Filtering Data

Filtering means showing a subset of your data: selecting points with the mouse, applying a range slider, or driving visibility from a dropdown. In Bokeh the pattern is the same everywhere - a ColumnDataSource holds the full data, and you update a second "view" source (or the CDS itself) with the filtered rows.

Selection-Driven Filtering

import pandas as pd
from bokeh.plotting import figure
from bokeh.models import ColumnDataSource, BoxSelectTool

df = pd.DataFrame({
    "x": [1, 2, 3, 4, 5, 6, 7, 8],
    "y": [5, 3, 6, 2, 7, 4, 8, 3],
})

source = ColumnDataSource(df)
p = figure(tools="pan,box_select,reset")
p.circle("x", "y", source=source, size=10)

# a second plot showing only the selected points
sel = figure(title="Selected")
sel_src = ColumnDataSource(data={"x": [], "y": []})
sel.circle("x", "y", source=sel_src, size=10)

box = p.select(type=BoxSelectTool)
def on_selection():
    idx = box.selected.indices
    if idx:
        sel_src.data = {"x": df["x"].iloc[idx].tolist(),
                        "y": df["y"].iloc[idx].tolist()}
box.on_select_event = on_selection

Range Slider Filtering

from bokeh.models import RangeSlider

slider = RangeSlider(start=0, end=100, value=(0, 100),
                     title="Value range", step=1)

def update(attr, old, new):
    lo, hi = new
    mask = (df["x"] >= lo) & (df["x"] <= hi)
    source.data = df[mask].to_dict()

slider.on_change("value", update)

Categorical Filtering (Dropdown)

from bokeh.models import Select

select = Select(value="A", options=["A", "B", "C"], title="Group")

def update(attr, old, new):
    source.data = df[df["grp"] == new].to_dict()

select.on_change("value", update)

Server-Side Filtering (Bokeh Server)

For large data, keep the full DataFrame in the server app and sync only the filtered rows to the browser CDS. The server callback pattern (covered in the Bokeh Server chapter) is the scalable version of the same idea.

Tips

• Filter by updating source.data (or a view CDS) rather than re-creating glyphs.
• Use boolean masks on the DataFrame; convert with .to_dict() for the CDS.
• Reset tools should restore the full dataset.

Summary

Filter Bokeh plots by updating a ColumnDataSource with a subset of the data: mouse selection via BoxSelectTool's selected indices, a RangeSlider for numeric bounds, or a Select for categorical values. For large data, run the filter server-side and sync only the visible rows to the browser.