("Science of Interstellar" it is...)

Just hoping Mma has no hard-to-spot bugs in symbolics, considering how much theoretical physicists tend to rely on it... ;)

Any other Mac users experiencing frequent (1/day) front end crashes with v12.0? I'm happily typing, then poof

just sent a note to support with the crash details

@ChrisK I can't remember when I would have seen crash last time, and even then they have been "expected."

I seem to recall a question about this (can't find it), so I must not be the only one. Anyhow, I'm sure Wolfram will figure it out sooner or later

@kirma Can you give some more context?

What is this paper, why is it on your screen?

I know Interstellar and I love the movie, the music and everything. It's a movie I always watch alone though.. you know, being a father changed a lot of movies for me. So I usually cut some onions when I watch this kind of stuff.

@halirutan A short flash on "Science of Interstellar" documentary. I assume it was some sort of a draft to back up movie script. Well, it features quite a bit of Kip Thorne, certainly a Mma user.

@kirma Ah.. alright. Just rewatched the scene where Cooper views all his messages after Miller's planet. This always gets me.

Maybe I should watch the movie again. Today it's election and ice hockey, though! (I'm not particularly thrilled of ice hockey, but more than the election, though.;)

Was there ever a discussion about adding "low effort question" to the close reasons? I'm not really advocating for it, but I am curious.

@kirma You've probably seen it, but there are a few blog posts on the WRI/SW blog about Interstellar that are entertaining reads.

There are also two or three long videos about the behind-the-scenes of making up the language in Mathematica somewhere.

@CarlLange Probably I did read them, and don't quite remember them any more after couple years. :)

(going through the actual notebooks for the language generation and stuff, which I thought was quite cool)

Maybe I should take a look at it.

Trying to find them now to link to

Ah, what am I talking about, I meant Arrival

Also a worthwhile film

Heh. Almost the same! ;)

I remember Wolfram involvement on that, and actually purchased it on Apple TV recently. (Mostly to add up on collection of Dolby Vision test movies, I did watch it in the theatre too, which is pretty rare for me.)

@anyone want an asymptotically faster Mathematica?

@CarlLange You're probably mixing up Arrival with Interstellar.

@halirutan As he noted. :)

Ahh.. I should read to the end next time.. You already noted it.

Given that WL/Jupyter is probably gonna become an ever larger portion Mathematica programming it might be time to revisit the ToSVG stuff I was working on to make it more robust and complete...

Jupyter can (probably) embed SVG naturally and so if we can get good taggable SVG export that could enable some minor Dynamic capability

Add that on to a re-imp of the Jupyter link that's done with a MathLink connection and we'd have a pretty good FE.

I figure I can probably do both of those as I have my Python MathLink connection and of course that SVG stuff.

On the other hand I'm just not feeling the inspiration to do it...

@b3m2a1 The whole Jupyter affair basically demands a larger planning. I already noted some days ago that I'm not happy to use a Jupyter kernel where the communication between WL kernel and Jupyter is implemented by the same kernel used for the calculations.

@halirutan exactly that's why I figured I could use my Python/MathLink stuff and use python as the manager language

It's hard to debug and I'm not sure I trust its stability. As already said by others, a simple Dialog[] crashes the kernel.

The thing already can be run in a Jupyter notebook (using python) and now finally compiles on Linux, Windows, and Mac

@b3m2a1 There is already a Jupyter kernel for WL written in python.

What's wrong with that one?

@b3m2a1 I didn't try it and Szabolcs couldn't get it running.

Ah I see. Do you still have the link? I could take a look and see if I can figure out why when I feel inspired.

@b3m2a1 I already said to CE that it's sad we don't seem to have any Java devs here. If the network layer of the kernel was in Java, I would probably join and work on things like autocompletion and syntax highlighting.

Because no one really mentioned that this is missing too.

Basically, my idea is the following: we have Java code for the LinkSnooper which already shows how main, service and pre link are working. The Jupyter Kernel (JK) also should have several links.

Ah. Yeah I can see why that'd be useful...

For one, Jupyter expects a "heartbeat" from the kernel. Second, we need to think about what to do with Graphics, Graphics3D and other fancy stuff.

Mathics used real 3D graphics for plots if I recall correctly.

Right. That's where I was imagining just going with the SVG for Graphics although for 3D graphics a legit library makes more sense.

I wonder what they use on the Cloud...

@b3m2a1 Well, they have the sources of the front end and in the end, it's just OpenGL or DirectX.

So I'm pretty sure, they converted it into a JS lib that can show 3D graphics.

Yeah I was more wondering which JS lib they used

Only WRI knows.

Although I'm pretty sure that Jan (ex dev of Mathics, now kernel dev at WRI) could answer this.

If Jan's WolframNotebookEmbedder stuff reaches sufficient maturity to embed arbitrary stuff we could also use that to embed more complex stuff like this.

But I emailed him a bit ago and currently it needs to be hosted in the Cloud. If there were a free upload API that'd be fine since we could upload then embed, but I don't think there is...

I feel like embedding WRI stuff directly for complex content is probably the most robust approach, on the other hand writing an exporter for simple 3D stuff to a JS 3D lib is something I've wanted for my blogging stuff for a while.

@halirutan is there any way it would be realistic for someone like myself to port your very nice syntax parsing/autocomplete stuff for IntelliJ to python? Like how much code is there? I ask because looking at the Jupyter docs it looks like writing a python kernel that acts as a client is by far the simplest thing to do in Jupyter.

And since I can already go from Mathematica to python... well it feels like it makes sense.

@b3m2a1 For the advanced stuff, it is quite some work and I wouldn't consider it. The reason is that it's not only my algorithms for e.g. inspecting local structures (Module, Table). There is a powerful framework behind it for handling and working with AST.

I see. I probably wouldn't need to implement a full AST package since python already has one, but I'm sure there'd be a lot of stuff that isn't already covered to handle.

@b3m2a1 To make even a bit advanced stuff, you need a full parser.

@b3m2a1 You should checkout PyCharm jetbrains.com/pycharm (or use IntelliJ w/ the Python plugin)... by far the best IDE for python. @halirutan's plugin also installs on PyCharm, so you can have a project where you flip between mma and python and get autocomplete for both :)

@rm-rf I already use PyCharm :)

Well Intelli J with the python plugin...

I'm thinking entirely about doing this kind of stuff in Jupyter for other people, not myself.

I use PyCharm for python and my mini-IDE in Mathematica for my Mathematica development

Ah, ok. Get them to use PyCharm pro for jupyter support :P

I figure this kind of thing is good Mathematica community out-reach for some day when I'm in a generous mood and don't want to work.

Oh, wait... you were asking about Mathematica autocomplete within Jupyter? Sorry, I totally misread 🤦

@rm-rf Yes, basically we are talking about a Jupyter kernel for the new free Wolfram Engine

So many new things.... I suppose that's what happens when I drop into chat once a year :D

@rm-rf You are now using Python exclusively, right?

Almost exclusively, yeah. I downloaded v12 two weeks ago, but haven't activated my trial yet... not sure what I'd use it for

@rm-rf for most stuff you don't even need a trial anymore

Because of that Wolfram Engine thing which is basically Mathematica but without the FE

I mean there was also always the cloud too if you're down for a buggy FE

That might be fun, actually. I loved the core language more than the "knowledge entity" stuff that was being pushed on us v10 onwards... Although, I'll surely miss Manipulate :)

If you can get the ipython toolkit to play nice with the WL engine stuff you could keep some semblance of Manipulate.

Not quite as nice of course, but probably workable