« first day (1323 days earlier)      last day (3151 days later) » 

1:07 AM
"rubber duck debugging" - I vastly prefer the term originally used by McConnell: "confessional debugging". And yes, it is a frequent case for me: attempts to write a question in an easily understood form often leads me to solutions.
@barznjy, as noted by Michael, in calculus as well as in Mathematica, there are special rules for handling limits/integrals/sums/whatever that have infinity in them. Mathematica has those rules programmed in, so that it does what is (usually) expected.
@bob, could you post a screenshot?
 
 
1 hour later…
2:20 AM
@Guesswhoitis. Will do when I get to my office tomorrow - I've got MMA working on this display for the RPi and want to tweak X and MMA to make the experience a bit more inviting.
 
 
9 hours later…
11:09 AM
@MichaelHale that was useful. Thanks.
 
@halirutan Do you know how shared libraries are handled on Linux? I really don't know much about how dynamic linking works and I'm running into this problem: I compile a LibraryLink library that depends on some pre-installed package on Linux. It appears to link to this package dynamically when compiled with the default settings. Now my LibraryLink library will only work if I explicitly LibraryLoad the dependency.
Is there a way to avoid having to explicitly LibraryLoad that?
 
11:30 AM
0
Q: How to de-clutter MSE to ease searching?

Jerry GuernEarlier today, had a file go corrupt The WR documentation gave a solution that didn't work. So I posted this Question: How to restore corrupted files? In the first Comment, @belisarius posted a link to pertinent Search results. I clicked through every link, which led me to several un-Answered ...

 
11:44 AM
@halirutan There's this, but let me clarify: is it normal that Mathematica crashes if I don't manually LibraryLoad?
 
@Szabolcs I don't have to load every library my programs use, but the specific non-standard things, yes.
@Szabolcs I'm using something like this in my init.m and it works on all three operating systems:
$ipcuFound = FindLibrary[$ipcuLibrary];

If[ $ipcuFound =!= $Failed,

  $ipcuDependentLibs = {};
  Which[
    $SystemID === "Linux-x86-64",
    $ipcuDependentLibs = {"libtbb"},
    $SystemID === "MacOSX-x86-64",
    $ipcuDependentLibs = {"libtbb", "libiomp5"},
    StringMatchQ[$SystemID, "Windows*"],
    $ipcuDependentLibs = {"tbbmalloc", "tbb", "libiomp5md"}
  ];

  (* Load the dependent libraries before the functions from the ipcu_library *)
  Check[
    LibraryLoad /@ $ipcuDependentLibs,
    Message[ipcu::ldlibs]
@Szabolcs And I should tell you that, although it doesn't crash, Mathematica gave me an error message on Windows too when I did not pre-load the libs. Was hard to find out, because the message was pretty unspecific.
So it took a long while until I found out it is the dendency-lib that was making trouble and not my own.
 
@halirutan Do you have anything other than 10.2 installed on Linux? You mainly use Linux, right?
 
@Szabolcs Linux at home, OSX in office, yes.
@Szabolcs Yes, 8.0 and probably 9.1
 
@halirutan But not 10.0 or 10.1?
 
I mostly delete the real crappy versions like 9.0
@Szabolcs Not sure. It's possible that I haven't delete all of the crappy 10.x versions yet :-)
 
11:59 AM
I'd say 10.0.0 was the crappiest ever ... by 10.0.2 a lot got fixed.
 
@Szabolcs There was this 9.0 version that couldn't be installed on Windows or smth like that though
@Szabolcs Btw, have you seen that I implemented your Enter behavior in the plugin last night?
 
@halirutan Yes! I have it installed but haven't played much yet.
It works :)
I was obsessed with trying to fix some problem in my code so I didn't stop to test it out right away ....
Maybe these problems are due to LD_LIBRARY_PATH being changed by the Mathematica launcher script.
But then I don't know how the default library locations like /usr/lib are set ...
 
@Szabolcs I can assure you that it is not normal that Mathematica crashes. Even on Linux.
 
@Szabolcs Yes, I always have to adapt my Mathematica start icon so that it includes the additional lib folders
@Szabolcs But then again, it's only for the Intel libraries that are not part of Mathematica
@Szabolcs Which library exactly is it that troubles you?
 
12:09 PM
@halirutan It's my igraph stuff. Today I tried to make it work on Linux (Ubunut). I simply installed igraph with apt-get, linked against it with -ligraph, and this made it link dynamically.
now I have to do LibraryLoad["/path/to/libigraph.so"] before loading the package, to avoid a crash.
 
@Szabolcs And does LibraryLoad["libigraph"] work?
 
igraph is installed in what I assume to be some default system location, specifically /usr/lib/x86_64-linux-gnu/libigraph.so
@halirutan No.
 
