« first day (316 days earlier)      last day (4174 days later) » 
00:00 - 19:0019:00 - 00:00

7:06 PM
I mentioned in this answer that using x[_] was possible but slower, I didn't realize it was this much slower :O i.imgur.com/zX27x.png
 
acl
@Szabolcs yes I was wondering whether they are downloadable. seems strange for them not to be
@Szabolcs Someone else did before I had time to.
actually I just used the new stpchastic process functionality to solve the KPZ equation in a few lines of code.
took me a few months to get this exactly right (and fast enough) during my phd...
 
7:21 PM
@P.Fonseca Not at all, ever since I got MMA9 everything has worked like a charm
It even stopped raining, my toliet started flushing properly again
Gaza and Israel stopped bombing each other
@ArnoudBuzing Thanks, I've been reached out by support
 
acl
@Rojo what happened in the end? is your w8 system dead?
 
@acl I can use it in safe mode :P
 
acl
@Rojo and all this due to a font problem?
 
I don't know, I don't think so but who knows
All I know is, it happens whether I open MMA or not, randomly but consistently before 5-10 mins of usage, even if I do nothing
It didn't happen before installing MMA9
Reinstalling MMA9, and uninstalling MMA8 didn't help
and in my first MMA9 installation, I had MMA8 open, didn't reboot, the fonts didn't display properly
Reinstalled, they got fixed, but I got a couple of MMA9 crashes before the first BSOD
I also got suspicious of my wacom bamboo tablet which I hadn't used for long and happened to plug in and use right about that time
Then I decided to just move back to W7 (which I have in very bad shape, sluggish)
This time I installed it with MMA8 closed
Couldn't activate it, of course. My MMA8 now doesn't like the piecewise character, gices an internal self test error
 
Hi! Just got MMA9 but it seems like the unit stuff using wolfram alpha does not actually work :/
 
7:36 PM
@nixeagle What's the problem? It seemed to work for me
 
AlphaIntegration (the package/namespace) is blue
My guess is that is not getting loaded or some such
=1+2 gives AlphaIntegration`....[<my input>, "Text Form"] (roughly, this is not exact)
and uh... now it works?
close and restart mathematica ftw?
Unitconvert butchers some inputs you would expect to work. Otherwise works fine
 
@nixeagle I also have troubles with the units. If I do something different than return, it blocks the interface for a while, and then leaves the "square" empty. Sometimes when I press return, it waits for 4 or 5 seconds and it returns the disable dynamic message. Other times it works fine. I know my internet connection isn't great, but it is not that bad...
 
Yea I get that too
The "blocks the interface" thing
 
It would be interesting to inspect the boxes. What's dynamic about the result of some units?
 
Also UnitConvert[<input>, "ft"] does not work as you would expect. Only input it'll take for imperial length is "Feet" (not "feet")
 
7:43 PM
Annoying...
And the temporary freezing of the interface with other stuff is much more frequent than on M8
 
I would have thought with all the datamining from Wolfram|Alpha that the company would have the simple units covered.
 
@nixeagle you mean having that part of the intelligence local?
 
@nixeagle if you are having issues with the WA connection, then its no wonder it can't even recognize "feet", it's not a lack-of-knowledge issue
 
@Rojo, well it works for "Feet"
this is what I get in the "predictive" interface
 
@nixeagle That's because "Feet" is the proper way to imput it
input
 
7:46 PM
after pushing 'convert'
UnitConvert[Quantity[1, "Meters"], "meter"]
or if you put "m" it fails too x)
 
IIt's clear
 
Dunno, I was just sorta expecting that simple stuff to be done right there.
 
it's not connecting properly to WA
 
99% of the time, I only use a small portion of the 4100 units available. Here some mixt up examples: ºC, m, s, m.s-1, m3.s, m3.h, L.s, m.s, Pa, Bar, atm, m3, m2, N, kg m-3, N m-3. How can create a shortcut for a placeholder + the unit in already its outputform?
 
but you can use it with the proper names. Plural, starting with a capital letter
 
7:47 PM
@P.Fonseca to get good use out of Units you will need to make heavy use of ctrl+=. Type something like UnitConvert[ctrl+= 2 in, ctrl+= "ft"]
 
Well I get the result correctly after doing =1m, which gives the Quantity[1,"Meters"], just the predictive interface sorta blows up after that. You have to put into it "Feet" (note the quotation marks).
 
