Can I update a view based on a change in a cell in a DataFrame?

Panel
1

What am I doing wrong in the code below? The view_stats does not update when I change values in the dataframe.

import param
import pandas as pd
import panel as pn
pn.extension('tabulator')

class Example(param.Parameterized):
    dataframe               = param.DataFrame(pd._testing.makeDataFrame())
    
    @param.depends('dataframe', watch=True, on_init=True)
    def _update_data(self):
        self.stats = self.dataframe.describe()
    
    @param.depends('stats')
    def view_stats(self):
        return pn.widgets.Tabulator(self.stats)
        
    def view(self):
        return pn.widgets.Tabulator(self.dataframe)

e = Example()

pn.Column(e.view_stats, e.view)
2

Try this: pn.Column(e.view_stats, pn.Param(e.param, widgets={"dataframe": pn.widgets.Tabulator}))

1 Like
3

Awesome, That works. But why does it work?

4

I think that when you call pn.widgets.Tabulator it is disconnected from the class, where pn.Param will keep it inside the class.

So the view could look something like this:

def view(self):
    return pn.Param(self.param, widgets={"dataframe": pn.widgets.Tabulator})

or this

def view(self):
    return pn.widgets.Tabulator.from_param(self.param.dataframe)

(or just moved up into __init__)

2 Likes
5

Yes I agree with @Hoxbro, @dharhas in your snippet there’s nothing that links the dataframe Parameter to its Tabulator widgets. The code is just rendering the dataframe when view is called.

There’s also another small issue with this line @param.depends('stats'). stats isn’t a Parameter, it’s just a regular Python instance attribute, so you can’t depend on it. Param should actually raises an error in this case (this is a small regression after some big improvement done by Philipp on handling dependencies of subparameters, there’s an open issue).

Here’s another approach:

import param
import pandas as pd
import panel as pn
pn.extension('tabulator')

class Example(param.Parameterized):

    stats = param.DataFrame()

    dataframe = param.DataFrame(pd._testing.makeDataFrame())
    
    @param.depends('dataframe', watch=True, on_init=True)
    def _update_data(self):
        self.stats = self.dataframe.describe()
            
    def view(self):
        return pn.Param(self.param, widgets=dict(dataframe=pn.widgets.Tabulator))
    
e = Example()

e.view()
1 Like
6

Ok doing this

import param
import pandas as pd
import panel as pn
pn.extension('tabulator')

class Example(param.Parameterized):

    stats = param.DataFrame()

    dataframe = param.DataFrame(pd._testing.makeDataFrame())
    
    @param.depends('dataframe', watch=True, on_init=True)
    def _update_data(self):
        self.stats = pd.DataFrame(self.dataframe.apply(sum, axis=1))
            
    def view_stats(self):
        return pn.widgets.Tabulator(self.stats)
                        
    def view(self):
        return pn.widgets.Tabulator.from_param(self.param.dataframe)
    
e = Example()

pn.Row(
    e.view, 
    e.view_stats
)

initially makes the tables overlap until I edit a cell.

image

after I edit a cell the tables move around and I get the following

image

Also is there a way to link the indexes?

7

hmm, maybe this is a bug with Tabulator and pn.Row:

image
image733×168 17.6 KB

8

I think it is the same as I reported here: https://github.com/holoviz/panel/issues/3028

Can you add your example as a comment?

9

done

1 Like