How to access params of a Tabulator that is dynamically updated

Param
1

I have a dynamically updated tabulator, and I don’t know how to access its selected rows. Hope the example is understandable.

import param
import panel as pn
import pandas as pd
import numpy as np
pn.extension("tabulator")

class DataProfile(param.Parameterized):

    _df = param.DataFrame(doc="Values as Pandas DataFrame")

    button = param.Event(label='new random numbers')

    def __init__(self, **params):
        super().__init__(**params)

    def table(self):
        table_ = pn.widgets.Tabulator(self._df, selectable='checkbox')
        pn.bind(table_.param.values, self._df)
        return table_

    @pn.depends("button", watch=True)
    def _update_dataframe(self):
        self._df = pd.DataFrame(dict(zip("xy",np.random.rand(2,10))))


app = DataProfile(
    _df=pd.DataFrame({
        "x": [0, 5, 10],
        "y": [0, 3, 10]
    }),
    )

tbl = pn.panel(app.table)
layout = pn.Column(
    tbl,
    app.param.button,
    ### none of these work
    # tbl._pane.selection,
    # tbl._pane.param.selection,
    # tbl.selection,
)
layout.servable()

Ideally, I would like to have inside the class.

If you find some other suggestions to the code, please feel free to comment too.

1 Like
2

Hi @szarma

If you include tbl.param.selection instead of tblselectionin theColumn` it should work for you.

This refactored example works for me.

import param
import panel as pn
import pandas as pd
import numpy as np
pn.extension("tabulator")

def get_random_data():
    return pd.DataFrame(dict(zip("xy",np.random.rand(2,10))))

class DataProfile(param.Parameterized):

    data = param.DataFrame(doc="Values as Pandas DataFrame")
    update_data = param.Event(label='New Random Numbers')

    def __init__(self, **params):
        super().__init__(**params)

    def create_table(self):
        table_ = pn.widgets.Tabulator.from_param(self.param.data, selectable='checkbox')
        return table_

    @pn.depends("update_data", watch=True)
    def _update_dataframe(self):
        self.data = get_random_data()


app = DataProfile(data=get_random_data())

tbl = app.create_table()
layout = pn.Column(
    tbl,
    app.param.update_data,
    tbl.param.selection,
)
layout.servable()

table-selection

3

Thanks, @Marc. This does work in deed.

Do you also happen to know is there a way to have the Tabulator inside the app somehow, as a parameter? I have the feeling it would offer a more natural interaction between the objects.

The way it is now, I need to go “outside” of the scope of the class, and then “back inside” to use its methods.

Let me illustrate

import param
import panel as pn
import pandas as pd
import numpy as np
import holoviews as hv
hv.extension('bokeh')
pn.extension("tabulator")

def get_random_data():
    return pd.DataFrame(dict(zip("xy",np.random.rand(2,10))))

class DataProfile(param.Parameterized):

    data = param.DataFrame(doc="Values as Pandas DataFrame")
    update_data = param.Event(label='New Random Numbers')

    def __init__(self, **params):
        super().__init__(**params)

    def create_table(self):
        table_ = pn.widgets.Tabulator.from_param(self.param.data, selectable='checkbox')
        return table_

    @pn.depends("update_data", watch=True)
    def _update_dataframe(self):
        self.data = get_random_data()
        
     ############# the only different part #############
    @pn.depends("data")
    def plot_selected(self,selection=None):
        # this is a method that I would like to use
        if selection is None or len(selection)==0:
            selection = self.data.index
        else:
            selection = list(selection)
        return hv.Points(self.data.loc[selection])


app = DataProfile(data=get_random_data())

# here I create the Tabulator outside of the DataProfile scope 
tbl = app.create_table()
layout = pn.Column(
    tbl,
    app.param.update_data,
    # here I'm back to using the DataProfiler method
    hv.DynamicMap(app.plot_selected, streams=[tbl.param.selection]),
)
layout.servable()

This is how my code looks like currently, but I feel like the holoviz developers have provided all these awesome and elegant solutions and bindings, that I am certain I am just missing something when I find myself needing to reach for what feels like a hacky implementation.

All in all, I am probably overthinking this :smiley:

4

I have the feeling Marc’s suggestion is close to the best you can get. It is true that panel can map param Parameters (Number, String, Boolean, Selector, etc.) to widgets and thanks to that you can evolve in param’s world to code your apps. Representing in a GUI interface like panel one of these simpler Parameters (e.g. a Number into a FloatSlider) is rather straightforward. But all the ways you can interact with a Dataframe mean that it would certainly not make sense to change param to support them, param isn’t only about GUI programming after all. So in these cases you have to go into panel’s world more. I use from_param quite a lot to do that, I find it a natural pattern to keep things in sync.

3 Likes
5

Thanks for providing the context. And for pointing out the from_param method. I somehow missed it while going through tutorials.

1 Like