@AndyRoss but for that I need a fast answer from WA, which it seems I'm not getting
 
Yet I'm typing into a text box, so I kinda would hope it would already be put into a string for me. x)
Anyway still a nice release, just has a few UI quibbles atm
 
I think it caches the result. If that's the case you are not getting a fast answer because you are not getting an answer
 
@P.Fonseca hmm.. its pretty fast for me. I usually do ctrl+= and then evaluate that in place to replace it with the proper Quantity[...] expression
 
7:49 PM
but, Fonseca, you could create your input aliases
such as esc ft esc
 
@AndyRoss how do you evaluate in place?
 
highlight and ctrl+shift+=
incredibly useful trick
 
just get the 5 second ui freeze followed by blank box x)
Mind you it would be simple (though a tad involved) just to write a lookup table/patterns for the common case of <number> <unit> that puts it in a Quantity form.
Quantity[1, "Meters"] + Quantity[2, "Furlongs"], very neat
 
@AndyRoss it takes around 3 or 4 seconds to get each result... And for example, I first tried 2 in, and it returned the message saying no results for my expression, and then I tried 2in, and it worked, and then I tried 2 in and it worked. So, not great. The fact that I'm in Europe should make the ping take a little bit longer, but shouldn't justify timeouts by 5 seconds...
I agree with @nixeagle that it should be possible to add a lot of aliases on the units, and then only leave to WA, the big ben height kind of stuff
 
@P.Fonseca I find it odd that it didn't work with 2 in. It does for me. Each time I try a new unit it takes about half a second to return the result for me.
 
7:59 PM
The couple of tests I could do before having my issues, showed that when I used quantity with some unit alias, only the first time it took a little time
 
@Rojo first time generaly returns a "time out", and then it takes some seconds
 
I'm in a dilemma. My activation key was reset. Should I activate my W7 mma9 and start playing with it (that's what I want), or wait for support to see if my W8 can be fixed?
 
@Rojo I don't want the Quantity on my inputs (it is unreadable), so I'm thinking on taking the formatted output, substituting its magnitude by a placeholder, and then creating an Esc m Esc for that expression. Haven’t tried it yet
 
I'll do that too for sure. In fact there was a question and I posted an answer on that
but, that was just a general input alias, not one for each unit
an esc qu esc, with two placeholders
 
Yah ... wolfram support fixed my install ... so so happy
 
8:07 PM
@Rojo just read it quickly. Seems that I have there all I need.
 
Grr, i'm gonna activate on W7, that's it. If they need to reset it again later, then so be it
 
Hi @rojo.
 
@Rojo did you manually delete the Mathematica fonts from your system dir? in Win8?
 
"The front end encountered an error while processing a NotebookPredicitons packet. Nice start
@Gabriel No, I didn't
Let me see
 
I'm going to call in a bit with some instructions on how to resolve this issue.
 
8:13 PM
@Searke Hello
Ok :)
 
Well I will leave you to the experts ... will be sweet ... back to my version 9. Good luck!
 
Enjoy and thanks Gabriel
 
Hi @Rojo ! long time no see!
have just tried version 9. ParametricNDSolve and WhenEvent is great!
 
WhenEvent is so, so, so great.
 
8:17 PM
I would pay full price for the upgrade for this feature alone
 
It can get this optical reflection simulation very neat and simple.
 
acl
is it that different from the EventLocator method? when I needed to calculate trajectories of balls bouncing off walls some time ago (the examples given for EventLocator in the docs), EventLocator did the job
but if nothing else, the WhenEvent seems much easier to access and understand
 
@Gabriel yes me too! It makes things SO elegant!
 
@Silvia That's very nice! Can you share this code so I can try this?
 
@Silvia the number of times I made functions to stitch up particular runs of NDSolve solutions so that I could change one of the state variables made me want to cry.
Can't wait to update some of my answers to use this cleaner method!
 
