Re: [Numpy-discussion] Large symmetrical matrix
Pretty simple. I don't do any transformations. It is a euclidean distance matrix between n vectors in my data space, so I use it for lookup of minima and I recalculate portions of it during the processing of my algorithm. It is the single largest limitation of my code, both in terms of performance and scalability. A fast and efficient solution to this issue would make a huge difference to me. On Wed, Jun 11, 2008 at 1:05 AM, Robert Kern [EMAIL PROTECTED] wrote: On Tue, Jun 10, 2008 at 18:53, Simon Palmer [EMAIL PROTECTED] wrote: Hi I have a problem which involves the creation of a large square matrix which is zero across its diagonal and symmetrical about the diagonal i.e. m[i,j] = m[j,i] and m[i,i] = 0. So, in fact, it is a large triangular matrix. I was wondering whether there is any way of easily handling a matrix of this shape without either incurring a memory penalty or a whole whack of proprietary code? To get through this I have implemented a 1D array which has ((n-1)^2)/2 elements inside a wrapper class which manpulates the arguments of array accessors with some arithmetic to return the approriate value. To be honest I'd love to throw this away, but I haven't yet come across a feasible alternative. Any ideas? What operations do you want to perform on this array? -- Robert Kern I have come to believe that the whole world is an enigma, a harmless enigma that is made terrible by our own mad attempt to interpret it as though it had an underlying truth. -- Umberto Eco ___ Numpy-discussion mailing list Numpy-discussion@scipy.org http://projects.scipy.org/mailman/listinfo/numpy-discussion ___ Numpy-discussion mailing list Numpy-discussion@scipy.org http://projects.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] Large symmetrical matrix
clustering yes, hierarchical no. On Wed, Jun 11, 2008 at 4:55 PM, Charles R Harris [EMAIL PROTECTED] wrote: On Wed, Jun 11, 2008 at 9:33 AM, Simon Palmer [EMAIL PROTECTED] wrote: Pretty simple. I don't do any transformations. It is a euclidean distance matrix between n vectors in my data space, so I use it for lookup of minima and I recalculate portions of it during the processing of my algorithm. It is the single largest limitation of my code, both in terms of performance and scalability. A fast and efficient solution to this issue would make a huge difference to me. So are you doing hierarchical clustering? Chuck ___ Numpy-discussion mailing list Numpy-discussion@scipy.org http://projects.scipy.org/mailman/listinfo/numpy-discussion ___ Numpy-discussion mailing list Numpy-discussion@scipy.org http://projects.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] Large symmetrical matrix
You could do something like this, where MyDIstanceFunc is written with weave and some intelligent caching. This only computes the upper diagonal elements in a sparse matrix: self.NumDatapoints = len(self.DataPoints) self.Distances = sparse.lil_matrix((self.NumDatapoints,self.NumDatapoints)) for index, x in enumerate(self.DataPoints): self.Distances[index,0:index] = array(map(lambda y:self.MyDistanceFunc( x, y ), self.DataPoints[0:index] )) If an approximation is ok, you can save space by only calculating entries for random samples of the rows and doing a matrix factorization like NMF on the matrix with missing elements. Depending on the dataset and accuracy needed, this can be a big space savings. On Wed, Jun 11, 2008 at 8:55 AM, Charles R Harris [EMAIL PROTECTED] wrote: On Wed, Jun 11, 2008 at 9:33 AM, Simon Palmer [EMAIL PROTECTED] wrote: Pretty simple. I don't do any transformations. It is a euclidean distance matrix between n vectors in my data space, so I use it for lookup of minima and I recalculate portions of it during the processing of my algorithm. It is the single largest limitation of my code, both in terms of performance and scalability. A fast and efficient solution to this issue would make a huge difference to me. So are you doing hierarchical clustering? Chuck ___ Numpy-discussion mailing list Numpy-discussion@scipy.org http://projects.scipy.org/mailman/listinfo/numpy-discussion -- Peter N. Skomoroch [EMAIL PROTECTED] http://www.datawrangling.com http://del.icio.us/pskomoroch ___ Numpy-discussion mailing list Numpy-discussion@scipy.org http://projects.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] Large symmetrical matrix
