@halirutan for analyzing large CSVs, there's no comparison. A 250MB timestamped numeric file comes into MATLAB with a single readtable call in about 5 seconds and I end up with a table that has nice named columns and very fast operations against them, using about 210MB of RAM. MMA Imports the same file at about 750MB, which I can get down to the same 210MB after an N[] and ToPackedArray[]. But, to get it as a Dataset instead takes up 4.25GB of RAM, and the queries are comparatively slow.

smaller files work quite well, and I use MMA for those, but I look forward to the day I can slurp in a 1GB file to MMA with the same efficiency I get from MATLAB or R!

@kirkus Exactly this is my point. This has nothing to do with what people complaining about (at least in my case). We have large NIFI files (and here I mean several GB sometimes) and we need external importers anyway. But once the data is loaded, it's basically a n*x*y*z matrix of values and everything you do are operations on the numerical data. People think that Mathematica is bad in number-crunching the data which is just not true.

And yes, I agree, the CVS import of Mathematica leaves a lot of things to wish for.

@halirutan yep, my comment wasn't totally related to your issue, but I felt like venting anyways - thanks for listening! I truly do dream about the day where I only need to have two apps open during the day, a web browser and Mathematica.

If you try evaluating 2*(1/2)^Binomial[2^k 2^k, 2] with the suggestion bar enabled, does it cause weird behavior for anyone else? From a fresh start, the first time I evaluate it it doesn't show the suggestion bar. If I try it again, it blows up memory usage.