Please read: Our unanswered questions is almost 1.3k or nearly 10% of all questions on this site. It has been pretty high for quite some time now, but I don't think that most of them are unanswered/unanswerable. It is likely that the problem was trivial and someone answered in the comments or OP posted a not-so-interesting question over a year ago and disappeared, and no one has bothered to answer it in all this time (and probably won't).

These questions can probably be closed+deleted with no major consequence to the site or the community, so I want your help in reviewing them. To start with, here's a search that returns all unanswered Qs with a score of 1 or less, asked between Jan 2012 and Nov 2013. There are about 600 such questions as of this message.

If the question is good/answerable, leave it alone (or edit to improve). If the OP has abandoned/question is trivial or missing info, please flag/vote to close as appropriate. It would also help if you left a comment explaining your reasoning (for the benefit of others that follow). Even if you aren't interested in going through the pile, please keep this in mind when you see an old post in the review queue.

@rm-rf Is it your fault that I'm getting votes in that weird MMA quine question?

@sjoerdc.devries @halirutan @belisarius @yvesklett @michaele2 @bobthechemist @artes @sektor @m_goldberg @cormullion @oleksandrr. @rojo @rcollyer @szabolcs @kuba @istvanzachar (and anyone else reading the chat transcript) please see my messages above. (cc @mr.wizard)

@Rojo You're welcome :D

@Rojo Had you ever have a vote for something not weird?

@belisarius Humm, I had a few stars when I complained about the chat room being hidden

@Rojo Oh, really. I forgot that very event

@rm-rf Sometimes, I think about answering some of them, but it's so useless to put the comments into an answer and not getting any rep for it. Mostly, it's trivial stuff and I always wonder why the commenters not just put there solution as answer.

@halirutan half and half. Half of them are trivial, and the other half unanswerable

perhaps there is still a boring third half

@belisarius yes, for me those are the questions which try to tweak shortcuts into the initialization files..

and something with Ticks

@halirutan We're all guilty of it... sometimes the stuff is just too boring for the commenter's interests to take the time to write an answer, but they nevertheless leave a comment to help the OP "right now". If someone (preferably a newish user who needs some rep) picks it up and answers, it's ok, but it'll eventually end up closed/deleted.

or Frames, spaces and arrangement of graphics and, most importantly, legends

@halirutan You forgot manipulating algebraic expressions to show up in the way I want

@rm-rf Do frogs feel electric guilt?

@belisarius Yes, I had lately a liaison with multiplication signs in TeXForm..

Here's a nice one for Mma, for those who care

@Rojo You still here?

@belisarius Yes, I'm here when I get pinged. Just alt-tabbing

@halirutan I see the site as having two important goals — to help the user with a genuine question as soon as possible and to be a repository of useful and quality information. If the first goal is satisfied by a comment and the question is not interesting/useful enough to keep around for the second goal, then closure is the logical solution.

Now the problem comes when we have to deal with the "community aspect" of it. We need a balance between furthering the site's goals while not hurting people's "feelings". That's why we don't randomly kill questions just because they're boring. What I've suggested here is to go through mostly old stuff, and specifically ones where the OP has more or less abandoned or is no longer interested.

@Rojo I've just finished watching a film I've seen many many years ago and it's still a great one. Do you mind taking note?

@belisarius Not so much guilt, but frogs do get electronic orgasms when deleting your answers =)

@rm-rf I'm looking at the third question and I'm just not able to decide what to do..

@belisarius Tell me

@Rojo thepiratebay.se/torrent/5715545/…

@rm-rf You should write comedies. For frogs.

About 600-2 questions to check :-(

Going to sleep. Night all.

@halirutan Night

@belisarius Thanks

@rm-rf "That's why we don't randomly kill questions" ha!

@halirutan Nights!

@belisarius The key being "randomly"... I didn't say we wouldn't target you :D

@rm-rf SeedRandom[42]

@rm-rf I love you frog. I know one of these days someone will dare to kiss you and having the spell removed you'll reveal all that beauty inside you to the open world. Till that moment, please shut up.

