In the process of avoiding doing work ... code of mine that was used to generate graphs in a pre-print of mine may have a somewhat subtle bug in it that would affect the graphs. And, I really don't want to tackle it ... :|

Seems like I need to brush up on my $f:\mathbb{R}^2\rightarrow\mathbb{R}^2$ notation... anyone have a good primer as to what that means?

damn it, thought latex was handled here... I mean in this question: mathematica.stackexchange.com/questions/5362/…

Hey, traffic is rising again

We just passed Homebrewing. And I thought this was interesting:

On average, everyone visits at least once a day. Which means a few of us are a bit more obsessed than we might admit :)

@Verbeia never! I never check more than once, ever ever.. I swear!

More math notation question, is there a simple way of writing this (n[1]n[2]+n[2]n[3]+n[1]n[3]) as a Sum?

It might be too early in the morning to think straight on that one...

woke up enough: Sum[n[i]n[j](1-KroneckerDelta)/2,{i,1,3},{j,1,3}]

@tkott "f maps a pair of reals to a pair of reals"

@tkott Sum[n[i] n[j], {i, 1, 3}, {j, i + 1, 3}]

?

@tkott Or perhaps Total[Times @@@ Subsets[Array[n, 3], {2}]]

Thanks @acl, I like your way :) Taking me a little while to get up to speed this morning

@Heike ooh, pretty mathematica. I was looking for the Sum form merely so I could latex-ize it easily :)

@acl is there a good overview of these "math set" notations?

@tkott Fair enough. What do you mean with "math set" notations?

@tkott I don't know

@Szabolcs re our discussion in comments/ meta about ndroock's question: I would like any question about Mathematica to be on-topic here. I just think that to be a specific question, the asker needs to specify what would be an "improvement" in their eyes.

@Verbeia I'm here. I'm really on the "brink" about code reviews, I can come with arguments for and against. The real question is whether the answers would be useful or not, I think, and whether people would be willing to answer (too much work).

Sometimes a beginner wouldn't even know what to ask about specifically, yet experienced users could find several potential problems in the code

@Szabolcs Exactly. I think we have had "code-review" type questions already - see the meta discussion, where I posted a question of my own as an example.

for example, x_?NumberQ vs x_?NumericQ --- the former doesn't match Sqrt[2].

Another good example is Are these nested Tables necessary? - very much a code review, but specific in its aim.

and many times people suggest improvements even if the OP didn't ask for any

@StackExchange meta's RSS feed has more delay than the main site's

If ndroock specified whether he is looking for more concise /readable code, more functional style, or faster execution, then the question is specific enough to answer.

A small edit should do the trick

But aren't the specifics always going to be almost the same? I mean, isn't that implicit in the request for review? Things of interest are: are there edge cases when it would fail? Am I doing something redundant (such as reimplementing something that's built in)? Is there some obvious big performance problem, or potential performance problem?

I'm thinking about halirutan's highlighter, and how it got broken after it was integrated into the site

See here. If I implemented the highlighter, it would never have occurred to me to ask about such a problem explicitly

But the SE staff caught it right away.

@Szabolcs true, but I think the asker can make it a specific (and therefore a "good" StackExchange) question by indicating which issue is more important to them.

Edge cases might not be important if you are sure what kind of input you are getting.

The point is: sometimes there are some improvements which are very much worth noting but it would be difficult to ask precisely about them unless the OP is already familiar with that issue

@Szabolcs and we've generally noted them, e.g in comments - or at least Leonid has! :) I have lost count of the number of times he has mentioned a really important glitch or proposed an improvement, in comments somewhere.

I still think it would be helpful for askers to specify e.g. "I am looking for performance speed-ups and possible cases where the function would not work as expected" or, "this code seems really convoluted, can anyone suggest some ways to make it more readable". Then answerers know where to start.

Yes, that's true

If the question is more specific that's always better

Like I said, as long as they are not too burdensome, or too "plz fix my codez it don work", I would like code-review questions to be on-topic.

Let's see what others say. Right now I have (counter)argument for both sides, so that fact that I was arguing with your point doesn't mean that I want to allow code-review questions :-) I just don't know what would be better.

I'm wondering if they were allowed on SO before CodeReview was launched

My proposed solution to the conundrum is to have coding-style, functional-style or performance-tuning tags, but not a general code-review tag. There might be one or two other tags we could create to indicate the goals of any review.

@Szabolcs I don't know. Maybe there's something on meta (paging @yoda)

I didn't have time to read through, as I have a driving lesson in 12 minutes (I need to log off now). In the "tavern" they suggested this one:

good luck with lesson - I'm off to bed shortly anyway

It says pretty much the same what you said, @Verbeia

@Verbeia Good night!

@ndroock1 Instead of deleting, you could edit it a bit and explain what areas you'd like us to focus on. Correctness? Edge cases you might have overlooked? Performance problems?

class delayed ...

@tkott it is, by a script you run. See this link.

@Verbeia I visit many, many times per day. The figures may be skewed ... :)

@rcollyer thanks for the link

@tkott you're welcome. It is also in the starred list on the right.

