@MarianoSuárez-Alvarez unfortunately ConvexHullMesh[] is only built for < 3 dimensions.

@MarianoSuárez-Alvarez It shouldn't be to difficult to code up a quickhull implementation for n dimensions though.....

@MarianoSuárez-Alvarez In fact, no need to even write it yourself - the code from this page works perfectly - I am able to use Loren Petrich's code to generate convex hulls for arbitrary dimensions using the CHNQuickHull function. I can verify that in 2D and 3D it gives the same convex hull as Mathematica's own implementation

@MarianoSuárez-Alvarez If you have trouble making it work just ask it as a question on the main site

@JasonB this fails for unrelated reason :) {dataset}[[ ;; ]][[2]]

@Kuba lol, you're telling me list[[All,2]] is not the same as list[[All]][[2]]?!?!

@JasonB Yes, [[All, 2]] is "from each take the second" and [[All]][[2]] is "from All taken, take the second one"

Yeah, it took me too long of staring at it to realize that

@JasonB I'd say, maybe the original problem is expected to happen since Dataset is atomic, but so are Associations.

@Karsten7. Doesn't matter, 2 referes to the next level. Even with one association in assoList it will work while one dataset in datasetList isn't.

@Karsten7. assoList[[{1}]][[All, 2]]

@JasonB Length /@ datasetList gives {2} so [[;;, 2]] may be expected to work.

Right, but isn't it interesting that Dimensions@datasetList does not give {1,2}?

 Depth /@ assoList
 Depth /@ dataset
 Depth /@ datasetList

@JasonB I don;t think so, Dataset is atomic.

 Depth /@ assoList[[{1}]]

@Karsten7. yep Depth /@ datasetList is {1}, yet dataset[[2]] works so there is inconsistency somewhere.

They have different Heads.

@Karsten7. Do you agree there is something going on or not, because I don't know :) And please be more descriptive in your comments :P

@Kuba I think Part isn't implemented for Lists of Datasets. The developers properly don't even have Lists of Datasets on their radar to consider. You could ask support what functionality is supported for such structures or just contact them to put it on their radar.

You can presumably get some information about this from @Taliesin Beynon.

@Karsten7. my point is association[[1]] works and{association}[[;;,1]] works. Then dataset[[1]] works so one could expect {dataset}[[;;, 1]] to work too. I don't know how that's related to the fact that list of associations can be an input for Dataset, could you explain further?

I over simplified my example in this question as I mentioned in the comments to the answer. Do I edit the question or create a new one? mathematica.stackexchange.com/questions/115955

"I’ll be opening each event, and will take the opportunity to give some of the first public glimpses of Mathematica 11." ... I assume no Mma 11 then on quarterly schedule of releases which has been going on recently, and would sort of imply something coming out in couple of weeks? :I

@Kuba @Karsten7. My guess is that this behaviour is due to the fact that the kernel implementation of Part knows about lists and associations, but datasets are implemented in high-level code outside the kernel. {dataset}[[All, 2]] will be interpreted by the kernel function. Since the kernel is unaware of datasets, it just sees an atom. On the contrary, dataset[[2]] is fielded by an up-value on Dataset for Part which delegates to Dataset`DatasetPart.

Since an up-value cannot be triggered by the appearance of Dataset in deeper argument structure, it is hard to patch Part to handle the case without completely overriding the built-in definition -- a tricky prospect. Ideally, of course, datasets would be first-class data types for the kernel.

Does anybody know a simple way to get a .cdf file that corresponds to the look of printing a notebook with with "Printing Environment">"Working" i.e. colored local variables in Module, blue font for free Symbols etc.? That would be extremely handy for creating tutorial files.

Did anybody try to use RunProcess in subkernels (on Linux) and run Python 2 code with utf8 encoding? Somehow subkernels behave differently than the main one. BTW: since there is RLink, which installs an opensource R on demand, why is there no similar PythonLink? I mean, Python is even much more used than R, or?

Oh, and why is there no "wpm" (Wolfram package manager), like "npm" (github.com/npm/npm) ? I still remember MathSource (web.archive.org/web/20000303161553/http://www.mathsource.com), but there is no successor ( well, maybe the Wolfram Cloud could be ).

@RolfMertig Package managers have been discussed on and off; mfvonh made one and Leonid Shifrin has made a project installer that could be used for this purpose. When I created PackageData.net we talked about building a package manager for/around it. I'm still open to the idea, but how to make something that people want to use?