@Szabolcs Then it seems to me that it is not installed in a path the Mathematica knows
@Szabolcs Or is this location in $LibraryPath?
 
LibraryLoad["/usr/lib/x86_64-linux-gnu/libigraph.so"] works. But I didn't expect this kind of problem and I'm not really prepared to deal with having to explicitly set library locations when I'm not shipping the library ...
@halirutan Yes, it seems so. But how do other software that depend on it find it then? I assumed it got installed system-wide. Do you know what the 0 suffixes mean in ubuntu, i.e. why this was called libigraph0 instead of libigraph?
 
@Szabolcs Can you link it statically?
@Szabolcs Often you have names with a 0 appended and the real name is a link to this file.
@Szabolcs I just installed libigraph on OSX with port. It cannot be found either. How do you solve it there?
(there I have to specify the complete location too)
 
12:20 PM
@halirutan Perhaps the default linking on OS X was static? I didn't have the problem on OS X. (Sorry, had to step out for a moment). yes, I'm looking into how to link it statically without making it too painful to keep using CreateLibrary.
 
@Szabolcs It is not always possible to link statically, but maybe you are lucky.
@Szabolcs Could you send me your library for OSX? I could look at it with otool and probably find out more.
I only need the compiled lib. I have 10.10.5
 
@halirutan Actually this one was compiled against igraph SVN, not macports igraph ... dl.dropboxusercontent.com/u/38623/IGraphM.dylib
otool -L does list igraph.dylib, so perhaps the path is hardcoded ...
@halirutan If you uncompress this into $UserBaseDirectory/Applications, does it work at all? dl.dropboxusercontent.com/u/38623/igraphm.zip
I have to find another Mac.
@halirutan Don't spend time on this, I realize I need to look into how dynamic linking work in some more detail ...
 
12:43 PM
@Szabolcs Let me see
 
12:54 PM
@Szabolcs AFAICS everything is hardcoded. I had to rebuild my own lib. This has the same disadvantages as they exist in Ubuntu, because the header and library locations are specific to my machine :-(
 
1:04 PM
Hi all.
 
@mikeonly hello Mike
 
Mathematica behaves strangely when I try to find FindGeometricTransform. Could someone please try to reproduce the same output?
More specifically, I follow this question in order to digitalise the plot.
I'm trying to align {0,0} point on graph with {0,0} coordinate and etc after catching all the plot points.
But I get this in result.
 
@mikeonly I can see why that would be problematic. Try aligning the {0,-1000} point instead.
 
@OleksandrR. Actually I align three points like this: FindGeometricTransform[{{0, 0}, {0, -500}, {4, 0}}, pointsOnImage].
 
@mikeonly You need to make the transformation rigid
@mikeonly If you only want to move the graph, then try
 
1:18 PM
Three points should be enough for translation and scaling if I understand it right.
 
FindGeometricTransform[...., TransformationClass -> "Translation"]
@mikeonly What I think is happening is that you are not only scaling and translating!
 
@halirutan Yeah, I tried that.
But, well, there should be additional scaling rather than just moving.
 
@mikeonly it is also better if you delete unnecessary image elements using an image editor before digitizing the points. Although that doesn't seem to have been a problem for you in this case
 
That's the result after applying FindGeometricTransform[...., TransformationClass -> "Translation"].
 
@mikeonly Then use "Similarity"
 
1:23 PM
A question about using gcc instead of clang (for OpenMP) on OS X: community.wolfram.com/groups/-/m/t/558403
 
@OleksandrR. I clean axes and Binarize an image beforehand, then get black points from it and apply the transformation function to them.
That's what `Similarity' does. I guess it should be an easy process and I'm obviously missing something crucial in it.
 
@mikeonly I'm not sure the rest of your approach is correct. So you have all points of the graph in "pointsOnImage", right?
 
@halirutan Ok, I'll elaborate on that.
 
@mikeonly I know the question you linked, because I wrote an answer. The result was a list of points for the graph.
 
So I get 3 points from the image.
Then I do "Copy Coordinates" and save them to a set pointsOnImage.
After that I try to find a transformation:
`trans = FindGeometricTransform[{{0, 0}, {0, -500}, {4, 0}}, points, TransformationClass -> "Similarity"][[2]]`
{{0, 0}, {0, -500}, {4, 0}} is a set of coordinates that should be a result of trans when it is applied to points set.
 
1:33 PM
@mikeonly Yes, I see. And did you already extract the points of the graph with the methods in the question you linked?
 
@halirutan Yes, I removed the axes manually and got that.
Points look to correspond to their actual place on the original image.
 
@mikeonly Ok, now you do the following:
Select the point with the minimal x-value {xmin, y} and then you move all points by -xmin.
 