@rm-rf What reason should we give for closure for a boring post? Also this is the second post I've found where like one person has been able to reproduce the issue in the OP and the question hasn't been touched for a year. But what reason should one give for closure? mathematica.stackexchange.com/questions/16492/…

Time to go for a beer, c u l8r

Off topic -> other?

@Calle Yeah, that'd work. I wouldn't necessarily close it for being "boring", but more so because the issue is not reproducible/no activity or update in a long time, etc.

In this case, it's probably a localized bug with 9.0.0 (there were several) that's solved with the free upgrade to 9.0.1

ok

@belisarius Later

@Calle It's not really a dupe of mathematica.stackexchange.com/q/3926/8 but it's close enough that you could get away with saying it is.

@rm-rf Read it, will try to do my best. For the greater cause.

Maybe Parallelize is like Compile. For Compile I always imagined more things were compilable than was the case. I can imagine a lot of things to be parallelizable, but still I hardly ever see it used on this site. I will be on the lookout for that

@JacobAkkerboom writing parallel code is just difficult, especially when constrained to the scatter/gather model of the Parallel` package. There aren't really any limitations to what kind of functions can be used with it.

@OleksandrR. I was unable to make it work with LibraryLink functions. I think I tried something like Parallelize[{llfu[1], llfu[2]}], and the resulting computation time was considerably longer than just doing both funtions in sequence. For librarlylink functions at least you can use C++ stuff to parallelize I suppose, but I am not experienced with that.

@JacobAkkerboom The capabilities of Compile are bound to how much Mathematica supports it. On the other hand, writing parallel code isn't just replacing Table by ParallelTable. Often one could get the feeling that it is communicated that way. In real life, parallelization is just difficult as Oleksandr already said and it needs a good understanding of the topic to gain speed

@JacobAkkerboom Exactly this is one example where you need to understand what is really going on in order to make this work.

@halirutan I will try again to see if something basic will work or fail :)

does not work :(

But without Total it works :)

Having worked for years on products that attempt to scale to dozens of hardware threads and tens of millions of transactions per second, I'm not too surprised it's not that trivial.

@JacobAkkerboom I'm not convinced your kernels have the right definitions. Anyway, Amdahl's law is going to bite you in most cases.

Not to mention, if we consider typical floating point semantics, parallel addition is actually not easy to do. That is, if you actually expect to get same results as on non-parallelized run.

@kirma yeah but it's not like Mma cares about strictly correct FP anyway ;)

@OleksandrR. That's why I mentioned "typical floating point semantics"

... but I think it might care about the fact Parallelize[Total[...]] and Total[...] would return the same value.

@OleksandrR. or @halirutan Do either of you remember a question here about using the FE on local with the kernel on a remote machine?

@kirma I think really some problems are hard and some are not. I think this problem is about as perfect for parallelization as it gets

@rm-rf no, sorry. I'm sure there is one but this is not a case I've ever had reason to try so I would not have paid that much attention to questions about it

@JacobAkkerboom Here is a version which is at least comparable fast (on my machine). Watch out that I clear the compiled function on the main kernel, so that you can be sure that it is really computed on all parallel kernels:

@OleksandrR. could you explain to me what you mean by that my kernel has the right definitions?

cfu = Compile[{{list, _Integer, 1}},
   Block[{tot = 0},
    Do[tot += list[[i]]*list[[i]], {i, 1, Length[list]}];
    tot]];

mm = 1*^6;
rands = RandomInteger[1*^6, mm];

DistributeDefinitions[cfu];
cfu =.

ParallelSum[cfu[arg], {arg, Partition[rands, 10000]},
  DistributedContexts -> None] // AbsoluteTiming

I get timings of 0.055439 on the main kernel and 0.056698 if run in parallel.

@OleksandrR. Looks like this might be useful: library.wolfram.com/infocenter/Conferences/7250

@kirma I think that's probably debatable. The fact that Equal and SameQ both have a tolerance at which they consider values to be equal (and that this tolerance is not zero but actually very large, 128 ULPs for Equal) suggests to me that Mma isn't really designed with that kind of precision in mind. It would rather you just use arbitrary precision if numerical stability becomes an issue.

