Plotting in Bokeh using Custom Function? - python

does anyone know if/how one can use a "custom" function to plot in Bokeh using the Bokeh server? For example, I know you can use something like
plot = figure(toolbar_location=None)
plot.vbar(x='x', width=0.5, bottom=0, top='y', source=source)
But how can you plot using something like
def mplot(source):
p = pd.DataFrame()
p['aspects'] = source.data['x']
p['importance'] = source.data['y']
plot = Bar(p, values='importance', label='aspects', legend=False)
return plot
My current attempt is, here:
http://pastebin.com/7Zk9ampq
but it doesn't run. I'm not worried about getting the function "update_samples_or_dataset" working yet, just the initial plot to show. Any help would be much appreciated. Thanks!

Is this what you want? Note that I did not use the Bar function imported from bokeh.charts as this does not update upon updating the data source.
If you want to stick with using Bar from bokeh.charts you need to recreate the plot each time.
Note: to run this and have updating work - you need to execute bokeh serve --show plotfilename.py from the command line.
from bokeh.io import curdoc
from bokeh.layouts import layout
from bokeh.models.widgets import Button
from bokeh.plotting import ColumnDataSource, figure
import random
def bar_plot(fig, source):
fig.vbar(x='x', width=0.5, bottom=0,top='y',source=source, color="firebrick")
return fig
def update_data():
data = source.data
data['y'] = random.sample(range(0,10),len(data['y']))
source.data =data
button = Button(label="Press here to update data", button_type="success")
button.on_click(update_data)
data = {'x':[0,1,2,3],'y':[10,20,30,40]}
source = ColumnDataSource(data)
fig = figure(plot_width=650,
plot_height=500,
x_axis_label='x',
y_axis_label='y')
fig = bar_plot(fig, source)
layout = layout([[button,fig]])
curdoc().add_root(layout)
EDIT: See below a method that plots a bokeh plot but uses data from a dataframe as you wanted. It also will update the plot on each button press. Still you need to use the command bokeh serve --show plotfilename.py
from bokeh.io import curdoc
from bokeh.layouts import layout
from bokeh.models.widgets import Button
from bokeh.plotting import ColumnDataSource
from bokeh.charts import Bar
import random
import pandas as pd
def bar_plot(source):
df = pd.DataFrame(source.data)
fig = Bar(df, values='y', color="firebrick")
return fig
def update_data():
data = {'x':[0,1,2,3],'y':random.sample(range(0,10),4)}
source2 = ColumnDataSource(data)
newfig = bar_plot(source2)
layout.children[0].children[1] = newfig
button = Button(label="Press here to update data", button_type="success")
button.on_click(update_data)
data = {'x':[0,1,2,3],'y':[10,20,30,40]}
source = ColumnDataSource(data)
fig = bar_plot(source)
layout = layout([[button,fig]])
curdoc().add_root(layout)

I think you still have to attach your Bar instance to a Figure instance; a Figure is a set of plots, essentially, with niceties like the toolbar.

Related

Fix right-side of plot in place when zooming in/out

When I zoom out the following Bokeh plot:
A bunch of whitespace is added to both sides of the data:
How can I make it so that the right-hand side of the plot is fixed, so that zooming out will only create whitespace on the left-hand side? For example, this is my desired zoom-out look:
This is the code that initializes the plot:
from bokeh.plotting import figure, curdoc
from bokeh.driving import linear
from bokeh.models.tools import PanTool, WheelZoomTool
p = figure(sizing_mode="stretch_both", y_axis_location="right", x_axis_type="datetime")
pan_tool = p.select(dict(type=PanTool))
pan_tool.dimensions="width"
zoom_tool = p.select(dict(type=WheelZoomTool))
zoom_tool.dimensions="width"
This can be done with a CustomJS callback. Define a ColumnDataSource as source and look for the maximum x value on thof the source. If you have multiple renderers or multiple sources, you have to find the mximum of the maximums.
Minimal Example
from bokeh.plotting import show, figure, output_notebook
from bokeh.models import CustomJS, Line, ColumnDataSource
output_notebook()
p = figure(width=500, height=300)
source = ColumnDataSource(dict(x=data, y=data))
p.line(x='x', y='y', source=source)
callback = CustomJS(
args=dict(p=p, source=source),
code="""
p.x_range.end = Math.max(...source.data['x'])
p.change.emit();
"""
)
p.x_range.js_on_change("end", callback)
show(p)

