@EliLansey haha I shouldn't have visited that site. Now I'm fooling around

@CHM Kite is the mother to visit the site. Now it is illegal.

@yoda First phrase I tried: Please take this to the post office.

@Heike How did you figure that out?

@Heike Was your clue that it is ReadProtected?

@Szabolcs By accident. I tried to evaluate AbsoluteOptions[Histogram[RandomReal[10, 100]], PlotRange] which returned an error and something of the form {PlotRange -> PlotRange[...]}

Hm, even Graphics has DownValues ...

(It's boring, it's just Graphics[] = Graphics[{}])

I am wondering what is the best way to test if a notebook exists.

Suppose we create nb = CreateDocument[]. Then close the document. Operations on nb such as NotebookGet will return $Failed because the notebook does not exist any more.

I am simply wondering what is the best (i.e. fast) way to test if a notebook exists. NotebookGet might be slow and return a lot of data for big notebooks so I'd like to avoid it.

Alright, after a little spelunking I found MemberQ[Notebooks[], nb] used internally. I'll just use the same. It's in the formatting code of NotebookObject.

@Szabolcs I was about to suggest that.

I'm now looking at the definition of Notebooks

I found ArgumentCountQ, also undocumented

I think the purpose of this is merely issuing an error if the wrong number of arguments are passed

"Notebooks[fe] gives a list of notebooks open in a specific front end" <-- can we have multiple front ends connected to the same kernel? If yes, what is the use of that?

How long does it take for a proposal to be "prepared for private beta"? Chemistry.SE is 100% since yesterday, but they're "preparing".

@CHM It took about 2-3 days for Mathematica IIRC. I think they like to launch sites around Tuesday or Wednesday.

@Heike oh ok. I thought the process was fully automated.

@CHM Yes, they like to launch around Tue-Wed, while avoiding holidays. The reasoning is that those are the days when traffic is the highest and they want to give a good start to the site

@CHM StackOverflow traffic charts with Tue-Wed peaks

@CHM I'm pretty sure they prepare for it manually and make sure that everything is smooth

@Szabolcs ok. Nice stats.

@Szabolcs might be useful for remote kernels, or perhaps for developing front ends on new platforms. But more likely in my opinion is that whoever wrote the Notebooks function simply didn't want to hard-code an assumption of only one front end when multiple front ends is technically a possibility, regardless of whether it's actually useful.

@OleksandrR I dug into the source of UsingFrontEnd, and it seems that there is no explicit way of launching a second front end unless the kernel is used with a remote front end and a local front end is needed for some reason (I guess for exporting or for rasterization)

@OleksandrR I'm curious, do you have any special tools for spelunking? Here's one question I asked about this: mathematica.stackexchange.com/questions/1742/… (I use those functions a lot)

@Szabolcs could always launch it as a MathLink program using Install and then change $ParentLink I suppose.

Of course, I was just wondering if there is built-in support, which would suggest that this is something that is needed on occasion

@Szabolcs no, not really. I often hook functions using the Villegas-Gayley method or by putting some debug output into a Condition. But I think if you have a good idea how Mathematica's put together, high-tech tools aren't really necessary for spelunking.

Is anyone familiar with TaggingRules? Is the best way to set them CurrentValue (instead of SetOptions)?

CurrentValue[nb, {TaggingRules, "one"}] = 1 allows setting a "suboption" easily, while SetOptions doesn't seem to allow it.

I'll just go with assignments to CurrentValue.

@Szabolcs That's mentioned in "possible issues" of SetOptions.

@Heike Thanks! That gives me some reassurance

SetSystemOptions does support setting suboptions directly, e.g. SetSystemOptions["CompileOptions" -> "ApplyCompileLength" -> 500]

@Szabolcs also, when I encounter a situation that's slightly out of the ordinary (and/or documented) but not too obscure, I presume that someone else had the same problem before and might have implemented a function to solve it. Looking for likely keywords using Information often works, and if the functions seem stable, I (cautiously) use them. That's probably the reason why I tend to encounter undocumented functions more often than many others, rather than using different tools (for example).

@OleksandrR So you don't read source code of internal functions, you just search for guesses with ??

@Szabolcs if the functions have readable source, I do usually read it. But that's often not the case, especially for functions in Internal` context.

it has had its 200 committers for 10 to 15 days now

@Fx maybe SE people all use Blackberries? :P

but there is all-team SE supermeeting somewhere in the world (I bet on a Caribean island), so the team said they wouldn't be around much these days

@Szabolcs in regard to finding out the internal working of FindMinimum for example, the BFGS method is done totally inside the kernel and is a black box, so all I could do was to Trace with TraceInternal -> True and look for calls visible from top-level code that contained useful information. Fortunately there was one, so the problem was soluble.

@Fx You need more than just 200 committers

But it seems to have a 100% commitment score

@Heike yes, but in that case I believe the number of committers was the limiting factor

@Szabolcs But it's hard to tell how long they've been at 100%

Does anyone have the slightest idea what this code is supposed to accomplish?

@Fx Looking through the list of committers I suspect you're right.

@OleksandrR I think the OP is basically integrating a polygon.

At least up to stop 1.

He's using a rather complicated way to take the inner product of two vectors though.

I wonder why he's using all those ParallelTables.

@Heike I suspect "it's slow, so let's sprinkle Parallel* all over". Unfortunately in this case it seems to be a considerable hindrance.

@OleksandrR I think it might even make things slower in this case since he's hardly calculating anything.

@Heike agreed. On the other hand a couple of well-placed Simplifys could help a lot.

@OleksandrR I suspect so.

To be honest it's not the kind of thing I feel inclined to give an answer to (and probably that's true for everyone else as well). The method is seemingly needlessly complicated and ParallelTable/Parallelize certainly contributes to the memory overhead he complains about, if nothing else.

@OleksandrR I feel the same. I was thinking of just posting some remarks about his way of taking inner products and about the (ab)use of ParallelTable and leave it at that.

Also, memory is very cheap nowadays. If your calculation needs no more than 32GB, in my opinion it makes sense to just buy more (as long as the method isn't the cause of the high memory requirements through gross inefficiency).

@Heike I think that would be a very helpful comment to make.

@OleksandrR Sometimes that's not possible. My laptop for example can't take more than 4GB of RAM.

@Heike okay, fair point--but I also meant that computers capable of taking up to 32GB of memory are cheap, so no sense in using an inappropriate computer just for the sake of it. I personally think 8GB is a good minimum for serious Mathematica usage.

Above 32GB, the costs rise dramatically.

@Fx wow. Bizarre.

@CHM what is bizarre?

@Fx that the Windows Phone stack is being "prepared" for more than two weeks.