Charles R Harris wrote: On Wed, Jun 11, 2008 at 9:33 AM, Simon Palmer [EMAIL PROTECTED] mailto:[EMAIL PROTECTED] wrote: Pretty simple. I don't do any transformations. It is a euclidean distance matrix between n vectors in my data space, so I use it for lookup of minima and I recalculate portions of it during the processing of my algorithm. It is the single largest limitation of my code, both in terms of performance and scalability. A fast and efficient solution to this issue would make a huge difference to me. So are you doing hierarchical clustering? Chuck ___ Numpy-discussion mailing list Numpy-discussion@scipy.org http://projects.scipy.org/mailman/listinfo/numpy-discussion Lapack supports symmetric matrices (http://en.wikipedia.org/wiki/LAPACK) but, to my knowledge, these are not implemented in SciPy (and I am not sure if sparse will help). So you have the memory penalty or performance penalty unless you create the bindings necessary that you need. Bruce ___ Numpy-discussion mailing list Numpy-discussion@scipy.org http://projects.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] Large symmetrical matrix
On Wed, Jun 11, 2008 at 9:47 AM, Simon Palmer [EMAIL PROTECTED] wrote: clustering yes, hierarchical no. I think scipy-cluster [1] only stores the upper triangle of the distance matrix. [1] http://code.google.com/p/scipy-cluster/ ___ Numpy-discussion mailing list Numpy-discussion@scipy.org http://projects.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] Large symmetrical matrix
2008/6/10 Simon Palmer [EMAIL PROTECTED]: Hi I have a problem which involves the creation of a large square matrix which is zero across its diagonal and symmetrical about the diagonal i.e. m[i,j] = m[j,i] and m[i,i] = 0. So, in fact, it is a large triangular matrix. I was wondering whether there is any way of easily handling a matrix of this shape without either incurring a memory penalty or a whole whack of proprietary code? To get through this I have implemented a 1D array which has ((n-1)^2)/2 elements inside a wrapper class which manpulates the arguments of array accessors with some arithmetic to return the approriate value. To be honest I'd love to throw this away, but I haven't yet come across a feasible alternative. Any ideas? Well, there's a risk of shooting yourself in the foot, but here's a way to get at the elements quickly and conveniently: store the coefficients in a 1D array (as you do already). Then if your matrix is M by M, element (i,j) is at position (M-2)*j+i-1, assuming ij. Of course, you can simply use this expression every time you want to look an element up, as you do now, but that's no fun. So now create a view A of this array having offset -1, shape (M,M), and strides (1,(M-2)). (This kind of weird view can be created with the ndarray constructor.) Then A[i,j] addresses the right element. A[j,i] is of course totally wrong. But as long as you avoid ij, you can use all numpy's magical indexing tricks. Unfortunately you can't really use ufuncs, since they'll waste time operating on the useless lower half. Personally, I would live with the factor-of-two memory hit. Anne ___ Numpy-discussion mailing list Numpy-discussion@scipy.org http://projects.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] Large symmetrical matrix
Actually that is very close to what I currently do, which is why I want to throw it away. The factor of two memory hit is not really the problem, it is the scaling O(N^2) which is the limitation I suspect I may end up using the 1-d array plus arithmetic as it is at least efficient for retrieval. I have just started working on a stack which sits between the vector array and a lookup for a minimum, I think that could improve performance a great deal. Next up is a disk based memory map of some kind which might free me from the constraints of physical memory. Anyone know of an implementation of array() which automatically reads/writes to disk? To answer another suggestion, I'm reluctant to sample not because it is not a good approach but rather because the commercial application at the end of this is much harder to explain, justify and sell if I have to include reasoning for not using all the data. As scientists this is obvious, as a lay person, and particularly a business person, it appears to be madness. The fewer leaps of faith necessary the better. On Wed, Jun 11, 2008 at 7:34 PM, Anne Archibald [EMAIL PROTECTED] wrote: 2008/6/10 Simon Palmer [EMAIL PROTECTED]: Hi I have a problem which involves the creation of a large square matrix which is zero across its diagonal and symmetrical about the diagonal i.e. m[i,j] = m[j,i] and m[i,i] = 0. So, in fact, it is a large triangular matrix. I was wondering whether there is any way of easily handling a matrix of this shape without either incurring a memory penalty or a whole whack of proprietary code? To get through