@rm-rf saw it before. I thought it was useful at first and then decided it wasn't :)

@JacobAkkerboom I think Mma is not particularly intelligent in parallelization unless everything inside Parallelize is parallelizable. OTOH, predictable rounding of floating-point sums is not trivial subject for parallelization. (At the same, Mma indeed leaves precise semantics of machine-precision reals often open, but at least results should be mostly reproducable!)

@OleksandrR. Oh, I haven't opened it yet. Good to know :D

@halirutan for me your solution is about 2.5 times as slow. A big speedup is possible with CompilationTarget-> "C" (of course? Should at least be mentioned :))

@rm-rf I remember one where a guy wanted to connect through ssh, but I'm not really sure whether it was the same Q which was about parallel subkernels and I don't know how to find it.

@OleksandrR. 128 ULPs is pretty little for big parallelizations, though.

@rm-rf This here has no answer: mathematica.stackexchange.com/q/39020/187

@halirutan This is probably the one: mathematica.stackexchange.com/q/11682/5

Ah but I see you partition the list inside the compiled function. That is nice, that should take considerable time. Of course this needs to happen, but in my case I was cheating in favor of Parallelize by doing it outside of the Timing

@kirma ah yes, rounding of floating points is a good point. I'm glad I chose integers then :)

@rm-rf Hmm, yes, it feels like I have seen this ... Anyway, I guess ssh is a good search keyword.

Regarding this problem: blogs.utexas.edu/jdm4372/2012/02/15/…

@kirma well, true. But also MathLink has relatively poor performance and poor scalability, so I wouldn't like to try large-scale parallelization with this method anyway. Someone wrote an MPI interface I think.

@VitaliyKaurov Hmm... how is ScanIndexed different from MapIndexed?

ScanIndexed? ahh, I have been missing that function!

@rm-rf You can use Scan on a held expression, but not map. Scan[Print, Hold[1, 2, 3]]

@OleksandrR. Valid point of course. :)

@JacobAkkerboom True, but I was referring to @murta's post in Community... :)

@kirma very interesting

So ScanIndexed would also make it possible to get a list of all elements (Level) of an expression and their positions, using Sow and Reap, in one swoop.

@rm-rf it doesn't need to accumulate the list of results

Hmm too many interesting things all at once :P

@OleksandrR. but the usage there is accumulating it, no? I would think a more idiomatic ScanIndexed usage might be (using his example) ScanIndexed[If[# > 0.8, pos[First@#2] = #]&, list]... I'm not even sure if Return would work as indicated there.

@rm-rf maybe @Murta can comment

@VitaliyKaurov Sorry, I meant to ping Murta but accidentally replied to your link :)

@rm-rf np, I do it all the time myself

Anyone knows when a stylesheet is opened in "Style definitions for xxx.nb" mode, whatever that is, and when as a regular notebook?

@Rojo What do you mean "regular notebook"? That is a regular notebook...

@rm-rf I mean, a notebook that has as a title just the name of the notebook. A notebook that appears in Notebooks[], a notebook that actually gets saved after getting saved

@rm-rf I know that if you just go to edit stylesheet, you will see a notebook that doesn't really exist outside the StyleDefinitions->Notebook[...] option of the father notebook, but

I'm speaking about a real stylesheet in a file

I want to edit it and resave happily and some times all I get is the "Style Definitions for xxxx.nb" to open, which I don't trust any more :P. Not until I understand it at least

@Rojo Anything that uses the stylesheet PrivateStyleDefinitions.nb is by default titled as "Style Definitions for xxx.nb". The presence or absence of that doesn't indicate saveability... most often, these style files also have Saveable -> False

@rm-rf Ok, but, if I have MMA closed, and I open my stylesheet, it doesn't have that title, and it uses PrivateStylesheetFormatting.nb

and sometimes, when I open it through MMA, it is as you say

Furthermore, in those cases, even when Saveble is set to True, after saving, when I reopen Mathematica nothing really changed in the stylesheet

At least I'm under that impression, I haven't been too methodical

@Rojo I don't know if this helps, but look at the code for se = SENotebook`StyleSheet`MathematicaSE. Now compare CreateDocument@se and CreateDocument[se /. Verbatim[Saveable -> False] :> Saveable -> True]. You can change stuff in the second one and save it (and reload with the changes).

