numbers as categories should equal their string representations

[alexcjohnson]

I would expect this to generate 3 categories (a, b, 1) but it generates 4 - two identical 1's:

Plotly.newPlot(gd,[{x:['a','b',1,'1'],y:[1,2,3,4]}],{xaxis:{type:'category'}})

[rreusser]

Also a good find! Note: ['a','b',1,'1'] and ['a','b','1',1] both fail in the same way.

[alexcjohnson]

@monfera @etpinard I missed this issue when category arrays went in, particularly this comment - I actually think we want string coercion, although this may open a can of worms re: lexicographic vs natural sorting. Thoughts?

[etpinard]

Yep agreed. 1 and '1' should be considered the same category - similar to how valType: 'number' attribute can be input as number of numeric strings.

[etpinard]

... but maybe @monfera has something to say about this.

[monfera]

Generally I'm not an advocate of automatic type coercions so if asked yesterday, I'd have said, expecting 4 things would seem natural. Coercing to strings will make it sort them as strings by default, e.g. "2" > "10". But based on the above discussion it looks like coercion would be the more natural fit for the library and I have no overriding argument of any kind.

[etpinard]

Found this. Will be fixed as part of #2489