如何使用框图使用Bokeh绘制分类数据？

Question

I have a simple dataframe with 5 columns and 10 rows of data that I would like to display as a single diagram using boxplots.

df_times[:5]
    0   1   2   3   4
0   211.635771  45.411404   17.134416   20.062214   185.544333
1   234.500053  49.166657   17.052492   17.056290   205.531887
2   234.224389  49.342572   17.868082   15.981429   193.489293
3   221.880990  47.189842   17.054071   17.052869   198.318657
4   223.811611  49.991753   17.052005   17.590466   219.541593

The drawing code as follows:

colors        = ["red", "olive", "darkred", "goldenrod", "skyblue", "orange", "salmon"]
charts_times = figure(plot_width=900, plot_height=350, title='Query Runtime')
base, lower, upper, stds = [], [], [], []
for i, table_name in enumerate([x[26:] for x in tables_to_be_analyzed]):
    run_time = df_times[i]
    run_time_mean = run_time.mean()
    run_time_std = run_time.std()
    stds.append(run_time_std)
    lower.append(run_time_mean - run_time_std)
    upper.append(run_time_mean + run_time_std)
    base.append(table_name)
    color = colors[i % len(colors)]
    charts_times.circle(x=table_name, y=run_time, color=color, size=6)

charts_times.title.text = 'Query time ' + ' std: ' + str([round(x,1) for x in stds])
source_times = ColumnDataSource(data=dict(base=base, lower=lower, upper=upper))
charts_times.add_layout(
    Whisker(
        source=source_times, base="base", upper="upper", 
        lower="lower", line_width=1.5))
show(charts_times)

The only problem is that I am not sure how to have categorical data with boxplots because as it is bokeh just throws an error:

ERROR:bokeh.core.validation.check:E-1001 (BAD_COLUMN_NAME): Glyph refers to nonexistent column name: orig_parquet_0 [renderer: GlyphRenderer(id='f3703748-b7f9-43f6-807c-a8a24bdaab32', ...)]
ERROR:bokeh.core.validation.check:E-1001 (BAD_COLUMN_NAME): Glyph refers to nonexistent column name: test_orc_opt_3 [renderer: GlyphRenderer(id='b86d8724-ff66-45ac-8a09-eaa70cadf348', ...)]
ERROR:bokeh.core.validation.check:E-1001 (BAD_COLUMN_NAME): Glyph refers to nonexistent column name: test_orc_opt_4 [renderer: GlyphRenderer(id='ee5f644e-b334-4bce-8184-9f7d8da8bba1', ...)]
ERROR:bokeh.core.validation.check:E-1001 (BAD_COLUMN_NAME): Glyph refers to nonexistent column name: test_orc_opt_6 [renderer: GlyphRenderer(id='55eb8c19-344f-4010-b254-222330b76203', ...)]
ERROR:bokeh.core.validation.check:E-1001 (BAD_COLUMN_NAME): Glyph refers to nonexistent column name: test_parquet_opt_7 [renderer: GlyphRenderer(id='81ad57e4-4e01-4762-9a6e-0ce57cb51dd7', ...)]

Answer 1

You pass the name of a column for x :

charts_times.circle(x=table_name, ...)

But you have not created and passed any ColumnDataSource to as the source argument to circle . You need to create a CDS with column names like "orig_parquet_0"

Note that you cannot mix and match passing column names and literal lists to the same glyph at the same time, ie

p.circle(x="colname", y=[1,2,3,...], # not possible

If any of the data for a glyphs is referred to "by name" from a CDS, then all the data has to be in a CDS:

p.circle(x="xname", y="yname", source=source)

Alternatively, you can pass all the data as literal arrays or lists and not pass a source at all (you just can't mix and match):

p.circle(x=[1,2,..], y=[4,5,...])  # also OK