Let me see

@rm-rf An exact copy of the same stylesheet file opens in "regular" mode while the original in "Style definitions for..."

and is not evaluatable even though it has the option on, isn't found in Notebooks, etc

Hello, Is there a way to tell Mathematica to always output in TraditionalForm?

@Alizter Yes

@Alizter If you want a global change, then going to Edit->Preferences->Evaluation

is the user-friendliest way probably

Thank you @Rojo!

Welcome

It's a stylesheet that's inherited by my default stylesheet

If I change the default, it works

If anyone has a suggestion, ping me up

Anyone have any thoughts on why the "wolfram language" doesn't give upvalues for TimeSeries datatypes for things like TimeSeriesMap etc? Very excited for all the new datatypes but it seems like it might be a bit of a bumpy transition if they make a bunch of non generic ways of using them

@bobthechemist much easier ... even when they are competing with the dust on my tiny laptop ;)

@Gabriel Thanks - the dusty monitor serves as an interesting analogy to the quality of an instrument's detector...

@Gabriel this is the way that sparse arrays are implemented as well. I expect that eventually (if they are being careful about it) all the time series stuff will be moved out into its own context and upvalues will be added to deal with them in the appropriate contexts.

sounds good. Not a major issue ... I just began to worry about having 10 different Maps ;)

@bobthechemist I agree with the proposition. But I'm not so sure if this accurately demonstrates the point? The dynamic range of emission measurements is larger (relative to a zero background) and there isn't such a problem with shot noise. But by having all the signal pixels be zero in the bright field image rather than e.g. 0.9 as it might be in practice, and by having a noiseless background, both of these points are lost

@OleksandrR. I agree that these are two extremes that don't adequately represent reality; I am trying to visually describe something from my students' textbook, which essentially boils down to saying that it is easier to measure an "increase in something from nothing" than it is to "subtract two large numbers with a small difference between them"

@bobthechemist I think this can best be understood as: in both cases the absolute difference is very small, and although in one case the relative difference is also small, in the other it is large and can be amplified and detected

@OleksandrR. That's a good way of putting it.

@rm-rf and @VitaliyKaurov ScanIndex advantage is that it do not need do accumulate the result, and you can have the position of some point of interest. Explain this here

It solves this problem:

with this: ScanIndexed[If[#1>0.8,Return[#2[[1]]]]&,list]

it's not a problem of code shortage, but memory consumption. The OuterScan is a better example of calc that can be done in a much more efficient way.

@VitaliyKaurov nice you liked it! I just miss in Wolfram community some way to comment a answer, in a form similar do SE. I know I can comment, but it get mixed with answers in a way that I can't separata answers from comments.

@rm-rf, do you believe that this question that I made in Wolfram Community would be out of topic here? We can say it's a question, but I feel it's strange format. I'm curious to see what kind of good ideas we can see.

@Murta It would definitely be off-topic. In addition, such a question was once asked here, then migrated to meta where it was closed (and eventually deleted). There were 17 suggestions there, some good but most were minor or about solving some localized pet peeve of theirs. I didn't think it was a constructive post and I don't think this format is suited well for such discussions.

@rm-rf ok... Let's see if we get something nice in WC

@Murta really, unless it's done in an official forum and instigated by someone with actual power to have the suggestions implemented, this kind of discussion is kind of pointless. Not only would it be off topic here, but that would be for good reasons.

More importantly, there were a couple of WRI developers that commented that such threads are pointless, because the company has its own vision and plans for future versions, and community/user threads have very little to no impact on their development plans.

Well.. there are some cases that are strong exception. Undo is one.

I would very much like if they would extend LibraryLink so that one can manipulate arbitrary exprs and not just packed arrays. Then we can implement Scan* functions ourselves if we like

Is there no way to delete a comment on the wolfram comunity site? All this fuss about doing Sort[DeleteDuplicates[list]], when there is just Union with one argument. Argfstf but I'll stop there because I'm sure rm-rf's download isnt finished yet :P

good night!

@Murta Also, a lot of the suggestions there were stuff that could be easily implemented by themselves/third parties. This, to me, points to a larger problem with the Mathematica user community — Everybody wants WRI to implement algorithms for their specific field X instead of doing it themselves/collaborating, but when they do, they don't bother polishing what they have written so far and sharing with the community.