8:23 PM
I don't know the underlying method of WhenEvent. But I personally feel it much more easy to use than EventLocator.
@ArnoudBuzing of course.
 
Totally. I found EventLocator to enough of a pain that I just did it the fortran way as a series of intial value problems ;-)
 
With[{T=10},
sol=ParametricNDSolve[{
(x^\[Prime])[t]==a[t],(y^\[Prime])[t]==b[t],
x[0]==-5,y[0]==0,
a[0]==Cos[\[Theta]],b[0]==Sin[\[Theta]],
WhenEvent[-2-x[t]+x[t]^3+y[t]^2==0,
{vx=(4 b[t] (1-3 x[t]^2) y[t]+a[t] (-1+6 x[t]^2-9 x[t]^4+4 y[t]^2))/((1-3 x[t]^2)^2+4 y[t]^2),
vy=(4 a[t] (1-3 x[t]^2) y[t]+b[t] ((1-3 x[t]^2)^2-4 y[t]^2))/((1-3 x[t]^2)^2+4 y[t]^2),
a[t]->vx,b[t]->vy}]
},{x,y},{t,0,T},{\[Theta]},DiscreteVariables->{a,b}];
Show[{
ParametricPlot[Evaluate[Table[{x[\[Theta]][t],y[\[Theta]][t]}/.sol,{\[Theta],0\[Degree],30\[Degree],.3\[Degree]}]],{t,0,T},
 
@Silvia If someone's copying, they should replace ^\[Prime] with '
 
sorry I'm too exciting to format it ;D
 
@Silvia Hey Silvia :) long time
 
8:28 PM
With[{T = 10},
 sol = ParametricNDSolve[{
    Derivative[1][x][t] == a[t], Derivative[1][y][t] == b[t],
    x[0] == -5, y[0] == 0,
    a[0] == Cos[\[Theta]], b[0] == Sin[\[Theta]],
    WhenEvent[-2 - x[t] + x[t]^3 + y[t]^2 == 0,
     {vx = (
       4 b[t] (1 - 3 x[t]^2) y[t] +
        a[t] (-1 + 6 x[t]^2 - 9 x[t]^4 + 4 y[t]^2))/((1 -
          3 x[t]^2)^2 + 4 y[t]^2),
      vy = (
       4 a[t] (1 - 3 x[t]^2) y[t] +
        b[t] ((1 - 3 x[t]^2)^2 - 4 y[t]^2))/((1 - 3 x[t]^2)^2 +
        4 y[t]^2),
 
Yes, WhenEvent and ParmetricFunction are one of the things I'm most excited to test out
 
@Silvia Thanks! I'd like to mention demonstrations.wolfram.com, since this could benefit from some nice (user provided) new demonstrations.
 
@Rojo Let me know if version 9 works after reinstalling. If it doesn't, I'll be glad to call back.
 
@Searke Thanks, I'll let you know :)
 
@ArnoudBuzing i'll do it when I have some more interesting ideas
 
8:33 PM
I'd be glad to do some demonstrations too. So far I didn't mostly because I get confused by what generates security warnings, etc, and got lazy to figure it out
 
it seems very easy to construct an optical system with these new functions.
@Rojo lazy me too
 
8:48 PM
So far, no BSOD, and it's been a couple of minutes already :D
I'm getting hopeful
 
I installed the trial version on a laptop with no previous version installed, and it works fine.
 
@Silvia It seems issues begin when you install it where you had previous MMA fonts lying around
 
oh so I'm lucky! I installed on a brand new system
 
I don't want to get ahead of myself but it seems to be fixed :D
 
I encountered some problems when trying mma 8 on a machine which already has mma 6, so this time I'm careful
@Rojo It's very glad to hear that :)
 
9:01 PM
Good girl, smart
Yess, I'm glad to hear that too, I was already losing hope. Now it's time to play
I fear reinstalling MMA8 now. I won't for the time being
 
@rojo If version 8 has any font errors, you can replace the fonts manually. They are available for download here: support.wolfram.com/kb/10427
2
 
It hard for me to leave 9 now. Can't image how to face mma8 after 30days.
 
