« first day (2088 days earlier)      last day (2379 days later) » 

1:22 AM
@CoryWalker do you have a link for this project?
 
@user21 What on earth are you doing here at this time? :)
Look here:

 Expreduce

Discussion about the OS implementation of the Wolfram Language
and here
 
@halirutan, it's 9.30 in the morning here ;-) So I am well into my operating hours :-)
 
Ahh.. OK. Not German time then.
 
@halirutan, seems need to go move to a time zone (UTC+8) to have some overlap with you....I am in Taiwan.
 
@user21 Yes, you mentioned it recently.
I remember.
 
1:27 AM
@halirutan bits and bytes are just the same here... and NDSolve also works; well most of the time... haha. So I am good.
 
@user21 But it's vacation, isn't it? You're not going to live there.
 
@halirutan, I have been traveling on and off for a while by now. I don't have a 'home' in Germany right now.... So I am going to stay here for a while and then move on. I just like to travel, eat good food and I am always in for a good hike....
@Szabolcs thanks for your help with stochasticlifestyle.com/…. This looks more realistic by now, though it's still a bit contrived.
 
@user21 On a similar issue regarding LLVM and Julia: Do you think it will be helpful to send Tom Wickham-Jones a fan-letter so that he knows there are people out there who figuratively would die to see the new Compiler?
 
@halirutan, they are working on it - but it's a massive project. I am not sure how much of this is going to be discussed at this years WTC but I could well imagine it to be on the agenda. So you could stay tuned for that. So I'd wait for until after the conference and see what that has to offer; I do not know myself.
 
@user21 His last talk on WTC about it was impressive and I do realize how much work it is. I so wanna try this..
 
1:41 AM
Got to go now, NDSolve is waiting.
 
Good luck and happy eating and hiking!
 
Thanks.
 
 
7 hours later…
9:01 AM
@user21 Sound like the life I'm sometimes dreaming about :-) But sometimes I imagine it on a sailboat. My wife says the boat must have a proper shower at an absolute minimum to live on it full-time, which is probably beyond our means :(
When an array is passed to a LibraryLink function with that accepts any array type (Real, Integer, Complex), Mathematica must determine the type of that array and pack it if necessary.
Is there a way to do this with excellent performance manually?
I am hoping that there would be some internal functions to help with this.
Why?
I am trying to implement a system where the type would be automatically detected and a C++ template function would be called with the appropriate template argument. E.g. fun<mint> or fun<double> based on whether the passed value on the Mathematica side was an integer or real.
For scalars, this detection is easy in pure Mathematica with pattern matching.
For arrays, this detection is easy on the C side: just pass the array as is, without detecting its type. Then on the C side use MTensor_getType.
But I want to the flexibility to have functions analogous to this: template<typename T> tensor<T> multiplyTensor(T x, tensor<T> arr) (C++ish peusocode). This would multiply an array with a scalar of matching type.
This function has both a scalar and a tensor argument. The scalar type can only be detected on the Mathematica side and as for the tensor element type: I only know how to detect it on the C side with good performance (which is critical).
I need to detect them on the same side, either C or Mathematica, then dispatch to the appropriate specialized function.
I see two solutions: either detect the array type in Mathematica, or instead of sending scalars "as-is", wrap them into a list and send them like that: that allows postponing the type detection to C.
Of course there's still the problem of deciding what to do if the function is called with non-matching types :(
I really need to speak to someone about this before I sit down to implement it (and discover at the end, after a lot of work, that I was taking the wrong approach). It's a pity that there's no one here(where I live) who uses Mathematica and also LibraryLink
I will try to break these down into smaller and clearly written questions and post them on main. Even if I don't get an answer, it should help me organize my thoughts. The next best thing after being able to talk to someone "irl".
 
9:32 AM
Hm ... it seems that if I write a LibraryLink function that takes a tensor of unspecified type and rank, and pass it something that is not homogeneous, it will freak out and give strange error messages. No doubt this is an unexplored area of LibraryLink.
 
10:07 AM
ToPackedArray has a second argument, the target type. It should appear in the usage message.
 
 
2 hours later…
12:06 PM
Anyone had any issues with the help files being really slow to open in v11.2?
 
12:33 PM
@KraZug Do you have any third-party packages installed?
 
12:52 PM
@KraZug Also, check Import[FileNameJoin[{$UserBaseDirectory, "Logs", "DocumentationSearch", "MissingIndexCreation.m"}], "ExpressionList"] and look for lots of repeated entries. The result should be short without repeated entries.
 
1:20 PM
I don't actually have that file, or the Logs directory at all. Yes, I have MaTeX and one other installed.
 
@KraZug What version of MaTeX?
 
1.5.0
 
Please do upgrade to 1.7.2 :-) All it takes is running a short script. github.com/szhorvat/MaTeX#upgrading-or-uninstalling
MaTeX 1.7.2 won't cause any trouble with the documentation in 11.2 (nor will earlier versions—they will just have their indexes auto-converted once which takes a few seconds).
But other packages may, in principle. Especially if they integrate with the documentation center. It may be worth temporarily uninstalling them (move them out of $UserBaseDirectory/Application) and see if the problem persists.
 
The other package was something I tried once and didn't use again, so just deleted that.
Running your update script now, very nice to be able to do it like that
 
 
4 hours later…
5:43 PM
@Searke @halirutan I asked support about using the header files (in a non-commerical, open source project), and they said it was okay.
 
5:57 PM
For string patterns EndOfLine matches to "\n" but what does "\r" match to? Is there a named symbol for this or do we just have to match "\r"?
It is a Carriage Return
 
 
1 hour later…
7:20 PM
Is it possible to use a dispatch table Dispatch with StringReplace?
 
I installed 5.1 on windows 7 OK. But it does not like any input I give it. It seems to think it is all text that I am typing and not Mathematica commands. I called support and they said they do not support 5.1 on windows 7. Everything I type gives the error

"You tried to evaluate an unevaluatable cell. You can make the cell evaluatable by choosing Cell Evaluatable in the Cell Properties submenu."

any one here by any chance running old version of Mathematica on windows 7?
It might be because 5.1 is 32 bit and the OS is 64 bit that it does not work.
 
7:56 PM
Just discovered EvaluationData. Very useful.
 
8:30 PM
@Edmund No.
@Edmund Just \r. If you Import as "Text", \r\n on Windows should be auto-translated to \n.
 
9:23 PM
@Szabolcs Thanks. Note that I am on Windows and it comes in as "\r\n".
@Szabolcs That is unfortunate. So I guess I'll stick to a list of rules for now.
 
 
1 hour later…
10:30 PM
@Edmund With "String" it should come as \r\n and with "Text" as \n, as I remember. But I may be wrong.
 

« first day (2088 days earlier)      last day (2379 days later) »