This, in turn, is probably because Mathematica's market share is abysmally low in most fields and it doesn't make sense for them to invest the time to create and maintain a package. Well, then why would WRI bother? That's why they seem to be going with the popular crowd these days instead of the traditional science/engineering areas.

@rm-rf What do you mean by "That's why they seem to be going with the popular crowd"?

@rm-rf I'm going to be honest it took me almost 30 seconds to figure out that this is Wolfram i.stack.imgur.com/nJiYS.jpg

@belisarius Haven't you heard? They're implementing undo

I believed that upvotes could filter the best suggestions, just like what occurred with Undo here. But I just saw that Wolfram Community forum structure is not a good place for this kind of request. The post it already very confuse, with just few interactions.

@acl nah, that's a myth

@belisarius just joking

there must be some point of principle behind this decades-long reluctance to implement it. possibly even a religious motivation

@acl I know :)

in the sense that whenever we're faced with ancient unexplainable artifacts, we conclude they must have had religious significance

anyway

@acl @belisarius ... I was in WTC2013

have to say, I'm programming in a more conventional IDE these days. undo is nice

@Murta WaTerCloset ?

@Murta so, any chance of undo soon?

@Murta oh really?

Let me say... people can change religion..!

I searched the documentation but I couldn't find what I was looking for. Is there a continued fraction function similar to Gauss' CF notation? If not how else would I construct an infinite continued fraction?

@Murta OK that's great.

@acl I have to delete.. sorry.. but it's true

@Alizter doesn't FromContinuedFraction do what you need?

@Murta no sure no problem

@acl Can it have a repeating pattern?

@belisarius This whole Big Data science and Internet of Things, etc. are going to be $XX Billion industries (they probably are already) and WRI clearly wants to jump in and leapfrog over their traditional rivals, Maple/MATLAB, by linking with R and Hadoop and throwing in some cluster buster tools instead of trying to compete for a slightly larger market share in the traditional engineering/science fields (which is hard).

As in go on infinitely?

@Alizter well, if you look at the docs, the third example under Details does what you want

I think

@acl Ah! I missed that! Thank you very much.

@Alizter no problem

Don't get me wrong, that's a perfectly nice and valid business direction, but seems a bit too rushed and obvious and perhaps not of much use to their existing core user base...

@rm-rf But do you really see them competing in that area? I mean apart from the talk on their website, I wonder whether they can really provide robust tools

@rm-rf I'm in the Buniness Inteligence field. There is a huge space to MMA. But the have to make data interaction simpler, and are working on it with Dataset, Associations, and a lot of nice data structure code. I believe MMA can be a very powerfull player in this market.

I mean I have no idea, but looking at most of the "fancy" functionality (ie things above linear algebra, say), it seems flimsy to me

@Murta The problem with upvotes for requests/features is that mediocre stuff that got posted when the question is "hot" gets upvoted (and stays that way) way more than good stuff posted when the discussion has quieted down (say, after a couple of weeks)

@rm-rf the big problem, as it seems to me, is that to be successful in e.g. big data, the tools have to have superb performance and be absolutely bulletproof. I can't say that I'd really describe most of Mma packages that way.

@acl They're not competitors yet (not every kid throwing a punch is a pro-wrestler), but they certainly seem to be orienting themselves that way.

@OleksandrR. yes, exactly (especially the bulletproof bit)

Scalability is also a massive problem for Mma-based technologies. I bet they had to work all kinds of black magic to make W|A behave properly at scale.

@rm-rf SW announced so many disruptive tech and science concepts around the corner lately that ...

@rm-rf We'll see. But if I find most of mma flimsy and non-scalable, I doubt industrial use is possible

@OleksandrR. yes, precisely

@belisarius such as what? I've not been following

@OleksandrR. The main problem with adopting Mathematica for big data/cluster/distributed computing (and why I probably will not) is the licensing cost. If you want to run on 500 clusters, you need 500 licenses! And a 3 GB bloatware on each cluster. I would probably just go with native hadoop + R/python for all of that.

@rm-rf well, volume licensing can give a huge (>90%) discount. Also most large clusters will use a shared filesystem, so it's 3GB bloat on one disk only.