The fonts folder is in the installation directory under SystemFiles
 
@Searke Thanks, it all seems to work fine now
 
You replace the Fonts folder with the folder you download from the support website
I am glad to hear it is working
 
9:14 PM
I'm excited about midi import too
 
can anyone with mma9 run the following code
data = Select[WeatherData["Champaign", "Temperature", {2012, 8}],
FreeQ[#, Missing] &][[All, 2]];
temp = TemporalData[data]
eproc = EstimatedProcess[temp, ARMAProcess[3, 3]]
 
@Gabriel yes, but its really really slow and sometimes just hangs.
 
and tell me how fast the last line runs roughly ... thought I had the hanging solution fixed, but with the new install it is back
 
running, still running
 
ugh
at least it is not just me ...
 
9:20 PM
aaand still going x)
 
lol
 
I'm not familiar with midi. But I was writing so code to simulate my Guqin a few month ago. maybe I should learn more before action.
 
@Gabriel the good news is that we shipped with RLink. Obviously we intend to improve the performance of these estimators but for the time being you could always use RLink and say arima or auto.arima to get your estimates and then use them in M.
 
'some code' not 'so code'. sorry for the typo
 
@AndyRoss yeah that is what I have been doing ;-) super fast ...
so strange, I had the solver working in under 5sec last night ... but now it is slow again
 
9:22 PM
@Andy, how hard is it to add more *Links? Like for example one to haskell or lisp.
 
Yeah, there is obviously something weird going on that made it under our radar. I assure you it is high priority to get this working better.
 
@Silvia btw still going x)
 
@Leonid is a better person to ask about that. I've heard it isn't too difficult using JLink but I am about as far from an expert on that as one can get.
 
@AndyRoss great! I love mma9 sooooo much. Like all things once you get a new awesome feature ... it is easy to forget you got on without it :-)
 
@Gabriel still updating data
 
9:23 PM
actually uh, I just crashed the kernel x)
 
For example I find whenever the new predictor doesn't give me suggestions my life is a little worse now :-) Even though 2 days ago I didn't even now I missed this
 
One nice thing about the new TemporalData object is that you can evaluate it in place to store data like this so you never have to evaluate it again. You can then happily copy and paste it around etc.
 
oh very nice
I did not even realize that was a new feature lol
 
hmm.. haven't crashed so far
 
@AndyRoss yeah it is great. I really like it for doing line plots, like I would with R's matplot ... gets rid of a lot of ugly 'Transpose' calls I used to do to specify the x-axis
 
9:25 PM
@Silvia actually I think the reason for the crash/delay is the second line (temporal data), mostly because it has to download and initialize all of that.
 
Any reason it was called TemporalData and not something more general? I find I want to use it for things that are not only x-axis with time
 
speaking of names, Quantity is a great name for the units thing.
 
@Gabriel sometimes one just has to pick a name :)
 
@Gabriel it's 107 seconds first time, and 13 seconds when evaluated again
 
@AndyRoss I'm surprised it was not "temporarydata" if it is not specific .Usually wolfram does a great job with names.
 
9:27 PM
@nixeagle why "Temporary" ?
@nixeagle the idea is that it stores time-indexed data.
 
yeah I was thinking 'AssociatedData' or something. Still I can see time being the most common use case ... love it so much
 
@AndyRoss hmm, well I was thinking of my usecase which was just using it as a temporary stash for imports x)
^ Why naming is hard
 
@nixeagle totally ... I always dread naming library functions ... something Mathematica does better than any package I have ever used
 
@nixeagle @Gabriel there is always a balancing act between being too general and too specific. If you go too specific people won't recognize potential clever uses. If you go too general you need to come up with bizarre names in the future for other things.
 
Finally got a result on that.
 
9:30 PM
@AndyRoss makes sense ... I still have to remind coworkers about general functions for this very reason
 
@AndyRoss yea I understand. Naming is the last issue I have with MMA. Wolfram usually does a great job with it. What is killing me atm is this 5 second pause after Control + shift + = and doing anything other than Shift + Enter.
 
Its worth noting that a couple of other *Data objects came out with this release as well. EventData and WeightedData.
 