this I have implemented a 1D array which has ((n-1)^2)/2 elements inside a wrapper class which manpulates the arguments of array accessors with some arithmetic to return the approriate value. To be honest I'd love to throw this away, but I haven't yet come across a feasible alternative. Any ideas? Well, there's a risk of shooting yourself in the foot, but here's a way to get at the elements quickly and conveniently: store the coefficients in a 1D array (as you do already). Then if your matrix is M by M, element (i,j) is at position (M-2)*j+i-1, assuming ij. Of course, you can simply use this expression every time you want to look an element up, as you do now, but that's no fun. So now create a view A of this array having offset -1, shape (M,M), and strides (1,(M-2)). (This kind of weird view can be created with the ndarray constructor.) Then A[i,j] addresses the right element. A[j,i] is of course totally wrong. But as long as you avoid ij, you can use all numpy's magical indexing tricks. Unfortunately you can't really use ufuncs, since they'll waste time operating on the useless lower half. Personally, I would live with the factor-of-two memory hit. Anne ___ Numpy-discussion mailing list Numpy-discussion@scipy.org http://projects.scipy.org/mailman/listinfo/numpy-discussion ___ Numpy-discussion mailing list Numpy-discussion@scipy.org http://projects.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] Large symmetrical matrix
The factor of two memory hit is not really the problem, it is the scaling O(N^2) which is the limitation Have you thought about using kd-trees or similar structures? If you are only concerned about the minimum distances between two points, that is the ideal data structure. It would reduce your running time / memory usage for generating / using the data from O(n^2) to O(n log n), with a O(log n) lookup time for the closest distance to each point. Without doing something like that, looking at the distance between each pair of points is \Omega(n^2). --Hoyt -- +++ Hoyt Koepke UBC Department of Computer Science http://www.cs.ubc.ca/~hoytak/ [EMAIL PROTECTED] +++ ___ Numpy-discussion mailing list Numpy-discussion@scipy.org http://projects.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] Large symmetrical matrix
On Wed, Jun 11, 2008 at 15:14, Simon Palmer [EMAIL PROTECTED] wrote: Actually that is very close to what I currently do, which is why I want to throw it away. The factor of two memory hit is not really the problem, it is the scaling O(N^2) which is the limitation I suspect I may end up using the 1-d array plus arithmetic as it is at least efficient for retrieval. I have just started working on a stack which sits between the vector array and a lookup for a minimum, I think that could improve performance a great deal. Next up is a disk based memory map of some kind which might free me from the constraints of physical memory. Anyone know of an implementation of array() which automatically reads/writes to disk? Well, if you have a decent OS, you should already be free of the constraints of physical memory. The OS should swap data in and out as necessary. But of course, we do have mmap support with numpy.memmap. If you are using numpy 1.1.0, I would also recommend using numpy.lib.format.open_memmap() to use the new NPY file format in order to (hopefully) future-proof your data. Be careful with using mmap'ed arrays, though. Disk seek time could kill your performance. If this is the big hotspot in your program, you might want to consider coding up exactly the behavior you need in C. It doesn't appea -- Robert Kern I have come to believe that the whole world is an enigma, a harmless enigma that is made terrible by our own mad attempt to interpret it as though it had an underlying truth. -- Umberto Eco ___ Numpy-discussion mailing list Numpy-discussion@scipy.org http://projects.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] Switching to nose test framework (was: NumpyTest problem)
Could this also be added as a comment in the module docstring for test modules? It seems that anyone hacking on test code would be most likely to see it there...? Barry On Tue, Jun 10, 2008 at 2:57 AM, Stéfan van der Walt [EMAIL PROTECTED] wrote: 2008/6/10 Alan McIntyre [EMAIL PROTECTED]: Is the stuff Robert pointed out on a wiki page somewhere? It would be nice to have a Welcome noob NumPy developer, here's how to do NumPy-specific development things, page. There may be such a page, but I just haven't stumbled across it yet. We could add this info the the `numpy` docstring; that way new users will have it available without having to search the web. http://sd-2116.dedibox.fr/pydocweb/doc/numpy/ Regards Stéfan ___ Numpy-discussion mailing list Numpy-discussion@scipy.org http://projects.scipy.org/mailman/listinfo/numpy-discussion ___ Numpy-discussion mailing list Numpy-discussion@scipy.org http://projects.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] Switching to nose test framework (was: NumpyTest problem)