@acl WL is going to change the world, WA is going to change the world, NKS is going to change the world

et cetera

@belisarius None of these are new though!

@OleksandrR. And what if I want to scale my computation up to a 1000 cpus on the fly and then scale back? Get in touch with marketing for more licenses? Call again later to cancel some?

@acl RPi with Mma is going to change the world

@belisarius ah yes

actually, that has a much better chance of doing so than WA and NKS

chance of having an impact anyway

I am having trouble I write FromContinuedFraction[{1, 1, 1, [Ellipsis]}]

However I get a function back? Instead of a number

(3 [Ellipsis]+2)/(2 [Ellipsis]+1)

It then asks if I want to plot it or find the global min etc

@Alizter Why do you give it an ellipsis?

@rm-rf yes, well, that's obviously problematic. Personally I'm not too convinced that this kind of usage model really exists that widely in practice. But, if it does, I think they will probably try to address it via the Wolfram Cloud

... With full stops didn't work

@Alizter which actual fraction do you want? this syntax tells mathematica to use an ellipsis as a variable

Ahh I see

This is the error if I use ... with full stops

Syntax::sntxf: "FromContinuedFraction[" cannot be followed by "{1,1,1,...}]".

@OleksandrR. what kind of usage model? Scaling on the fly?

@rm-rf yes. I should probably say also that I think scaling to 1000 processes, on the fly or otherwise, is almost inconceivable with Mma as it is today

@OleksandrR. As far as it doesn't turn out to be a Wolfram Clog (TM)

@Alizter yes the syntax is wrong

What is the problem?

do you want 1+1/(1+1/(1+1/(1+...))) ?

Yes

@Alizter I think you need FromContinuedFraction[{1, {1}}]

So, read carefully the third paragraph in the details to see what this does

Oh the ellipses isn't part of the syntax?

@OleksandrR. Oh, no. Dynamic scaling is most certainly a very common model. Think of a website (say, one that allows you to input something, do some computations and return the output) that is designed to serve upto X users/min. What happens when they get posted on reddit/HN and are bombarded with traffic? They crash... or they can scale dynamically based on the current load and increase the number of cpus used temporarily.

@Alizter no

I read things to literally. Thank you once more!

@rm-rf well, a variable number of users or time pressure (e.g. financial report processing) are I think some of the few use cases that exist for this. And I wouldn't really think of doing either of them with Mma.

@OleksandrR. People do use clouds as rent-able supercomputers. When I was doing consulting and support for Microsoft's cloud those were by far the most interesting customers. Pixar would scale up their render farms during the end of production cycles. Most people were just trying to decide whether to switch over hosting for their pretty standard business websites though.

@MichaelHale But don't you think robustness has some way to go before this kind of problem is even on the horizon?

(I mean mma's robustness with scaling)

@MichaelHale I am quite surprised about that. Firstly the costs are prohibitive for cloud time versus building your own cluster unless you only need it VERY rarely. Secondly the cloud is not closely coupled by design so you have very poor inter-node latency and bandwidth compared to a dedicated HPC cluster

I haven't tried any distributed Mma stuff, but I've done quite a few overnight computations. For experimental stuff and prototyping I find it good enough to keep upgrading.

Well render farms fall in the "embarrassingly parallel" category, and like I said, talking to those clients were the brief breaths of fresh air.

They just send the scene descriptions on upload and get images to download.

@acl How can I have the "coefficients" change?

eg 1, 2, 3, 4...

@MichaelHale client is not cost conscious, workload is embarrassingly parallel. Yes, if I was offering cloud services I would consider that a breath of fresh air too :)

Yeah. They might not still be doing it, but apparently during "crunch time" they considered it worth a shot.

@Alizter this? FromContinuedFraction[{1, {1, 2, 3, 4}}]

@MichaelHale No of course, that is what I do too. I use it for up to a few weeks-long calculations. But I find that trying to scale to more than 7-8 parallel processes it starts falling apart.

well not falling apart, but things start being unpredictable.

Ah, I just have the home license, so I haven't tried with more than 4.

@Murta @Murta choose Tree View to see replies-to-post

@acl No I mean such that it increments every time