Strange Labels on Bokeh Pie/Donut Chart

I'm following the response in this question: Adding labels in pie chart wedge in bokeh
I'm trying to add labels to my Bokeh chart so users can see data values. For some reason, the chart is rendering like this:
I'm not sure why this is happening. I tried commenting out the z variable to see if this made a difference.. it does not.
Here is my code:
import os
import pandas as pd
import pyodbc
from bokeh.plotting import figure, show
from bokeh.io import export_png
from bokeh.models import LabelSet, ColumnDataSource
from bokeh.palettes import Category20
import matplotlib as plt
from math import pi
from bokeh.transform import cumsum
lst = ['On_Time', 'All']
lst2 = [8, 85]
df = pd.DataFrame(list(zip(lst, lst2)),
columns =['Column', 'Value'])
df
df['angle'] = df['value']/df['value'].sum() * 2*pi
df['angle']
df['color'] = ['#084594', '#2171b5']
#z=110*(df['value']/df['value'].sum())
#df['value']=z
#df
p = figure(plot_height=350, title="", toolbar_location=None,
tools="", x_range=(-.5, .5))
p.annular_wedge(x=0, y=1, inner_radius=0.15, outer_radius=0.25, direction="anticlock",
start_angle=cumsum('angle', include_zero=True), end_angle=cumsum('angle'),
line_color="white", fill_color='color', legend='column', source=df)
df["value"] = df['value'].astype(str)
df["value"] = df["value"].str.pad(35, side = "left")
source = ColumnDataSource(df)
labels = LabelSet(x=0, y=1, text='value', level='glyph',
angle=cumsum('angle', include_zero=True), source=source, render_mode='canvas')
p.add_layout(labels)
p.axis.axis_label=None
p.axis.visible=False
p.grid.grid_line_color = None
show(p)
The labels are "underneath" because you have set
level='glyph'
in the call to LabelSet. There is generally not very many reasons to disturb the default render levels. If you remove this, the label will show up "on top" the way annotations such as LabelSet are intended to.
Also note the "str padding" that the other answer used to position the label alignment. That's slightly hacky but serviceable. Font differences between viewers' browser might make small differences. The alternative is to compute actual, exact x, y positions around the circle where you want the labels to go.

Bokeh skip tick labels for categorical data

I'm using Bokeh version 0.12.13.
I have a mixed numerical and categorical data. I only have one categorical data on the x-axis and the rest is numerical. I converted everything to categorical data to do the plotting (might not be the easiest way to achieve my goal). Now my x-axis tick labels are way denser than I need. I would like to space them out every 10th value so the labels are 10,20,...,90,rest
This is what I tried so far:
import pandas as pd
from bokeh.io import show
from bokeh.models import ColumnDataSource
from bokeh.plotting import figure
from bokeh.models.tickers import FixedTicker
# create mock data
n = [str(i) for i in np.arange(1,100)]
n.append('rest')
t = pd.DataFrame([n,list(np.random.randint(25,size=100))]).T
t.columns = ['Number','Value']
t.loc[t['Number']==100,'Number'] = 'Rest'
source = ColumnDataSource(t)
p = figure(plot_width=800, plot_height=400, title="",
x_range=t['Number'].tolist(),toolbar_location=None, tools="")
p.vbar(x='Number', top='Value', width=1, source=source,
line_color="white")
#p.xaxis.ticker = FixedTicker(ticks=[i for i in range(0,100,10)])
show(p)
Ideally, I would like the grid and the x-axis labels to appear every 10th value. Any help on how to get there would be greatly appreciated.
An easier way to do it is to keep the numerical data and use the xaxis.major_label_overrides. Here is the code:
import pandas as pd
from bokeh.io import show
from bokeh.models import ColumnDataSource
from bokeh.plotting import figure
from bokeh.models.tickers import FixedTicker
# create mock data
n = np.arange(1,101)
t = pd.DataFrame([n,list(np.random.randint(25,size=100))]).T
t.columns = ['Number','Value']
source = ColumnDataSource(t)
p = figure(plot_width=800, plot_height=400, title="",
toolbar_location=None, tools="")
p.vbar(x='Number', top='Value', width=1, source=source,
line_color="white")
p.xaxis.major_label_overrides = {100: 'Rest'}
show(p)
You can do this now (in Bokeh 2.2.3) using FuncTickFormatter:
# This prints out only every 10th tick label
p.axis.formatter = FuncTickFormatter(code="""
if (index % 10 == 0)
{
return tick;
}
else
{
return "";
}
""")
Sometimes you might want to do this instead of using numerical axis and major_label_overrides e.g. in a heatmap to get positioning of the content rects in the right place, or if you don't have numerical data at all but still want gaps in the axis labels.

How to create a multi-line plot title in bokeh?

