I wonder how many lines of code Mathematica 11 is now. Last one I know about is version 6.0 at 2.5 millions lines (I think this is for Kernel and front end but not sure).

I also wonder what percentage of Mathematica is written in Mathematica vs. C/C++. Last I remember seeing something about this, it was 50%

I have troubles with documentation center on v11, without some effort (which I haven't really sorted out in detail) many details show up as gray boxes, like disabled dynamics. This includes many components, like even some intra-document links and graphs. Any ideas what's going on? (I have made my best to enable dynamic evaluation.)

@kirma I didn't have that problem, but for me Manipulator didn't show up at all (only a red error box). After renaming my $UserBaseDirectory folder (so Mathematica creates a new one), everything worked fine. Maybe this resolves your problem too.

I suspect I don't have a particularly esoteric configuration, but I suspect this would be some sort of an upgrade issue. Definitely it wouldn't pass release QA if this was a common situation (...sure?).

@Karsten7. I have to check that out after I'm done at work... :)

@MichaelE2, idle question: have you already read Hairer/Nørsett/Wanner, by any chance?

I installed trial V 11 two days ago. Now I downloaded V11 from my portal (not the trial, actual official version). Then I un-installed the trial. Rebooted, then installed my official V11 (from portal download). But now when I start V11, it still tell me I am on trial, and when I try to activate using original official activation key (not trial one), it does not like it. How do I get rid of trial version? It still thinks my V11 is trial.

Ok, fixed it. Please ignore. I had to enter new activation code for V11. I was using V10.4 activation code. Sorry about this. All OK now. Trial banner is gone now.

@J.M. I've read some of vol I. I saw one of your references to it and checked it out of our library. I've been more or less using it as a reference reading a chapter here and a chapter there. (Vol II was either unavailable or missing.)

Don't assume Stern-Brocot sequence would nicely map into Hilbert curve and produce relatively uniformly distributed points... (first 32768 values here).

Hm. I can't believe I would have done it right.

Hum. I guess Stern-Brocot sequence doesn't really have the values very uniformly distributed, even if you look relatively deep to the tree.

@Karsten7. What you suggested wasn't enough alone(!), but I think it resolved part of the problem.

Or just on some documentation pages. WTF is this?

@MichaelE2 Tss, that's too bad about volume 2; if your library has access to SpringerLink, the electronic version might be accessible. (But volume 1 is definitely lovely, especially the pictures in their chapter on symplectic methods.)

How do I make sure PlotLegends uses the same marker size as the plot itself? Because right now for me it has a completely different size, even when using something like PlotMarkers -> Automatic. As an aside, is there a place to find some more documentation for PlotLegends? The built in documentation is super concise and doesn't even include how you can move it for example. I'd rather not have it outside my plot but somewhere inside.

How do I make sure PlotLegends uses the same marker size as the plot itself? Because right now for me it has a completely different size, even when using something like PlotMarkers -> Automatic. As an aside, is there a place to find some more documentation for PlotLegends? The built in documentation is super concise

Oops, sorry for the double post.. Can't seem to delete it

We can just disregard it altogether, using something like PlotLegends -> PointLegend[] works quite well.

Oh well. Gave up and sent a support request.

If I have two datasets x and y and I want to sort x in ascending order, how do I then make sure that y is sorted in the same order? I think it is something with SortBy but somehow I can't figure it out

@user3183724 There are many ways of doing this. The simplest I can think of is:

First join x and y together into list of {x,y} pairs.

The common trick for this is to use Transpose (which assumes that x and y of the same length). pairs = Transpose[{x,y}]

Now we can sort on pairs, but we want to sort based on the first value of each pair, which is x

SortBy[x, First]

You can extract the x values by "sortedPairs[[All,1]]" and the y values by "sortedPairs[[All,2]]"

Its a bit cumbersome to combine and then separate them again, but still, it works! Thanks :)

@J.M. I like the writing too. Clear, succinct, organized, and with a historical perspective. Geometric Numerical Integration is another nice one I've read a few chapters of, but with Lubich instead of Nørsett. Have you seen it? And Analysis by Its History literally just arrived in the mail.

@MichaelE2 Yes, those last two are also lovely. Someday I'll get my own copies instead of hogging the library's. :D

@user3183724 You've added extra information that wasn't there before. Your x and y aren't simple lists of numbers.

What are you sorting?

@Searke Much worse than that, I copy and pasted the wrong thing. data = Transpose[{x,y}]}. I can't believeI didn't even read what I wrote!

I don't understand your question. Is it similar to sorting numbers? Are x and y lists of numbers or something similar like dates?

@Searke No no, they are just numbers. The extra information I added was an accidental Ctrl V. Your first interpretation of the question was correct, that works. I use data because it is a dataset.

np. This isn't the most elegant way of doing it. But it's the way I would recommend first learning.

