: it does not show bar chart.

Similarly, user could build the config using a dict:

# built by dict.
fig_config = FigConfig({"hist.bins": 10,
                        "hist.show": True,
                        "bar.show": False})

We allow user customize plots by either input FigConfig or a dict:

plot(df, x, config = fig_config)

# when config is dict, we internally transform it into FigConfig.
plot(df, x, config = {"hist.bins": 10, 
                      "hist.show": True, 
                      "bar.show": False})

[jnwang]

I like this design. Two small comments.

1. Will there be any global parameter that applies to all plots in plot(df, x)? If so, can you add it to your design?
2. For the “how to guide” feature, we could show the config of each plot, e.g., {"hist.bins": 10, "hist.show": True} for histogram. Can this design support this feature?

[jinglinpeng]

@jnwang For the first question, yes. The global parameter includes height and width, and they can be set in a similar way: FigConfig.set_width(200).set_height(100) or FigConfig({"width": 200, "height": 100}). There is a question: whether we allow user to customize height and weight for each plot. If this is the case, we need to resolve the conflicts between global parameter and 'local' parameter.

For the second question, yes. We can display the important parameters for each plot, such as '{"hist.bins": 10, "hist.show": True}', and we show the dict in "how to guide". In this way user can easily customize the dict and pass it to plot.

[jinglinpeng]

Probably we should have two global parameters display and not_display to decide which to show or which not to show (at the same time only one of them is valid). By default they set as 'auto' and dataprep will show all the plots based on the function type (plot(df) or plot(df, x)). User can set display = ['wordcloud'] to show only wordcloud, or set not_display = ['spearman'] to avoid the computation of spearman correlation.