On Tue, Jun 10, 2008 at 04:57, Stéfan van der Walt [EMAIL PROTECTED] wrote: 2008/6/10 Alan McIntyre [EMAIL PROTECTED]: Is the stuff Robert pointed out on a wiki page somewhere? It would be nice to have a Welcome noob NumPy developer, here's how to do NumPy-specific development things, page. There may be such a page, but I just haven't stumbled across it yet. We could add this info the the `numpy` docstring; that way new users will have it available without having to search the web. Build instructions don't really belong in docstrings. Put it in the README.txt or DEV_README.txt. -- Robert Kern I have come to believe that the whole world is an enigma, a harmless enigma that is made terrible by our own mad attempt to interpret it as though it had an underlying truth. -- Umberto Eco ___ Numpy-discussion mailing list Numpy-discussion@scipy.org http://projects.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] array vs matrix
On Sun, Jun 8, 2008 at 3:54 AM, Anne Archibald [EMAIL PROTECTED] wrote: 2008/6/7 Robert Kern [EMAIL PROTECTED]: On Sat, Jun 7, 2008 at 14:37, Ondrej Certik [EMAIL PROTECTED] wrote: Hi, what is the current plan with array and matrix with regard of calculating sin(A) ? I.e. elementwise vs sin(A) = Q*sin(D)*Q^T? Is the current approach (elementwise for array and Q*sin(D)*Q^T for matrix) the way to go? I don't believe we intend to make numpy.matrix any more featureful. I don't think it's a good idea for you to base sympy.Matrix's capabilities in lock-step with numpy.matrix, though. There are very different constraints at work. Please, do what you think is best for sympy.Matrix. Let me elaborate somewhat: We recently ran across some painful quirks in numpy's handling of matrices, and they spawned massive discussion. As it stands now, there is significant interest in reimplementing the matrix object from scratch, with different behaviour. So emulating its current behaviour is not a win. For consistency, it makes a certain amount of sense to have sin(A) compute a matrix sine, since A**n computes a matrix power. But looking at the matrix exponential, I see that we have several implementations, none of which is satisfactory for all matrices. I would expect the matrix sine to be similar - particularly when faced with complex matrices - so perhaps needing an explicit matrix sine is a good thing? Also worth noting is that this problem can be evaded with namespaces; matrix sin could be scipy.matrixmath.sin, abbreviated perhaps to mm.sin, as opposed to np.sin. I see. Thanks Robert and Anne for the replies. That's all I needed to know. Ondrej ___ Numpy-discussion mailing list Numpy-discussion@scipy.org http://projects.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] Switching to nose test framework (was: NumpyTest problem)
On Wed, Jun 11, 2008 at 6:09 PM, Robert Kern [EMAIL PROTECTED] wrote: On Tue, Jun 10, 2008 at 04:57, Stéfan van der Walt [EMAIL PROTECTED] wrote: 2008/6/10 Alan McIntyre [EMAIL PROTECTED]: Is the stuff Robert pointed out on a wiki page somewhere? It would be nice to have a Welcome noob NumPy developer, here's how to do NumPy-specific development things, page. There may be such a page, but I just haven't stumbled across it yet. We could add this info the the `numpy` docstring; that way new users will have it available without having to search the web. Build instructions don't really belong in docstrings. Put it in the README.txt or DEV_README.txt. I agree; I'm far more likely to look at the READMEs than in module docstrings for build/test instructions. ___ Numpy-discussion mailing list Numpy-discussion@scipy.org http://projects.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] Switching to nose test framework (was: NumpyTest problem)
2008/6/12 Robert Kern [EMAIL PROTECTED]: On Tue, Jun 10, 2008 at 04:57, Stéfan van der Walt [EMAIL PROTECTED] wrote: 2008/6/10 Alan McIntyre [EMAIL PROTECTED]: Is the stuff Robert pointed out on a wiki page somewhere? It would be nice to have a Welcome noob NumPy developer, here's how to do NumPy-specific development things, page. There may be such a page, but I just haven't stumbled across it yet. We could add this info the the `numpy` docstring; that way new users will have it available without having to search the web. Build instructions don't really belong in docstrings. Put it in the README.txt or DEV_README.txt. No, but running the tests is something every user should do before using NumPy. Unless you object, I'll add that bit to the docstring. Cheers Stéfan ___ Numpy-discussion mailing list Numpy-discussion@scipy.org http://projects.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] Switching to nose test framework (was: NumpyTest problem)
