Selection windows does not appear with quick box selection

Panel
1

Hi, I have a small app composed of a dataframe and a layout of 2 plots. The bottom plot is linked with a BoundsX stream, which extracts the range selected, add it to the dataframe and display a Vspan on the first plot.

In the end, the user would make several quick selection, but I have 2 problems.

  • If the user is too fast, the grey box for the selection doesn’t appear, It could be irritating for some user.

test_lattence6
test_lattence61112×614 42.1 KB

  • and if the process called by the stream takes too long, the script skip some selections.

My questions are :

Is there a way to dissociate the selection and the other process activated by param.depends, in order for the browser to keep the select box displayed ? (Maybe gather all selections and update once or parallelize each callback called by each selection)

Is there a minimum amount of time between two callback and is there a parameter to change it ? Or there is a way to manage more precisely the event buffer ?

I have posted those problems in panel, because It happens mostly when some panel view are refreshed by the stream.

the code :

import param
import pandas as pd
import hvplot.pandas
import holoviews as hv
from holoviews import streams
import panel as pn



class App_debug(param.Parameterized):
    list_sp=[]
    df_st=param.DataFrame(pd.DataFrame(columns=['beg','end']))
   

    def __init__(self,**params):
        super().__init__(**params)
        index=pd.date_range(start="2018-01-01",end="2018-12-31")
        self.plot=pd.DataFrame(data=range(len(index)),index=index,columns=['val']).val.hvplot(datashade=True).options(toolbar=None,default_tools=['wheel_zoom','reset'],active_tools=['box_select','wheel_zoom'],yaxis=None)
        rangex=streams.BoundsX(source=self.plot,boundsx=('2018-12-01','2018-12-02'))
        rangex.add_subscriber(self.recup_range)
        self.p_=pd.DataFrame(data=range(len(index)),index=index,columns=['val']).val.hvplot(datashade=True).options(toolbar=None,yaxis=None)
        self.p_.opts(alpha=0)
    
    @param.depends('df_st')
    def view_df(self):
        return hv.Table(self.df_st)
    
    
       
    @param.depends('df_st')
    def span(self,*events):
        sp=hv.Overlay(self.list_sp)
        
        return (self.p_*sp).opts(show_legend=False,yaxis=None,toolbar=None) 
    
    def recup_range(self,boundsx):
            b,e=boundsx
            df_=self.df_st
            df_.loc[-1]=[str(b),str(e)]
            df_.index=df_.index+1
            df_.sort_index(inplace=True)
            self.list_sp.append(hv.VSpan(pd.Timestamp(b),pd.Timestamp(e)).opts(color='#f89e9e',alpha=0.4))
            self.df_st=df_
            

    @param.depends()
    def view(self):
        return self.plot 
    
viewer=App_debug()
pn.Row(pn.Column(viewer.span,viewer.view),viewer.view_df).show()
1 Like
2

I don’t know anything in this area except that some widgets etc. have throttling parameters. I think that is the thing you should look for. My guess is that @philippjfr knows.

1 Like
3

Thanks
I finally found the throttling parameters. I’ve set :
streams.Stream._callbacks[‘bokeh’][hv.Streams.BoundsX].throttling_scheme=‘throttle’
streams.Stream._callbacks[‘bokeh’]['hv.Streams.BoundsX].trottle_timeout=0.1

and It’s seems to work, the server seems to not skip any input, I need to do more test to verify that.

I know there must be an easier syntax for setting those parameters, but for instance it does the job.

Although, the problem with the grey box is still there.

@philippjfr Is there a throttling parameter (or another mechanism) for RangeToolLink() ? I’ve search in the _callbacks attribute but I haven’t found it.

2 Likes
4

Great job @AurelienSciarra. Thanks for sharing.