Plot legend shows value instead of text specified

On a Plot legend (bokeh 0.12.3), if the name for the legend is the same as the name for the column in the data source, the value will be displayed int he legend instead of the name. Is this expected behavior?

Example:

from bokeh import plotting
from bokeh.models import ColumnDataSource, Range1d
from bokeh.io import curdoc, show
import pandas as pd

data = [{"right1": 10, "right2": 14, "right3": 17,
         "right4": 22, "right5": 37, "right6": 44,
         "right7": 49, "right8": 59}]

source = pd.DataFrame(data)
source = ColumnDataSource(source)

fig = plotting.figure(width=1200, height=900)

fig.hbar(y=1, right="right1", left=0, height=0.5, fill_color="blue", source=source, legend="right1")
fig.hbar(y=1, left="right1", right="right2", height=0.5, fill_color="green",
         line_color="black", source=source, legend="right2")
fig.hbar(y=1, left="right2", right="right3", height=0.5, fill_color="yellow",
         line_color="black", source=source, legend="right3")
fig.hbar(y=1, left="right3", right="right4", height=0.5, fill_color="orange",
         line_color="black", source=source, legend="right4")
fig.hbar(
    y=1, left="right4", right="right5", height=0.5, fill_color="red", line_color="black", source=source,
    legend="right 5")
fig.hbar(
    y=1, left="right5", right="right6", height=0.5, fill_color="purple", line_color="black", source=source,
    legend="right 6")
fig.hbar(y=1, left="right6", right="right7", height=0.5,
         fill_color="grey", line_color="black", source=source, legend="7.")
fig.hbar(
    y=1, left="right7", right="right8", height=0.5, color="black", line_color="black", source=source, legend="*8.",
    fill_alpha=.12, line_alpha=.12)

curdoc().add_root(fig)

show(fig)

As you can see, in the legend for the first 4 hbar(s), the value for that hbar is displayed instead of the text specified in legend=

[bryevdv]

@Rubyj Yes, this is a new feature that allows legends to be automatically generated by grouping values in a column data source for a single glyph. In conjunction with a categorical color mapper this can often greatly reduce code needed. If you don’t want this behaviour, there are two options:

• change the column names to something else, so as not to match the legend string value
• be explicit that you don’t want the legend to look at field names: i.e passlegend=dict(value="foo") to the glyph function

ping @birdsarah this is the second time today this question has come up. Perhaps we need better documentation support

[foobarbecue]

The way to make this work is now:

from bokeh.core.properties import value
legend=value("Sunset")

because bokeh.properties has become bokeh.core.properties .

I have to say, I’m a new bokeh user and this legend thing was a major speedbump that cost me about an hour. Couldn’t find the answer until I came across it here.