Oh. If you want to use a data set, then it becomes a different problem. I believe people should first learn how to manipulate lists. They're more fundamental.

As a simple example consider this dataset:

You can sort it they way you want by specifying the column to SortBy:

dataset[SortBy["a"]]

or

SortBy[dataset, "a"]

Hm, well I'm not explicitly using Dataset though, I didn't know that was a thing. I called it a dataset because I obtained it from an experiment is what I should have said

So I'm just using a list for now, or an array as I see it

Arrays are something else as well. You have to be careful. In many programming languages "list", "array", and "dataset" tend to have a definite and specific meaning

Hm. But the documentation of Array says that it generates a list. So you mean that by talking about an array, people would think I mean the function Array[]? That makes sense.

Well I mean. For example, new programmers will be quick to call an array a vector if they happen to be treating it as a vector in the math sense.

this is a problem when helping people taking their first C++ class.

To be fair to them, vector is a terrible name.

I see what you mean

Mathematica uses lists, and dataset is a function. And one shouldn't copy paste and not read what they pasted. Or work without enough sleep. That last one is debatable.

Thanks a lot for the help in any case; the fact that you chose the approach that is better for learning is much appreciated.

np. dataset isn't really a function. It's a datastructure, just like a list is or an Association.

@kirma There's structure... but maybe not the one you were expecting. :)

@J.M. Yep. I must admit I had a hunch this might not work, but I was surprised by the fact it was so far from an even distribution.

Anyone seen the new layout of a Dataset representation? Scrollbars and drill down to subelements. Pretty cool!

Examples in the Dataset have to be re-run to see this BTW

@SjoerdC.deVries Yes, mostly nice, but it has some annoying habits, such as not wanting to use scientific notation. It would rather write out all the zeros. Also, the imaginary units is formatted in a hard-to-read manner compared to v10 datasets. And finally: the scrollbar doesn't sync with the |< < > >| buttons at the bottom.

Try this: Dataset[Table[{x, x*10^-20, x + x/Sqrt[2] I, x + x/(Sqrt[2]*10^20) I}, {x, 0., 5}]]

@Szabolcs: i'm sure we can fix all those for 11.0.1 or 11.1.

@Szabolcs: the scientific notation thing is really about how much scientific notation vertically bloats a Dataset. especially if some rows have it and others don't it looks really ugly. but perhaps there can be a cutoff so that REALLY large things end up in scientific notation

@TaliesinBeynon Sounds great!

@Taliesin Actually I don't care about the sync problem, but I do care about the number formatting. Overall readability is much better than in v10 of course.

@Szabolcs: or maybe it can be a hidden option

in terms of options we need the ability to have them carried around in the Dataset itself

then you can create a Dataset that has some option setting in it and it'll format more to your liking

@TaliesinBeynon You mean that the power in $10^5$ makes the line too tall? How about another notation such as 1*^5 then? I guess 1e5 is not appropriate because it suggests that Mathematica supports this notation (which it doesn't)

yes, exactly

it may not be as bad as i'm claiming. let me experiment with enabling it and see how it looks

Nag nag. I think my answer is the only one that actually constructs the answer using Mma, instead of prior knowledge. I am disappoint of the lack of attention: mathematica.stackexchange.com/q/123138/3056 :I

Mostly wondering of those mysterious ways people vote.

@kirma I hadn't seen it. So, nobody got votes from me. :)

:~(

randomly changing nose. :)

;)

@Taliesin I imagine making the new scrollable dataset display perform well must have been a real challenge. BTW how is the scrollbar made? Did it require adding some new notebook elements? Is that element something we can use? I notice it tends to overlap other things in the notebook.

Voting is definitely a random process, but one never knows what exactly is driving it.

@Szabolcs: it's far from perfect. unfortunately i could only work on it very 'part time' because i was so occupied with getting neural networks finished

yes it was complex to get to that point

lots of abandoned approaches and sudden insights etc.

it uses attached cells, which unfortunately have a tendency to become 'detached' or overlap other stuff

we'll try to remove or fix that for 11.0.1 and just have the scrollbar sit directly to the left of the dataset

the unfortunate thing about that is it means the dataset shifts over to the right whenever it is big enough to warrant a scrollbar

@kirma likely the viewing is modellable as a Poisson process which when combined with the front-page likely makes things "interesting".

the attached cell does actually solve that problem, but at too high a cost in terms of strange failures

@rcollyer Yes, if you idealize (or banalize) behaviour of people. I have doubts there are other factors than them just throwing dice or flipping coin all the time. That's what makes life pretty hard...

@kirma maybe. I suspect there are several factors, in particular the effort required to understand the method (always an "interesting" factor) vs time available is skewed heavily in favor of voting for simple "clear" solutions.

