Generalize the figure= keyword argument to all kind of plot

[Sigmun]

No description provided.

[Sigmun]

Should solve #86

[PatrikHlobil]

Hi @Sigmun, thanks for contributing. I will try to have a detailled look at the MR as soon as possible, I am quite busy now.

I already tried to implement this feature, but struggled how the API should be designed. There are a few critical issues that have to be settled:

1. Should we check the plotting type (line, bar, ...) and only allow certain combinations that make sense
2. How should we handle the case when the first plot has a different numerical range as the second (x/y range, ticks, etc)
3. Errorhandling in the code or indirect via Bokeh API

What do you think about that?

Best Patrik

[kkimdev]

Thanks for working on this! I just wanted to note that this is the default behavior of matplotlib backend, and has been very productive for me. We don't have to copy exactly, but maybe we can reference what matplotlib is doing in some aspects.

Regarding 1 & 2 : One option I think it makes is respecting the first plot.

fig = s1.plot(kind='line', show_figure=False)  #first
s2.plot(kind="bar", figure=fig)                #second

The first line is responsible for "creating" a figure, and second line is "using" an already created plot so I think such behavior is intuitive API wise as well.

[rraymondgh]

this looks like a very good change to me. without it the higher level API that pandas-bokeh offers removes a significant number of use cases. As well as working with matplotlib and plotly that offer this capability have also worked with geopandas explore()which provides similar type API to folium. effectively building up a figure from a sequence of layers

[PatrikHlobil]

@rraymondgh: As already mentioned I wanted this feature for nearly 1,5 years, but couldn't find the time to implement it. The problems are outlined above and the current code status is a bit messy, so in order to implement the feature, I would have to refactor the code quite a lot.

[algomaschine]

Gents, I'm sorry, still having issues figuring out how to plot 2 lines of different style in HTML.

Let's say here's an example. df would have different columns. I want to display one of it as normal line, but some others as discrete arrows (to show my signals) or circles or whatever, to mark the main line with some entries/exists. Also how can I make solid vertical line for xticks? They're kind of very pale. Thanks!

with open("osc_multimodel_forecast_plot.html" , "w") as f: html = df.plot_bokeh(kind='line', title="BB S&P", xticks=xticks, vertical_xlabel=True,figsize=(2000, 1000), show_figure=True,return_html=True, xlabel='date', ylabel='Price', hovertool=True, zooming=True,panning=True, legend='top_right') f.write(html+"\n") f.close()

[Sigmun]

Gents, I'm sorry, still having issues figuring out how to plot 2 lines of different style in HTML.

Let's say here's an example. df would have different columns. I want to display one of it as normal line, but some others as discrete arrows (to show my signals) or circles or whatever, to mark the main line with some entries/exists. Also how can I make solid vertical line for xticks? They're kind of very pale. Thanks!

with open("osc_multimodel_forecast_plot.html" , "w") as f: html = df.plot_bokeh(kind='line', title="BB S&P", xticks=xticks, vertical_xlabel=True,figsize=(2000, 1000), show_figure=True,return_html=True, xlabel='date', ylabel='Price', hovertool=True, zooming=True,panning=True, legend='top_right') f.write(html+"\n") f.close()

Hi @algomaschine, I don't see the link with this PR. you need to open an issue for your problem.