Should categorical bars / boxes always have a width of 1?

[bpostlethwaite]

http://codepen.io/bpostlethwaite/pen/eBvMWM

All data points except two have been filtered but the plot shows the full range of axis points and they are off center. I don't know if this is related to filtering or is more general but this is a way to trigger it.

cc @etpinard @alexcjohnson

[alexcjohnson]

that's a funny one, I think it's two bugs:

• the categories are being read from the original trace, pre-filtering, which doesn't seem like what we want
• then when we make bars only at a and f they each get a width of 5 categories - also probably not what we want, category axis bars should always have width 1 (one category), right? You can also trigger this second one with two overlaid traces:

Plotly.newPlot(gd,[
        {x:['a','b','c','d','e'], y:[1,2,3,4,5], type: 'bar'},
        {x:['a','e'], y:[4,2], type:'bar'}
    ],
    {barmode:'overlay'})

[n-riesco]

@alexcjohnson Forcing a width of 1 for category axis can be done by changing this line and this line.

[etpinard]

the categories are being read from the original trace, pre-filtering, which doesn't seem like what we want

Fixed in #1260

Changing issue title to reflect what's left of @bpostlethwaite 's bug report.

[alexcjohnson]

Should categorical bars always have a width of 1?

I think so - it's misleading otherwise, as otherwise a bar would bleed into neighboring categories, which in principle have no relation to each other.

[etpinard]

A similar problem happens for box traces on categorical position axes - see #4510 for more details.