And after that I hoped to automatically change their coordinates so that they will be placed relatively to the axes.
 
Select the points with minimal and maximal y-value and use those values to rescale all y values so that they are in [-1000, 500]
 
@halirutan Yes, that may tweak the transformation but not the scaling.
@halirutan I'll try that.
 
1:38 PM
Select the point with the maximal x-value and use this to scale all x-values to be in [0,20]
@mikeonly If you provide me the points for instance on hastebin.com I can give you a hand with this
 
But I am afraid the maximum y-plot value is not 500 as well as minimum is not -1000.
And I am quite annoyed with the way the produced this graph in the article.
I've aligned {0,0} point properly but scaling seams a bit harder for me.
 
@mikeonly Can you additionally post the 3 imagePoints you have extracted?
 
@halirutan {{175.56, 143.10}, {99.89, 141.39}, {98.53, 188.02}} is respectively {{0, 0}, {0, -500}, {4, -500}}.
My guess is that even subtly wrong points on the image after the scaling are the cause of that drastic rotation.
 
@mikeonly You have mixed something up. {0,0} is not your first image point
@mikeonly Anyway, give me a second
 
1:54 PM
Yikes!
@halirutan Of course it's not! Thank you.
@halirutan My fault is that I though that copy order is the same as the order of adding.
 
@mikeonly The approach is pretty simple. Here, with your points in the correct order:
pts = {{0, 0}, {0, -500}, {4, 0}};
imPts = {{98, 188}, {98, 141}, {173, 188}};

transform[{a_, b_, c_}, {aref_, bref_, cref_}] :=
 Composition[
  ScalingTransform[{First[c - a]/First[cref - aref],
    Last[(a - b)]/Last[(aref - bref)]}],
  TranslationTransform[a - aref]
  ]
and then you use it like
data = Import["http://hastebin.com/raw/buvevikebe", "Package"];

ListPlot[transform[pts, imPts][data]]
 
Lovely.
 
@mikeonly Btw, before you extract the points from the image, you should definitely do a Thinning on the graph-image!
Please look at my answer. I make this step along the way
curve = Thinning[Image[SelectComponents[comp, "Count", -1], "Bit"]]
Because now, you have a cloud of points when you inspect your graph closer.
 
@halirutan Thanks!
 
@mikeonly no problem
 
2:13 PM
@halirutan By the way, Thinning alone on a binarized image yield the following result.
That's why I skipped that step.
 
@mikeonly Can you post that binarized image?
 
If you are interested.
 
@mikeonly Look, when doing morphological operations, always remember white=1=signal
Therefore:
Thinning[ColorNegate[Import["http://i.stack.imgur.com/JHzew.png"]]]
@mikeonly You need to learn to debug! The first clue would have been, why you input image has a white background and your output image a black background :-)
 
@halirutan [Silly again]. Agh. Such silly mistakes.
 
@mikeonly Take them and remember how you solved them and why they happened.
 
2:19 PM
@halirutan Hm. So Mathematica tries to catch something in a white mess?
 
@mikeonly Yep, you see the thinned background.
 
@halirutan Wow, the pattern look great though.
@halirutan Thanks again. I really appreciate your help!
 
@mikeonly You're welcome.
 
3:11 PM
@Guesswhoitis. Here's the notebook that shows up every time the RPi version of M starts. It is called MathematicaNavigator.nb and I found that it is being loaded whenever DefaultFrontEndInit.tr is being parsed. Removing reference to the notebook in that tr file did the trick.
 
@bob, I'm glad you were able to solve it; it is strange that it does not have the usual checkmark in the lower right corner…
 
@Guesswhoitis. I suspect it's intentional since that's the only (obvious) way to get at the documentation - which for the RPi is only accessible on line.
Do you happen to know how to evaluate a notebook at startup and keep it persistent? Using the -ne argument (e.g. mathematica -ne notbook.nb runs the notebook but then exits.
 
I don't know offhand, @bob; sorry…
 
3:38 PM
Is there anyone around with and OS X version earlier than 10.10 and Mathematica 10 or later installed?
Especially 10.9 would be great.
 
 
1 hour later…
4:55 PM
@Szabolcs Yes, sadly I am still using 10.9.5 so I can help you, I think.
 
5:24 PM
anyone up for a fun challenge?
What could you do with MoMa dataset ?
 
@Silvia So, I am after an out-of-production connector from Honda Connectors
 
Ok
Then you traced it to this chinese site?
 
Exactly, I have tried writing an email
But nobody responded, of course it was not in Putonghua
And I suspect that is the issue
My question is -- Do you have any connections/ideas/friends that sell electronics/people in general -- that can point me to a website/storage/shop/whatever that sells those
 
@Sektor Who do you emailed to? From what I can tell, this site is a market pretty much like ebay.
 