Sure. :)

Then there's Leonid's theses. Always worth the read, and usually when I don't have time. :P

@kirma While the accepted answer is OK, this one actually does what is asked. The difference is clear from the plots. But it flew right under the radar.

@rcollyer Don't forget the "baby seal" factor -- those that make pretty pictures.

@MichaelE2 I always enjoyed "clubbing". (and yes sarcasm)

pretty pictures certainly helps.

The "how can I make this picture?" questions seem to get either +10 or -5 right away, and I don't understand what difference triggers the response.

good question. I think the early success of such questions helped spur them on, and then the inevitable backlash began.

The question @kirma answered wasn't even intended for the site. Probably lots of people skipped it. (I did.)

@MichaelE2 I was unsure of it all the time. Like fifty-fifty, does it belong here or not?

@kirma The first comment makes you wonder, and the OP doesn't even respond. (Or hasn't checked in. But I think one should check in after asking a question, assuming you can, to address any issues.)

@MichaelE2 I agree.

I guess the OP responded to some of the answers....but on first read, I don't think I got past the Q. Or it was before any answers. I don't remember now.

Also with V11 just coming out, and only checking the site every 3-4 hours, I miss stuff.

Oh well, maybe not worth worrying about too much. :)

There are enough things in the world to worry, mostly I want to keep my focus off all those worries. Otherwise... there's no end of it.

I'm just glad this is now easy to do with Mathematica:

(might be hard to 3D-print, tho)

@kirma admittedly, I had already done that demo back in version 4... :)

:)

@TaliesinBeynon The scroll bar size is a bit inconsistent as well. Sometimes it's as large as the data grid and sometimes like 50% of that. (see, e.g., this and this)

@SjoerdC.deVries Yes. Unfortunate, but not much can be done about that. The problem is that the amount of vertical space consumed by the grid is data-dependent. So as you scroll the displayed grid may grow taller or shorter. If we fix the size of the scrollbar to the size of the grid, it will literally oscillate because the relative y position as you drag it will change radically whenever the grid changes height, causing it to scroll again.

That's also why the scrollbar is on the left and not on the right: the width can change, which would displace the scrollbar out from under your mouse as you dragged it, which would be very confusing.

The best solution may be to drop the scrollbar all together, and have some other mechanism for scrolling, such as clicking and holding part of the Dataset and dragging the mouse. Then the delta can be computed from your first mouse click, irrespective of the size of the grid.

any thoughts welcome. it's a whole new world of Dataset navigation and i'm sure someone has some good ideas

like one thing i was prototyping was right clicking on a column header and getting auto-suggestions based on the type of that column, like 'Histogram', 'ListPlot', 'GeoHistogram', etc

also, be aware that with Dataset you can right an item in a dataset or a column header and get a bunch of options like copying to clipboard the position or the data itself. that's quite useful sometimes especially for heavily nested datasets

@Taliesin I must leave soon, but I want to grab the opportunity while you're here and ask: How stable are GeneralUtilities? It would be so useful to have a set of stable developer-oriented functions in Mathematica that are relatively safe to use in packages (i.e. low risk of breakage between versions).

@Szabolcs depends on what part, but some utilities are not going to change

@Szabolcs e.g. Scope, which I use a lot, will stabilize at this point.

and the whole macro mechanism

there's a lot in there so specific things you find useful that you want to know about go ahead and ask

Thanks! Are there any plans to make some of these semi-official? To find some reasonable middle ground between "fully supported" and "undocumented, therefore anything can happen"? There are many useful undocumented functions which haven't changed in many years. It would be nice to somehow indicate their de facto stable status, even if there's no support for them and no official guarantees.

agree

well a fair number of things that started out in GeneralUtilities became System symbols. like StringStartsQ, StringEndsQ, SubsetQ, etc

that's the ideal case of course

macros are too esoteric of course

but there are things like FileString (read a file into a string) that are rock stable because they're so obvious

if i had nothing else to do i would be happy to compile a list and then mark them stable in their actual usages

that's more likely to happen if someone has particular functions that would like to have a 'semi-official' promise of stability for, rather than the whole of GU, which is huge

change of topic. for anyone curious about neural networks, this may be interesting:

tensor graph abstraction that i'm working on as a stepping stone to supporting recurrent networks. that's a recurrent net unrolled to depth 4.

looked kind of neat and wanted to show it off :)

Thanks again, I might ping you about these in the future. I really must leave now or my wife will kill me. Waking up in 6 hrs to travel ...

k bye

@J.M. code or didn't happen

In due time, @Kuba. In due time… ;) But let me just say the code was surprisingly short.

@TaliesinBeynon Nice! I hadn't discovered that yet.