Checked out the JuliaPlots branch, and applied the pull request to StatsBase.
Non-uniform binning is no longer broken. (try, for example to histogram a normal distribution with the bins [-2.5, -2, -0.5, 0.1, 1.9, 2.5]. Using
histogram, the bars overlap.
With the new version I can just do
h = fit(Histogram, randn(10000), [-2.5, -2, -0.5, 0.1, 1.9, 2.5])
plot(h, seriestype=:steps) and everything looks as expected.
That's as easy as it should be. Admittedly this is not a particularly thorough test, but I hope this gets merged soon.