@AndyRoss I just started using them, but, if you were part of all the new statistics functionality, thanks a lot, will be very useful to me and so far I am liking it a lot
 
@AndyRoss yeah I haven't totally wrapped my head around those yet ... maybe not in my problem domain as much
 
those *Data objects
 
9:32 PM
@Rojo agreed I love the solution for wrapping data ... always wondered what a more idiomatic mathematica way woudl be for making these type of rich data objects ... sort of like R has
 
@Gabriel Yes. Don-t know R (probably will learn soon), but
 
@nixeagle I would like to download all the database onto my disk:) Have to go to sleep now. see you guys
 
I've had my own attempts of doing that
 
Any new way for handling "objects"?
 
can't wait until the properites of such objects will be suggested by the "smart bar"
 
9:33 PM
as in struct equivalents.
 
@nixeagle yeah ... I'm sure every mathematia user past a certain level has made their own solution of this type. These Data objects feel like the first time for me that I get a nice feeling for using something like this
time will tell
 
@Gabriel right, I am kinda hoping a nice idiomatic way for doing those comes.
 
@Gabriel make sure to send any suggestions you have to us, they are greatly appreciated.
 
it might have been released, but there are so many functions to check out that I have to actually dig around first.
 
agreed. Feels like they are starting to circle this problem. Statistical modeling seems to be the best problem domain for this ... as I feel R has done it better than any platform in a high level langauge
@AndyRoss good to know ... I am never sure what level of feedback is wanted. Used to opensource world until recently so I like trying out new ways of doing things!
@AndyRoss is using the suggestion link in the "smart bar" the right way to do this? I have been sending them whenever I notice something
 
9:38 PM
@Gabriel I hope it is the right way. Else my 3 suggestions will have gone to /dev/null x)
 
@Gabriel not sure who those suggestions go to. I would guess that whatever you suggest will get around to the right person at some point no matter how you do it.
 
oh hey, Quantity[1, "fps"] works as you would expect. Where it blows up is in UnitConvert.
 
Hey guys what's up? Was pinged by @Andy
 
That is Quantity does the right thing and takes random "human" form inputs like "fps", but ConvertUnit does not. Making ConvertUnit pass that along to Quantity would fix one of my issues x)
 
@LeonidShifrin they were asking about the difficulty of making links to other languages.
I said you would be the better person to ask about that.
 
9:45 PM
Alo Leonid
 
@nixeagle It "blows up" because "fps" actually stands for first person shooter =)
 
All wonderful here
 
@Andy That depends on a few things, such as how low-level we have to go (basically, are we forced to use C, or we can use Java), and how different that language is from M, and also, very importantly, how much experience one has in using various APIs. It also depends on how well that other language behaves in reality, in terms of intergation. I was very lucky with the first two aspects, not so much with the other two :)
Hi @Rojo
 
@LeonidShifrin yea I was asking about the difficulty of linking say... haskell or a lisp.
 
@nixeagle The other question is: Do you gain anything by linking. I would say you don't get very much from linking with Haskell or lisp, as you do with R.
 
9:47 PM
Since Compile is fairly useless for anything more than simple cases using machine sized data types.
 
@nixeagle I was thinking about Lisp myself. Probably not so hard, particularly if you don't request ultimate efficiency in data transfer.
 
@LeonidShifrin well yea the data transfer part is the part that is not that important.
 
Just keep your bag of parentheses handy... ;)
 
@LeonidShifrin so I guess next question is... what APIs and such do you need to code against and can it be done by someone without access to MMA source. (read, me)
 
@rm-rf Strategically, I think we can gain enormously. But that would require to reconsider what we think M is. I like to think about it as a potential technology integrator, something like a system which can catalyze the prototyping of hybrid systems written in many langauges.
 
9:50 PM
This works ListLinePlot[{Quantity[7, "Meters"], Quantity[1, "Meters"],
Quantity[10, "Meters"]}]
but this doesn't
ListLinePlot[{{1, Quantity[7, "Meters"]}, {2,
Quantity[1, "Meters"]}, {3, Quantity[10, "Meters"]}}]
 
@nixeagle I did not use any access to mma source in my work, although I had that access. All RLink is completely top-level M plus Java
 
It seems units have to be on both axis or none
 
@LeonidShifrin oh cool, I smell a side project provided I can figure out the appropriate APIs to use and such.
 