@Heike I mean the notations like in this question: mathematica.stackexchange.com/questions/5362/… with the script R etc. The example given acl answered, but I've seen similar (and more complicated) expressions that I feel like I should know. Just wondering perhaps what the right google term for that is so that I can find some resources, or just the resources in general

@rcollyer haHA! That's all the way over on the right side of the screen. Much too far to look.

@tkott That's a good question ... Mathematics accepts questions on notation.

@tkott don't I know it. =)

@tkott That particular notation, $f:\mathcal{V}\to\mathcal{W}$, is a way of stating that $f$ takes as input members of set $\mathcal{V}$ and transforms them into members of set $\mathcal{W}$.

@tkott You could read the wikipedia page on mathematical functions.

Hi all:

@ajasja sorry to spoil the fun

It's an interesting question and good for a discussion, but not generally in the Q&A section

@SjoerdCdeVries Don't worry, I half-expected it to be borderline

My expectation would be that you would not see CompilationTarget->"CUDA" anytime soon

It would be useful for massively parallelizable simple stuff,but most programs wouldn't fit that description

Hmm, too bad. Is OpenCL any different?

OpenCL is just another (language) way to approach the same hardware architecture. I don't think it makes a difference.

Oh well, I guess we'l'l have to write kernels manually until then (not that I've written one yet).

@Ajasja I have a round-about way of accomplishing CompilationTarget->"CUDA". It involves SymbolicC` and CUDALink` , and using their specific compiler functions, but I think it is doable. I haven't tested it, yet, of course.

Of course, it is an answer to the question asked, but it may have the same results.

@rcollyer Really? Would anything that is compilable to C be compilable to CUDA in that way, or would one have to take even more care?

@Ajasja that is one of the issues I don't know about. The examples given are not that different from regular c.

The other question is how to use SymbolicC to turn a Mathematica expression into a C expression. Right now, I'm hunting that down.

Found it!

From there, it is a bit murky. But, it may be doable.

I'll play with it over the next week, and if I come up with something, it may be worthwhile re-opening your question.

@rcollyer Yes, especially how to parallelize, especially how to take care of those Those ThreadIdx and BlockIdx...

@Ajasja Parallelization is a question, but ThreadIdx and BlockIdx have SymbolicC counterparts. But, transforming a function from std SymbolicC to CUDALink` or OpenCLLink` will have some ambiguities that will have to be dealt with.

@rcollyer Hmm, it would be interesting. Even If you get just a simple example working it would be worth reopening the question.

@Ajasja I'll let you know. Probably won't get to it until early next week, though.

I should run, for now.

@rcollyer No hurry. Take care.

What do people think of 500's questions (a lot of them), which link to an external notebook (hosted on his site/dropbox) and require you to run it or need it for the question to be meaningful?

The entire question is then "too localized", as it doesn't help anyone else reading it. Besides, the question is at his mercy and any time he pulls the notebook away from his server, the question (and answers) stop making sense..

@RM To be honest, I still don't know what he's really asking for in his latest question.

yeah, there's that too...

For this particular question I don't really see why he the data file is necessary. I think some arbitrary data generated with RandomGenerate or so would probably work just as well and would make the question a lot more accessible.

As an interesting comparison, Mi Yodeya just graduated. Currently, we beat them in terms of users (1.2k v. 1.1k), traffic (15k v. 770), and questions/day (13 v. 6.8). They're beating us at number of questions (1.5k v. 4.7k) and number of answers (3.3k v. 10k).

My MMA frontend is eating up a core of my cpu on win7x64. This doesn't happen often, and not reproducibly that I can tell. Any clues as to debugging this? (Or should this be an actual question :))

@R.m I'd say this question looks 'too localized'. I don't really see how others could profit from its answers.

it seems like mma uses many processors when multiplying (sparse) matrices:

all I do is construct and multiply lots of sparse matrices

(this image shows that MathKernel is using 20 cores, for those not used to the output of top)

@rcollyer Have we had any communication about where we sit relative to their expectations for a launchable site?

Congratulations to Mi Yodeya, by the way. I remember when Isaac, their mod, dropped by because we'd referenced their "Question Week" idea in meta. I'm not Jewish but they seem like a really friendly coherent SE community.

@Verbeia Not that I'm aware of. I am just in a habit of comparing us to the newly launched sites to see how we match up.

@rcollyer @Verbeia This is exactly why I stressed early on that people not worry about statistics... from what I can tell, SE launches sites when the toot fairy toots

In this case though, mi yodeya (Me yoda, yeah?) was an SE 1.0 community which merged with SE2.0, so they came in knowing what they were getting into... also, Joel might have a fondness for the site.

@whuber, I am actually gonna a go with you I don`t think this simulation makes sense anymore. I have to implement the effect of the Center of Gravity. Thank You very much again.

@All, Hi ! I just want to mention that Mathematica, this place, is something really special and different from other programming language community. there is always rigorous, beautiful scientific thinking and methods her. I definitely found all of you and this platform as "one" of my mentors. I am sorry sometimes I have to do something in a rush and thus ask imprecise, time waisting questions. I will make efforts. Thank You Aall agin.