How do you create a multiline plot title in bokeh?... same question as https://github.com/bokeh/bokeh/issues/994
Is this resolved yet?
import bokeh.plotting as plt
plt.output_file("test.html")
plt.text(x=[1,2,3], y = [0,0,0], text=['hello\nworld!', 'hello\nworld!', 'hello\nworld!'], angle = 0)
plt.show()
Additionally, can the title text string accept rich text?
In recent versions of Bokeh, labels and text glyphs can accept newlines in the text, and these will be rendered as expected. For multi-line titles, you will have to add explicit Title annotations for each line you want. Here is a complete example:
from bokeh.io import output_file, show
from bokeh.models import Title
from bokeh.plotting import figure
output_file("test.html")
p = figure(x_range=(0, 5))
p.text(x=[1,2,3], y = [0,0,0], text=['hello\nworld!', 'hello\nworld!', 'hello\nworld!'], angle = 0)
p.add_layout(Title(text="Sub-Title", text_font_style="italic"), 'above')
p.add_layout(Title(text="Title", text_font_size="16pt"), 'above')
show(p)
Which produces:
Note that you are limited to the standard "text properties" that Bokeh exposes, since the underlying HTML Canvas does not accept rich text. If you need something like that it might be possible with a custom extension
You can add a simple title to your plot with this:
from bokeh.plotting import figure, show, output_file
output_file("test.html")
p = figure(title="Your title")
p.text(x=[1,2,3], y = [0,0,0], text=['hello\nworld!', 'hello\nworld!', 'hello\nworld!'], angle = 0)
show(p)
Addendum
Here is a working example for plotting a pandas dataframe for you to copy/paste into a jupyter notebook. It's neither elegant nor pythonic. I got it a long time ago from various SO posts. Sorry, that I don't remember which ones anymore, so I can't cite them.
Code
# coding: utf-8
from bokeh.plotting import figure, show
from bokeh.io import output_notebook
import pandas as pd
import numpy as np
# Create some data
np_arr = np.array([[1,1,1], [2,2,2], [3,3,3], [4,4,4]])
pd_df = pd.DataFrame(data=np_arr)
pd_df
# Convert for multi-line plotting
data = [row[1].as_matrix() for row in pd_df.iterrows()]
num_lines = len(pd_df)
cols = [pd_df.columns.values] * num_lines
data
# Init bokeh output for jupyter notebook - Adjust this to your needs
output_notebook()
# Plot
p = figure(plot_width=600, plot_height=300)
p.multi_line(xs=cols, ys=data)
show(p)
Plot

Migrating code from Bokeh 0.10.0 to 0.11.0

I have the below example of real time streaming in Bokeh 0.10.0 from reddit thread.
import time
from random import shuffle
from bokeh.plotting import figure, output_server, cursession, show
# prepare output to server
output_server("animated_line")
p = figure(plot_width=400, plot_height=400)
p.line([1, 2, 3, 4, 5], [6, 7, 2, 4, 5], name='ex_line')
show(p)
# create some simple animation..
# first get our figure example data source
renderer = p.select(dict(name="ex_line"))
ds = renderer[0].data_source
while True:
# Update y data of the source object
shuffle(ds.data["y"])
# store the updated source on the server
cursession().store_objects(ds)
time.sleep(0.5)
I know that there is no cursession from 0.11.0 version. What would the code be like in Bokeh 0.11.0? Here is my attempt at it. Am I missing something? Basically, what I want the below code to do is to run as an app so that when I provide live streaming data, I can update the source and plot it realtime.
from bokeh.models import ColumnDataSource, HoverTool, HBox, VBoxForm
from bokeh.plotting import Figure, output_file, save
from bokeh.embed import file_html
from bokeh.models import DatetimeTickFormatter, HoverTool, PreText
from bokeh.io import curdoc
from bokeh.palettes import OrRd9, Greens9
plot = Figure(logo=None, plot_height=400, plot_width=700, title="",
tools=["resize,crosshair"])
source = ColumnDataSource(data=dict(x=[], y=[]))
plot.line([1,2,3], [10,20,30], source=source, legend='Price', line_width=1, line_color=OrRd9[0])
curdoc().add_root(HBox(plot, width=1100))
you probably want to add a periodic callback, something like:
def update():
ds.data["y"] = shuffle(y)
curdoc().add_periodic_callback(update, 500)
But also you actually need to put the data into the column data source, , and tell line the columns you want to use, instead of passing list literals to figure:
source = ColumnDataSource(data=dict(x=[1,2,3], y=[10,20,30]))
plot.line('x', 'y', source=source, legend='Price',
line_width=1, line_color=OrRd9[0])

Categories