For Lisp (as for pretty much any other langauge), I would first use JLink to prototype (need some Java bindings for Lisp, but I think there are plenty of Lisps written in Java or otherwise with good bindings), then port to LibraryLink (may be)
 
@LeonidShifrin actually if we are going to use java, I'd just use clojure
That is a lisp that is on top of the JVM
 
9:53 PM
@nixeagle Well, there is some link to Clojure (called Clojuratica) already in existence. Did not test it though.
 
oh :o
and I assume this would work the same for stuff like haskell, basically code against JLink or LibraryLink?
 
The potential problem with Clojure (not sure how much it is really a problem) is that tail calls are not optimized on the JVM
@nixeagle Yep,I would say the same basic steps
 
@LeonidShifrin so? If you are using the link, you simply don't do tail calls x).
 
@nixeagle Yes, but then many recusrive algorithms may be ruled out
*recursive
 
actually the Common Lisp specification does not mandate that tailcall optimization must be done. If you want that assurance you need to look at Scheme.
Not that clojure is common lisp, but still.
and @LeonidShifrin, I assume clojure has idioms to avoid that kind of issue anyway. Just like we have idioms to make MMA go fast.
 
acl
9:57 PM
@rm-rf actually, all you need is emacs
 
@nixeagle Well, I am no expert in either, they are still on my wish list. May I ask if you have any connection to nix package manager, by chance?
 
nope
 
@LeonidShifrin question I have after using RLink for a bit is do we have any way of sharing DataType definitions? RLink is awesome, but most R functions beyond the most basic return some gnarly output, I have started to work on some nicer adapters using your framework, but is this only useful for each individual? Do mathematica people have any way of sharing this kind of thing with each other?
 
@nixeagle Never mind )
 
@LeonidShifrin I think in a few weeks though I might try playing with getting either haskell or a lisp to link to MMA9.
 
9:59 PM
@LeonidShifrin @gabriel I too am interested in the best ways for dealing with "gnarly" output from RLink :)
 
@Gabriel I am currently working on a set of open source tools based on Github gists, which will make such sharing a snap
 
Should be interesting and might solve one of my annoying issues with MMA.
 
@Rojo Yeah, that should work.
 
@LeonidShifrin so so awesome. Can't wait to test/add to such a system!
 
I recently put on Github the first part, the installer. The next step will be a more refined installer based on Gists. I already implemented Githib gist client in Mma, just need to drop the curl dependence in favor of new M HTTP functions or some Java, @Gabriel
 
10:01 PM
I saw the awesome HTTP functions. x)
 
@nixeagle Me too, but did not have a chance to try it yet
 
same
But they look really useful. I bet it would be trivial to implement stuff like a twitter client in just a few lines using it.
I remember the prior hacks at it needed you to drop down to jlink
 
Yay. Such a great time to be a mathematica user :-) I really am so excited by the recent advances, as more of a data/numeric scientist.
 
Twitter is kinda silly, but it is a very simple "usecase". Kinda a "if you can do this, you can also do more complex things"
 
@nixeagle Re:haskell that would be just awsome. You should also chat with Andreas Lauschke some day, he did links to a number of langauges (Java, Scala, F#) already - basically, execution of code of those from within M.
@nixeagle Yep, those functions seem quite useful
 
10:07 PM
@LeonidShifrin oh interesting. I think I'll end up using LibraryLink with a dose of C as haskell has a FFI to C
 
@Gabriel Thanks for the feedback, by the way!
 
So I'm assuming (famous last words...) that I just write a glue layer, make it all install nicely and call it done.
 
@nixeagle I would then use symbolic C,to generate as much C code as you can. This is what CUDALink is doing, b.t.w.
@nixeagle Too much boilerplate to write it by hand
 
Makes sense, haskell's FFI has stuff to reduce boilerplate too. (As does common lisp)
My main question is how to make install simple/painless. Do you just mandate that users have to install haskell beforehand or do you attempt to install it for them.
 
