A game presents itself: Starting from the number 1, what interesting places can you get to with the predictive interface alone? No typing! My first attempt below:

(With a few wrong turns on the way.)

@AndrewMoylan If the predictive interface can be accessed programmatically, you could get all possible outcomes using mma =)

@AndrewMoylan Is the number 1 Turing complete?

@P.Fonseca I can't (publicly) speculate which features will be in what future version. The customization of the predictive interface is a very interesting suggestion and having better support for spreadsheets is a very good one too.

@AndrewMoylan ... oh wow ... such a great idea!

@SjoerdC.deVries now that is not to say it is OK to perform all available tests and pick the smallest (or largest) p-value to report. If you run all tests, you should report all test results. As long as you report all of the test results and don't base your conclusions on a single one of them you are fine.

@P.Fonseca - You can turn on drag and drop in the option inspector, so maybe I don't fully understand that part of your suggestion?

Also, you can play games like can you get from X to Y?

@P.Fonseca (Ctrl-Shift-O, set scope to Global and then lookup the DragAndDrop setting)

(I have no idea.)

Also if Predictive Interface was programmable then maybe we could try to make something awesome like the new matlab import wizard ... which is the only thing I find mathematica misses for me now that we have legends, and bracket matching ;-)

@AndyRoss I know, but I'm sure there will be many researchers who will be cherry picking. Come to think of it: I don't recall seeing a publication where the authors reported the results of say 20 tests on the same measurements.

also the bracket highlighting in the notebook interface should be a front page feature ! can't count the number of times I have wanted to punch my monitor wishing for this feature outside of workbench

@Gabriel What does the import wizard do?

@SjoerdC.deVries mathworks.com/products/matlab/…

Sort of like the new predictive interface but does "smart" parsing of input files

in a gui centered way

it is awesome

@AndrewMoylan I can get to {"bar"} ...

@ArnoudBuzing No way! How so fast?

@ArnoudBuzing oh man can't wait for my download to finish to try my luck

@AndrewMoylan 1. Characters["foo"] , 2. StringJoin[Riffle[{"f", "o", "o"}, "XbarX"]], 3. StringSplit["fXbarXoXbarXo", "X"] , 4. Sort[{"f", "bar", "o", "bar", "o"}], 5. DeleteDuplicates[{"bar", "bar", "f", "o", "o"}], 6. StringJoin[Riffle[{"bar", "f", "o"}, " "]], 7. StringSplit[StringJoin[Riffle[{"bar", "f", "o"}, " "]], " f o"]

Ah with help of keyboard

That may be essential ... I get the feeling it is really hard with just mouse.

well, by clicking predictive interface buttons

XbarX is typed in right?

But some predictive interface options give you an input field

right

Right

Two levels of challenges!

So I think one way to "bar" without typing may be via random... If a sufficiently constrained random can be got to.

Example:

First get a lot of random letters. Then get random subsets of them. Then once you have a nice subset get random permutations ... Then finally you can get to "bar".

Maybe.

Ouch, Partition requires keyboard entry! This may be near impossible without keyboard.

@ArnoudBuzing what I mean with drag and drop is some sort of block diagram capacity, with snappy points (like the system modeler interface). There are a lot of application that would benefit from this type of interface (are completely not related with system modeler capacities), and nowaday, doing it with locators is not realistic.

So within a Graphics[...] output?

Something like that.

I don't know which would be the Mathematica way of doing it

Some sort of Dynamic thing that would allow to receive objects within its space. And these objects could have some properties like accepting connection from others, etc. All this managed with graph capacities.

I understand. I agree this would be very very cool to have in Mathematica.

But needs a very well thought out start

The spreadsheet thing has the same problematic well thought thing, to make it powerful. How can I create a spreadsheet environment after a paragraph of text, and then manipulate its content from within it, or from outside. Access to it like if it was a variable/db, and add code to its cells, like we do in Excel.

@P.Fonseca Send me an email (arnoudb@wolfram.com) if you're interested in being part of a prerelease test, should we embark on any of these suggested features.

I'll do :)

@Buzing While we are on the topic of future functionality, it would be great to have compiled and vectorized statistical distribution functionality. Right now many of the distributions are not compilable, and thus random number generation and pdf/cdf computation is not very fast.

@asim Random number generation is vectorized AFAIK

@Sjoerd. You cannot pass vectors of means and standard deviations. m={1, 2}, s={0.5, 0.6}, RandomVariate[NormalDistribution[m, s]] will not work.

Has anyone played around with the new Legends interface?

@Guillochon a little. it works and is not unspeakably ugly, that's the very least that can be said of it

@acl Not unspeakably ugly, well that's better than the previous iteration of it

@Guillochon exactly

@acl From the docs it still doesn't look very professional, IMHO

@asim This works OK for me:

@SjoerdC.deVries Of course MapThread can be used, but it is not the same thing as Listability. Also, Compilation to C would be very useful. If access to the Rmath standalone library is available via LibraryLink, that would work great.

@asim Your point was that it was slow. I don't consider two million random numbers from a given distribution in 1/7 of a second as slow/

Anyway your mileage may vary. SIgning off for today. Bye!

@Guilloch What would you like them to look like? It may be possible to change their style from the default to something more like what you wanted

@Sjoerd. Yes, that is fast, but you cannot use many of these functions within compiled functions. That creates a problem too. I should head out to..

If not, it's probably a good suggestion on how they can be improved.

@Searke Perhaps it's just the choice of examples on the what's new page, but everything just looks a bit cartoony. Probably just changing the fonts/color scheme/line styles would make things look better.

I kinda thought it looked a little D3/SVG. Maybe cartoony. I don't think the examples are intended to look like publication graphics.

it seems that the suggestions bar (the thing under output that gives you a set of choices on how to proceed) doesn't appear when I use PlotLegends -> Placed[{"log(x)"}, Right] (say)

@acl On what OS? I get the bar under Plot[Sin[x], {x, 0, 100}, PlotLegends -> Placed[{"log(x)"}, Right]] on Ubuntu

@Searke os x

Actually I do get the bar, but it has less options than that for Plot[Sin[x], {x, -5, 5}] say

the second has frame, axes, image size, plot style, more... while the one with Legends just has a rollup sign, the Alpha equals thing and a feedback button (and all but the last are greyed out)

Same here

The type of the result however is very different

which is probably why the bar is probably not sure what to do with it. It isn't a simple graphics statement

Finally got my activation key! Arnoud said the trial version that we downloaded is identical to the full one if properly activated, right?

It would be a good suggestion that it should be able to do the same things with this output that it does with the output without legends

@Szabolc Yes. It is the same.

With Mathematica 8, it was a important to make sure you removed the trial version and then install the full version. This was improved in 9.

Mathematica now turns on the discrete graphics on a Mac laptop as soon as it's launched. Does this happen for everyone?

@Searke hm good point, I hadn't noticed

v8 didn't do this

@Szabolcs what do you mean "discrete graphics"?

@rm-rf the nvidia card instead of using intel graphics

I had to google discrete graphics too. I imagine this is because using graphics cards has become more common, but I wasn't aware of this.

@Szabolcs Oh... I think I set mine to use nvidia all the time

I installed gfxCardStatus when I couldn't get CUDA to work with v8

@Szabolcs I could get CUDA to work... you had to make the change in the OS

@rm-rf Yes, I got it to work after I forced it to always have the nVidia card on