Bokeh甘特图样式堆叠水平条形图空白及JS报错问题

Hey there, let's figure out why your Gantt-style chart is showing up blank and throwing that JS error. The main issues are with how you're handling the data source and misusing the hbar_stack function for this use case. Here's the breakdown and fix:

What's Going Wrong #

1. You can't pass a GroupBy object directly to ColumnDataSource
   When you do source=ColumnDataSource(grouped), you're giving Bokeh a pandas GroupBy object instead of a structured data format it can understand (like a DataFrame or dictionary). This confuses Bokeh's JS rendering, leading to the "Unable to get property 'length' of undefined..." error.

2. hbar_stack isn't meant for Gantt charts
   hbar_stack is designed to stack numerical values from multiple series (e.g., adding up task durations per org). But a Gantt chart needs to show tasks as time-based segments on each org's row, not stacked values.

Fixed Code & Step-by-Step Explanation #

First, Prep Your Data #

Make sure your date columns are properly typed, then calculate task durations (or just use start/end dates directly):

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

# If you already have your df, skip this sample data setup
sample_data = [
    ["o1", "t1", "2019-11-19", "2019-11-23"],
    ["o1", "t2", "2019-11-23", "2019-11-27"],
    ["o1", "t3", "2019-11-27", "2019-12-01"],
    ["o1", "t4", "2019-12-02", "2019-12-06"],
    ["o2", "t1", "2019-11-20", "2019-11-24"],
    ["o2", "t2", "2019-11-24", "2019-11-28"],
    ["o2", "t3", "2019-12-01", "2019-12-05"],
    ["o2", "t4", "2019-12-05", "2019-12-09"],
    ["o3", "t1", "2019-11-20", "2019-11-24"],
    ["o3", "t2", "2019-11-24", "2019-11-28"],
    ["o3", "t3", "2019-12-01", "2019-12-05"],
    ["o3", "t4", "2019-12-05", "2019-12-09"],
]
df = pd.DataFrame(sample_data, columns=["Org", "Task", "Start", "Finish"])

# Convert date columns to datetime (critical for Bokeh's datetime axis)
df["Start"] = pd.to_datetime(df["Start"])
df["Finish"] = pd.to_datetime(df["Finish"])

Now Build the Gantt Chart #

We'll use hbar instead of hbar_stack, since we need to plot each task as a time segment on its org's row:

# Get unique orgs for the y-axis
orgs = df["Org"].unique().tolist()

# Create the figure
p = figure(
    title='Project Schedule',
    x_axis_type='datetime',
    y_range=FactorRange(factors=orgs),  # Ensures y-axis shows org names correctly
    width=800,
    height=450
)

# Plot each task as a horizontal bar using start/end dates
p.hbar(
    y='Org',
    left='Start',
    right='Finish',
    height=0.8,
    source=ColumnDataSource(df),
    legend_field='Task',
    alpha=0.7
)

# Tweak the chart for readability
p.legend.location = "top_right"
p.xaxis.axis_label = "Date"
p.yaxis.axis_label = "Organization"

# Show the chart
show(p)

If You Actually Want Stacked Bars (Not Gantt) #

If you meant to stack task durations per org (e.g., total time spent per org across tasks), here's how to use hbar_stack correctly by reshaping your data to wide format:

# Reshape data to wide format (one column per task)
wide_df = df.pivot(index='Org', columns='Task', values=(df["Finish"] - df["Start"]).dt.days).fillna(0)

# Create data source
source = ColumnDataSource(wide_df.reset_index())

# Build stacked bar chart
p = figure(
    title='Total Task Duration per Org',
    y_range=wide_df.index.tolist(),
    x_axis_label='Total Days',
    width=800,
    height=450
)

# Stack the task duration columns
p.hbar_stack(wide_df.columns.tolist(), y='Org', height=0.8, source=source)

show(p)

内容的提问来源于stack exchange，提问作者Wf19