@nixeagle Basically, you'd need to figure out a small DSL for haskell linking in M, which would generate C code for LibraryLink. Nice task, would love to do if I had the time
@nixeagle It is best IMO to do both - like we tried to do with R.Have some default stuff coming out of the box, and give an option to link to something they have
 
10:11 PM
@Leo yea and I have this odd feeling that lisp would be even easier. Lisp has proper macros that do code generation in no time flat.
 
@nixeagle Yep, Lisp would be easier, at least should be.
 
@nixeagle Make it happen! I've just begun investing some time into haskell and a HLink` would be nice :)
 
@LeonidShifrin alright, I think for starters I'll just go with the "make sure you have ghc in your path" approach. Actually installing haskell can be a project unto itself. Especially if you want it to be trouble free.
Well the problem is not so much ghc as it is cabal.
 
@nixeagle Sounds good! Keep us up to date with this, I am very interested.
@nixeagle The package manager?
 
@LeonidShifrin yes the package manager. If you are going to install haskell on someone's system for them that "just works with MMA", it also makes sense to make it simple for them to extend it by getting packages off of cabal.
 
10:15 PM
@nixeagle Sure. I have the same problem with RLink. Package install currently only works on Windows, for the default R dsitro which comes with RLink.
 
And getting cabal to work is enough of a PITA that the haskell project creates a "haskell platform" that removes this problem from the user.
@LeonidShifrin exactly. Getting it to work nicely is the tough part.
 
@nixeagle I see.
 
So I think for this project I'm going to assume that the user installs haskell themself. Get that part working then try to make install simple.
Ok, as far as updates, though I can't promise you I'll get started on this until next week. But I'll have some time after that.
 
@nixeagle Yep, that sounds right to me
@nixeagle Re: next week- don't worry :) I have tons of stuff to do and it would be a pity to get a notification from you and not being able to even look at it )
 
Ok, checking out for now, I might have some questions for you @Leo on what to do with the apis and so on.
I will be using library link though as there is no java involved on the haskell end.
Plus I'm better at coding C than Java anyway.
 
10:19 PM
More mma9 problems :-(
Can anyone try the following code for me with MMA9
 
@nixeagle Sure, will helpwith anything I can (which is not that much, alas, I have lots of gaps. I actually prefer C to Java too)
 
eq = {0.` + 1.` (1 - 0.5` R) R - (1.4058533129758728` C R)/(
1 + 0.05623413251903491` R) ==
0, -0.618575457709384` C - (0.7905694150420949` P R)/(
1.` + 0.0316227766016838` C) + (1.4058533129758728` C R)/(
1 + 0.05623413251903491` R) ==
0, -0.3478505426185218` P + (0.` + 0.7905694150420949` C P)/(
1.` + 0.0316227766016838` C) == 0};
NSolve[eq, {R, C, P}]
 
thanks @leo, I'll probably just "ping" you over the chat when I get stuck. :)
 
works in MMA with no problems ... get a Solve::ratnz error in 9
 
@nixeagle Sure. See you !
 
10:21 PM
@Gabriel confirmed, I get the error too
and now see ya all!
 
@Gabriel I reported the issue to WRI.
 
@AndyRoss great thanks! Is there a way for users to report issues? Seem to be uncovering a bunch with all my trials. Happy to write any up if that is useful
 
thanks
 
@Gabriel works for me
{{R -> -16.5822, C -> 0.446209, P -> 11.9396}, {R -> 0., C -> 0.,
P -> 0.}, {R -> 0.451163, C -> 0.564829, P -> 0.}, {R -> 0.799363,
C -> 0.446209, P -> 0.327138}, {R -> 2., C -> 0., P -> 0.}}
with a Solve::ratnz: Solve was unable to solve the system with inexact coefficients. The answer was obtained by solving a corresponding exact system and numericizing the result. message
 
10:34 PM
@Rojo but the result is different than for 8.0.4.
 
Oh, that's the error?
I see
 
@Rojo yeah sorry should have been more clear
 
@AndyRoss is it possible to create your own processes, by for example adding a couple of built in processes?
 
@Rojo you can define them from stochastic differential equations using ItoProcess or StratonovichProcess. I'm not sure if that will do what you want or not.
 
