Standardize outputs for plot functions

[danielolsen]

Currently, we have a mix of:

• showing a plot
• returning a figure

It would be nice to standardize to returning the figure by default, and allow the user to specify different behavior via optional keyword arguments (e.g. showing the figure, saving the figure, returning nothing, etc.) that are also standardized across our plotting functions.

[BainanXia]

Recall what we have discussed during the standup meeting:
I would suggest to return the axes object anyways and show the figure as well.

• This design is aligned with the design of pandas plot module.
• Returning the axes will give the user maximum flexibility of modifying the figure layout and the user could choose to capture it or not.
• Having axes returned, one could get the figure object by calling ax.get_figure() if he/she would like to modify figure level attributes.
• There are multiple ways to suppress figure from showing in any interactive environment such as Jupyter notebook: https://stackoverflow.com/questions/18717877/prevent-plot-from-showing-in-jupyter-notebook.
• With the figure shown by default will relax the requirement of importing matplotlib.pyplot from the user end.

[danielolsen]

So the standard call signature for our plot functions would be:

def plot_foo(*arg, **kwargs, show_plot=True):
    ...
    if show_plot:
        plt.show()
    return ax

?

[BainanXia]

So the standard call signature for our plot functions would be:

def plot_foo(*arg, **kwargs, plot=True):
    ...
    if plot:
        plt.show()
    return ax

?

We can remove that optional parameter: show_plot since some of the plots are generated using pandas plot function which is shown by default. We can leave it to the user if one would like to mute figure shows.

[danielolsen]

I think it would be good in all functions to give the user the option to not have plot windows showing up, regardless of the details under the hood (pandas vs. matplotlib directly).

[rouille]

I think it would be good in all functions to give the user the option to not have plot windows showing up, regardless of the details under the hood (pandas vs. matplotlib directly).

Does this mean that we need to suppress the showing when the plot is generated via the pandas.DataFrame.plot function?

[BainanXia]

I think it would be good in all functions to give the user the option to not have plot windows showing up, regardless of the details under the hood (pandas vs. matplotlib directly).

Does this mean that we need to suppress the showing when the plot is generated via the pandas.DataFrame.plot function?

I believe so.

[kasparm]

Lets make sure all the plotting functions follow the pattern.