Arg, this thing with a complex conjugate is driving me inasne

insane, even

@Pickett Beside other things, this statement is true until today

> He claims Mathematica is a computer language but the grammar in nowhere to be found.[ reference]

@halirutan ok, I never considered that. It might be true, but if it is it is a distinction without importance :)

@Pickett That is clearly not the case because the lack of a definite grammar makes it impossible to write a compiler/interpreter for the language. Can you guess how popular JavaScript would be if there was only one closed source interpreter?

@halirutan What is the difference between an interpreter and the kernel?

Ok so you mean impossible for others, I think I see now. I know that Fateman wrote a paper discussing how to build a non proprietary interpreter, but you are probably right. I haven't thought a lot about whether it is possible or not.

I agree of course that Mathematica would be much more popular if there was a free implementation, but this seems to me rather an attack on proprietary software :)

@Pickett I'm not entirely sure how a free compiler would work because the real treasure is not the language, but all the mathematical knowledge that is part of Simplify, Integrate, etc..

So it wouldn't be impossible to for implementers to re-create Integrate because need not only the language spec, but the internal algorithms.

It's all a mess. You can compare it to Java. How useful would Java be without all the zillion libraries? And all the core libraries are open-source. Something like that would not be possible with Mathematica.

Nevertheless, a clear definition of the language and the semantics of corner-cases would be really helpful for people like me that try to realise e.g. a parser.

@Pickett Haven't seen this question: Well, Mathematica is for most parts an interpreted language and the Kernel is the interpreter. But when you think about Compile then you can take Mathematica input and really transform it to byte- or c-code.

Hi guys

@Kuba I would say Good Morning, but it's almost noon here :-)

@halirutan as you probably know, the same here :P

@Kuba I know :))

@halirutan I agree with all that in principle. Like Fateman noted in his paper, in 1990, there are other CAS programs (written e.g. in LISP) that could be used to implement subroutines. I am very impressed by Rubi, which outperforms Mathematica's Integrate although it was created by one single person. I wouldn't have thought it would be possible if it weren't for that package but he made it look easy (even though of course it's not.)

Now, can it be exactly like Mathematica? No, since some details are hidden. But it might be possible to build something that is to Mathematica what Octave is to MATLAB.

And when you look what happened to Jan Poeschko who started the Mathics project, one could come to the conclusion that possible rivals will be bought by Wolfram.

Looking at how unfinished (and unusable) Mathics was, we can only guess how much effort and people it would need to make a system that actually can be used.

Purely hypothetically, I could image than one could at least start such a project when certain people would join. We have here on SE a very unique mixture of experts from very different parts of Mathematica. Take for instance Leonid and WReach who both have a very deep understanding of the language itself. I look at Oleksandre for low-level compilation and of course optimization algorithms. Jens and Nikie who have a great knowledge about graphics and image processing.

Of course, we have many allround talents too and you just need to look at our top users like MrWiz, Bel, kgulr, rm. Some of them like Szabolcs did things like graph-theory to a deeper extend than it is handled in Mathematica....

@halirutan I was just about to say that Poeschko is not the most qualified person to build such a system. What if five researchers at a university were given five years to build a Mathematica clone? This is how Julia was developed. Surely they would get much further than Mathics.

@Pickett Yes, but the big but is money.

At university you get payed and if you are lucky, you can do fun projects like building a Mathematica clone.

If we really want to make a difference, we surely would need people like Leonid and good people need to be payed.

@halirutan I know, but I surmise that interest in this type project is an even larger problem. After all, Julia got this kind of grants and the Jupyter project got grants to build just a notebook interface.

But I'm just talking off of my head here. I have no knowledge at all about funding for this type of projects, except for those two examples.

For sure, it would cost a lot of money.

@Pickett I think something like this is often started in a kickstarter portal where people can vote and donate money for projects.

But I'm not sure one could raise enough money with this. A good software delevoper (not a programmer) will be about 100k per year if you take taxes and all the stuff you have to pay as employer into account.

So for 10^6 Euro, you get 10 people. You wouldn't get 20, but maybe 15. This for 5 years makes 5 million Euro for a team of 10-15 people.

Not including things like offices or God forbid travel costs.

yep

@MichaelHale WolframLanguageData[Dynamic] does not return an entity. While it does for all the other functions; eg WolframLanguageData[Sin]

@Edmund WolframLanguageData["Dynamic"]

@MichaelHale Yes. Somehow I got into the habit of not using quotes. It works for everything I have used ... until now.

@halirutan compare Perl!

@OleksandrR. At least Perl had one in an earlier version.

@halirutan can you test something for me? ge.tt/api/1/files/1jlbVaW2/0/blob?download

Seems to be a thread-safety issue in MathLink

When two kernels (endpoints of a link) write to the link at the same time, if the message is big enough, there's either a deadlock with shared memory or a livelock (sometimes) with TCP/IP

I'd be grateful for tests on both Mac OS and Linux if possible

@Szabolcs Could you help me a bit? I use your package IGraphM to solve this question. However, IGLADGetSubisomorphism for the colored graph gives me LibraryFunction::cfct: Number of arguments 4 does not match the length 3 of the argument template. >> At the same time First@IGLADFindSubisomorphisms works fine. I'm not sure, maybe I use it wrong or there is a small bug in the package. Thanks again for IGraphM.

@ybeltukov It's late now, but I'll take a look tomorrow. Sounds like a bug :( You got the package from the releases page, not directly from git, right?

@ybeltukov Can you tell me what IGVersion[] says? In the dev version it's fine, it must have gotten fixed at some point. A new release is way overdue but I haven't had time to work on it recently. There are many improvements implemented already (most of them not even synced to GitHub).

@ybeltukov I'll see if I can release an update on the weekend, maybe with not all improvements, but at least this one should be fixed.

@Szabolcs I compiled the master version from git. Unfortunately, the precompiled release doesn't work (it cannot load the library). The version is: "IGraph/M 0.1.5dev; igraph 0.8.0-pre+ (Jan 21 2016); Linux x86 (64-bit)".

Any one of the three people who downloaded that notebook willing to tell me what the results were? :)

@Szabolcs I just commented out the line igCheck(igraph_clique_size_ashist(&graph, &hist.vec, min, max)); in IG.h because igraph_clique_size_hist was undefined.