@leonid I was wondering whether one can access Rcpp via RLink? Also, is it possible to use R formulas for specifying statistical models
 
10:39 PM
@Rojo It is possible via Java, e.g. I've got that answer ready, just a glitch prevents me from posting it
@Rojo Sorry,never mind
@asim You can use R formulas (I mean, import from R to Mathematica), but they are imported in such a way that they may not export fully correctly. This can surely be improved in the future if there is enough demand for that
@asim For Rcpp, I don't know. I did not yet had a chance to try it, although I know what it is, roughly
@asim But, on the general grounds, it should certainly be possible to use it if you use RLink with your own R distro
@asim Since if I understand correctly, Rcpp classes will load into main R process
 
@leonid: Another potential thing for the future could be to have access to RStan for Bayesian modeling and MCMC stuff
 
@asim But what prevents one currently from accessing that stuff via RLink (I don't know much about that stuff, so ask)?
 
0
Q: Structural analysis based on Mathematica

Siarhei A ArlouI just would like to discuss such theme. At this moment we see worldwide using of CAD/CAE systems based on FEM for engineering tasks solutions yet. Is exists future for symbolic calculations in such scale tasks? My opinion it is... As a first approach I build the TIMO software in numeric notat...

 
@leonid: I am not sure, as it involves parsing R formulas and accessing Boost and Eigen C++ libararies. I don't know whether such dependencies can be handled.
 
@asim are these packages that run in R or are they stand alone software that links to R?
 
10:50 PM
@AndyRoss: These are standalone packages that link to R via some Rcpp interface. I am not sure of all the intricate details. However, given their generality in terms of handling Bayesian inference tasks for practically any statistical model, this would be great functionality to have in Mathematica.
 
@asim In principle, RLink should be able to execute arbitrary R code, so if you can access that stuff from within R command prompt, you should be able to do that in RLink. In practice, it may be more subtle than that, expecially on Linux / Mac :)
 
Nice chatting with you all. Enjoy the rest of your day.
 
Bye
 
Have fun
 
11:08 PM
I must go too. It is 3 a.m. here. See you all later.
 
Good night
 
@ArnoudBuzing Have a nice evening, Arnoud.
 
@BrettChampion you may not be able to comment but the continued absence of built in two axis plots (for all relevant 2D plots) has brought howls of laughter from a MathCAD user I know well. Weren't you working on multiple axis features at one point?
Has anyone played around with CellObjects? Yet to download 9 but the documentation looks like this is a nice addition.
 
11:26 PM
@MikeHoneychurch I'm tempted to
Seems powerful
but, I just started playing a couple of hours ago. It's midis turn now
 
@Rojo I'm thinking it will make life a lot easier for some front end stuff
 
Definately
If you have anything concrete in mind give it to me and I'll test it
 
@Rojo nothing concrete but it seems to me that identifying specific cells and doing manipulations is now going to be so much easier (and shorter code blocks)
I got my download link today but probably won;t download until early next week -- got things to do and I know this will distract me :)
 
@MikeHoneychurch You're stronger than me
 
acl
11:53 PM
The nice thing about Mathematica is that you can use the notebook front-end to combine code, commentary and results. I do this often. Sometimes, looking back at my older work notebooks I find enlightening comments such as:
 
Is python a language that is less in demand these days in north america? Reason i ask is that while some form of python linking has been bundled with Mma for a few versions it hasn't been formally documented and updated -- so wondering if demand has dropped away. In Australia i was talking to CSIRO recently who decided on python for a relatively large project. ...
 
acl
Guess I need to redo the whole thing and work out what's going on. So much for documenting my work...
 
...Yesterday I was talking to a commercial organization (image processing predominantly) who have Mma, Matlab, R but are ramping up python usage. (BTW I know of pythonika)
 
acl
I am seeing a 40% speedup some of my code. This is linear algebra stuff with some compiled functions. This is fantastic
2
 
@MikeHoneychurch it is probably more that the overlap between MMA and python is larger so there is less reason for using both. Whereas things like R have a crazy library that is simply not possible to port over to MMA (most python science libraries I have seen are wrappers of lower language libraries so using the Python layer makes less sense)
 
00:00 - 19:0019:00 - 00:00

« first day (316 days earlier)      last day (4174 days later) »