@Silvia The shop owner I suppose, I tried QQ too, but he didn't respond
 
5:32 PM
@Sektor One of my friends is an IC engineer, I can ask her.
@Sektor Ah then I can ask the shop owner on QQ.
 
@Silvia That's great; let me give a link to the datasheet
 
@Sektor Great :)
 
If you can find anything related -- the crimping tools, the female connectors, different versions of it -- everything is a huge plus
 
@Sektor I'll forward that information to my friend. Hope she knows some ways!
 
@Silvia I hope so too ! Of course I will compensate you for your time and effort in a respectable way -- just name it :) And thank you !
 
5:37 PM
@Sektor Haha thanks in advance!
 
@Silvia ^^ Thank YOU :)
 
@Sektor Need to sleep now. Will ping you when I have more information! :)
 
@Silvia Superb; Go get some rest ! Wan an !
 
@Sektor Good night :)
 
Hi, I'm having trouble finding that post about the custom plot markers
I would like to use it as a starting point to creating some "status" graphics. Like a checkmark for Ok, a cross for not Ok, and things like this.
Found it.
35
Q: How can we make publication-quality PlotMarkers without version 10?

Oleksandr R.Suppose that for certain reasons we are not yet using Mathematica version 10, or we have a version with buggy PlotMarkers. It is well known that the default markers are font glyphs, and as a result they are difficult to size consistently, as well as exhibiting inconsistent alignment. Because of t...

 
6:04 PM
@Edmund find a font that has these drawn out. Then produce a PDF containing the character and import the character as an outline as Alexey shows. This is the simplest way to get such shapes
 
6:36 PM
@Pickett Can you test this work-in-progress package? Installation: extract, place the directory in $UserBaseDirectory/Applications. Test: <<IGraphM` , then IGVersion[], then IGBetweenness@RandomGraph[{10,20}], then IGBlissCountAutomorphisms[PetersenGraph[5,2]]. There should be no messages and please make sure the last one didn't crash.
I'm not sure if it works on 10.9.
 
7:08 PM
posted on September 02, 2015 by Giulio Alessandrini

I’ve taken pictures numerous times, either with a camera or with my phone, only to find out that the colors were completely off—they had bluish, reddish, or even greenish tints. Before I started working on image and color processing, this was quite mysterious to me. Moreover, I’d always noticed on my cameras a white balance [...]

 
@Szabolcs It works. No error messages or kernel crashes.
 
7:31 PM
Get ready for the flame wars to hit credit-card-sized computers. Matlab is now available on the RPi, although it looks like they aren't offering it for free as does Wolfram.
 
7:41 PM
0
Q: Protocol - Is it OK to write a long Comment as an Answer?

Jack LaVigneComment space is very short. Sometimes I have a very long comment where I would like to have the same capabilities as are available for Answers (e.g., paste long lists, show code, figures, ...) in order for my comment to be understood. I haven't done this yet but I could write the Comment as an...

 
 
2 hours later…
9:46 PM
In routine maintenance I prepared to delete this question but it has linked duplicates:
1
Q: Eliminate z from two equations in x, y, z and plot y as a function of x

user44840I have this equation to begin with: $$1 - 3 x y = 5 z - 2 x $$ I want to solve this equation for $z(x,y)$, then plug $z(x,y)$ I plug $z$ inside a second equation: $$ y = 7 x z - \frac{z}{x} $$ Finally, I want to plot $y(x)$ for $-5 < x < 5$ and $-5 < y < 5$. I tried numerically, but the fo...

The duplicates:
2
Q: Solve equations within range and plot

user44840Hi I have these pair of equations to solve: $$ 1 - 3 x y = 5 z - 2 x $$ $$ y = 7 x z - \frac{z}{x} $$ I want to eliminate $z$ and plot $y(x)$ within the range $-2 < x < 2$ and $-2 < y < 2$: I tried this code, but not sure why it's not working: eqn1 = 1 - 3 x y == 5 z - 2 x eqn2 = y =...

2
Q: How do I feed data points into an equation to solve NUMERICALLY?

user44840I start with this equation and solve it numerically for $z(x,y)$ in the range $1 < x < 5$ and $1 < y < 5$: $$ \frac{3}{xyz} - 2x - 3y - 5z = 0 $$ Then using the data points of $z$ above, I want to solve for this special condition: $$ x + y + z = 0 $$ Then I want to plot a graph of $y(x)$. E...

Should all of these be deleted or should something perhaps be reopened?
 
10:15 PM
@Mr.Wizard IMO, delete the first two, last one can be reopened if you want, or just left
 

« first day (1323 days earlier)      last day (3151 days later) »