Error bars on bar traces should adhere to 'tozero' findExtreme option

[henrykeithturner]

Hi all,

I've been greatly enjoying using plotly but I've run into the following problem which I've been told may be a bug. When i have a couple of small bars on a bar chart, adding error bars (which do not extend to the negative region) causes a large whitespace between the xaxis and the xticks - as shown below with the accompanying script:

A = [10, 8, 12, 5, 8]
B = [0.5, 0.4, 10, 0.1, 0.4]
A_error = [1, 2, 1, 1, 2]
B_error = [0.2, 0.2, 2, 0.05, 0.2]
x_data = ['alpha', 'beta', 'gamma', 'delta', 'epsilon']


trace1 = (py.graph_objs.Bar(x = x_data, y = A,
                            error_y = dict(type = 'data', array = A_error, visible = True), 
                            hoverinfo = 'x + y', name = 'A', opacity = 0.8, hoverlabel = dict(font = dict(size = 20))))    

trace2 = (py.graph_objs.Bar(x = x_data, y = B,
                            error_y = dict(type = 'data', array = B_error, visible = True), 
                            hoverinfo = 'x + y', name = 'B', opacity = 0.8, hoverlabel = dict(font = dict(size = 20))))   

data = [trace1, trace2]


layout = py.graph_objs.Layout(

    autosize = False,
    width = 1700,
    height = 1000,
    margin = dict(b = 100),
    showlegend = True,
    legend=dict(x=1, y=0.85, font = dict(size = 20), bgcolor = 'rgba(0,0,0,0)', bordercolor = 'black', borderwidth = 1),
    title = ('tick test'),
    titlefont = dict(size = 40),
        xaxis=dict(
            title  = 'x_variable',
        ticks = 'outside',
        tickwidth = 2, 
        titlefont = dict(
            size = 25),    
        tickfont=dict(
            size=15,
        ),
            tickangle = 270,
            automargin = True,
        ),

    yaxis=dict(
            title  = ('y_data'),
        ticks = 'outside',
        tickwidth = 2,
        titlefont = dict(
            size = 25),
        tickfont=dict(
            size=15,

        ),
            hoverformat = '.2f',
            automargin = True,


        ),    

)


fig = py.graph_objs.Figure(data=data, layout=layout)
py.offline.iplot(fig) 

The whitespace dissappears when I remove orange bars. There are more details in the link I first made below

https://community.plot.ly/t/large-white-space-between-xaxis-and-xticks/13739/3

Thanks,

Harry T

[etpinard]

Hmm. This appears correct.

Error bar extreme values are computed with an opinionated padding corresponding the 5% of the axis length in px.

In your second trace, some error bar positions are close to zero (e.g. the at x=delta the error bar starts at y=0.05. So padding that y=0.05 extreme an extra 5% gets you an auto yaxis.range value in the negative. This negative starting range value is the "white space" you're referring to.

Now, I agree our default behavior isn't great for all situations. See issue #1876 for a list of planned auto-range-related improvements.

[alexcjohnson]

I think this can be considered a bug - as long as the negative sides of the errorbars don't reach past zero, we should give autorange a tight bound at 0. Likely this can be accomplished by having the errorbar calc inherit the tozero option bars pass to Axes.findExtremes:

plotly.js/src/traces/bar/cross_trace_calc.js

Line 533 in 2a3dc26

var extremes = Axes.findExtremes(sa, sRange, {tozero: true, padded: true});

which does exactly that: as long as the pre-padding lower bound is above zero (or the pre-padding upper bound is below zero) it adds a tight bound at zero, but the padding comes back as soon as you have explicit bounds on both sides of zero.

Note #3031 complicates this a bit... so the two should probably be tackled together.