Bokeh - Do not show tooltip if it has missing valu

问题:

I'm working on a bokeh figure that shows cluster activity. When a user hovers over a particular processor, I want it to show statistics about the processor. Heres the code:

TOOLTIPS = [
    ("Usage", "@{usage}%"),
    ("Name", "@name"),
    ("PID", "@pid"),
    ("Command", "@command"),
    ("User", "@user"),
]

p = figure(title="Cluster Activity",
           plot_width=1200,
           plot_height=700,
           x_range=nodes,
           y_range=list(reversed(cores)),
           tools='hover',
           toolbar_location=None,
           tooltips=TOOLTIPS
           )

This works, but I don't want to show tooltips with a value of None. For example, if a particular processor, has a None value for User, the tooltip should not contain a user value, rather than showing "User : ???".

Is there any way to do this? I can't seem to find anything similar to this in the tutorials. I'd like to avoid writing custom JS.

回答1:

You can also create the tooltips dynamically using JS callback attached to the HoverTool (Bokeh 1.1.0)

from bokeh.plotting import figure, show
from bokeh.models import ColumnDataSource, HoverTool, CustomJS

pid = [1, 2, 3, 4, 5, 6]
user = ['user1', 'user2', 'user3', 'user4', None, 'user6']
name = ['name', 'name2', 'name3', 'name4', 'name5', 'name6']

source = ColumnDataSource(data = dict(pid = pid, user = user, name = name))

p = figure(x_range = FactorRange(*name), sizing_mode = 'stretch_both', title = "Test", toolbar_location = None, tools = "")
p.vbar(x = 'name', top = 'pid', width = 0.2, source = source)

code = '''  hover.tooltips = [["Name", "@name"], ["PID", "@pid"]];
            if (cb_data.index.indices.length > 0) { 
                index = cb_data.index.indices[0];
                counts = source.data.user[index]

                if (counts != null)
                    hover.tooltips = [["Name", "@name"], ["User", "@user"], ["PID", "@pid"]];                                       

            } '''
hover = HoverTool()
hover.callback = CustomJS(args = dict(source = source, hover = hover), code = code)
p.add_tools(hover)

show(p)

Result:

回答2:

I see two ways of doing this:

1. Checking if Name is None with Python and using multiple HoverTool

Since HoverTool is a bokeh.models.tools you can add it via

p.add_tools(hovertool)

So you could make two instances of HoverTool and split your data to two data sources:

p = figure(title="Cluster Activity",
           plot_width=1200,
           plot_height=700,
           toolbar_location=None)

without = p.square(name="without", ##your filtered data source without names)
with = p.square(name="with", ##your filtered data source with names)

hoverwith = HoverTool(names=["with"],tooltips=TOOLTIPS = [
        ("Usage", "@{usage}%"),
        ("Name", "@name"),
        ("PID", "@pid"),
        ("Command", "@command"),
        ("User", "@user"),
    ])

hoverwithout = HoverTool(names=["without"],tooltips=TOOLTIPS = [
    ("Usage", "@{usage}%"),
    ("PID", "@pid"),
    ("Command", "@command"),
    ("User", "@user"),
])


p.add_tools(hoverwith, hoverwithout)

With the names attribute of HoverTool you can specify for which glyps the hover is rendered. I haven't tested the code.

2. Using custom JS (just mentioning for the sake of completeness)

In case you have many different combinations of possible missing values, I only see JS as a way to do this, have a look here: https://groups.google.com/a/continuum.io/forum/#!msg/bokeh/4VxEbPaLqnA/-qYLDsbZAwAJ