Added implementation of var and std methods for ArrayBase

[kdubovikov]

I have found that ArrayBase is missing single-dimensional implementations for var and std. While you can use var_axis(Axis(0), 1.), it can become verbose and unnecessary in code with lots of single-dimensional arrays. I think sum_axis and mean_axis have sum and mean for this reason too.

Please consider merging this pull request.

[kdubovikov]

I see that the build fails at cargo clippy check un an unrelated file. Unsure if I need to fix that.

[bluss]

Thanks! Looks similar to #753 I don't mind whichever gets merged first

[ZuseZ4]

@jturner314 and @bluss I think it makes sense to merge this?
If we merge #848 first and rerun Travis here it shouldn't fail anymore. Just to be sure.

[kdubovikov]

@jturner314 , @bluss , @ZuseZ4 anything else I should do to help with the merge?

[bluss]

Looks great. I've rebased to current master so that tests will run properly with new CI. Given the commit history, I'd merge this with squash to clean it up a bit

[bluss]

Thanks!

[bluss]

cc @xd009642 Your comment would actually be interesting - w.r.t overlapping functionality with ndarray-stats. I don't think ndarray needs to include more of this, also unsure how we ended up here. Moving this method to ndarray-stats wouldn't be wrong IMO.

[xd009642]

Yeah I wouldn't be opposed to moving this functionality to ndarray-stats and then deprecating here

[bluss]

Oh so we ended up with var and std here - because we already had them in the axis variants here since long ago. It's hard to draw the line what's so basic that it could stay here - but from the summary functions I guess only .sum() is the so basic one, the rest could really move to stats.

[bluss]

I'll make a PR to improve the array traversal here - for performance. Fixed in 225d967