On Wed, Jun 11, 2008 at 19:51, Stéfan van der Walt [EMAIL PROTECTED] wrote: 2008/6/12 Robert Kern [EMAIL PROTECTED]: On Tue, Jun 10, 2008 at 04:57, Stéfan van der Walt [EMAIL PROTECTED] wrote: 2008/6/10 Alan McIntyre [EMAIL PROTECTED]: Is the stuff Robert pointed out on a wiki page somewhere? It would be nice to have a Welcome noob NumPy developer, here's how to do NumPy-specific development things, page. There may be such a page, but I just haven't stumbled across it yet. We could add this info the the `numpy` docstring; that way new users will have it available without having to search the web. Build instructions don't really belong in docstrings. Put it in the README.txt or DEV_README.txt. No, but running the tests is something every user should do before using NumPy. Right, which means README.txt, not something you only see after you've installed the thing. But what text exactly are you talking about? Nothing I wrote about inplace builds (which is I *thought* was under discussion) are relevant to running the tests in general but are targeted specifically at developers. -- Robert Kern I have come to believe that the whole world is an enigma, a harmless enigma that is made terrible by our own mad attempt to interpret it as though it had an underlying truth. -- Umberto Eco ___ Numpy-discussion mailing list Numpy-discussion@scipy.org http://projects.scipy.org/mailman/listinfo/numpy-discussion
[Numpy-discussion] [numscons] Would having two mandatory scons script per package be acceptable ?
Hi, Short story: it would make my life simpler to have two mandatory scons scripts per package instead of only one. Long story: for some reasons linked to scons' notion of build directory, having two scons scripts files per package (instead of one) would simplify quite a bit some parts of numscons, as well as simplifying the addition of custom builders (for swig, cython, etc...) by other people. One of the two files would basically be exactly the same as the current one, and the other one independent of the package (always the same 3 lines). This is a bit ugly, but less ugly than the actual contortions I have to do to make scons and distutils play nice together. Is there any strong feeling against this ? cheers, David ___ Numpy-discussion mailing list Numpy-discussion@scipy.org http://projects.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] [numscons] Would having two mandatory scons script per package be acceptable ?
On Wed, Jun 11, 2008 at 23:35, David Cournapeau [EMAIL PROTECTED] wrote: Hi, Short story: it would make my life simpler to have two mandatory scons scripts per package instead of only one. Long story: for some reasons linked to scons' notion of build directory, having two scons scripts files per package (instead of one) would simplify quite a bit some parts of numscons, as well as simplifying the addition of custom builders (for swig, cython, etc...) by other people. One of the two files would basically be exactly the same as the current one, and the other one independent of the package (always the same 3 lines). This is a bit ugly, but less ugly than the actual contortions I have to do to make scons and distutils play nice together. Is there any strong feeling against this ? Can you give me a longer story? What are the three lines? Why are they necessary in a separate, boilerplate file? -- Robert Kern I have come to believe that the whole world is an enigma, a harmless enigma that is made terrible by our own mad attempt to interpret it as though it had an underlying truth. -- Umberto Eco ___ Numpy-discussion mailing list Numpy-discussion@scipy.org http://projects.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] [numscons] Would having two mandatory scons script per package be acceptable ?
Robert Kern wrote: Can you give me a longer story? What are the three lines? Why are they necessary in a separate, boilerplate file? Sure: scons builds everything from the sources in the source directory, and build everything in the build directory. By default, both directories are the same as the directory where the scons script is. Now, this is of course unacceptable to use with distutils, and scons has this notion of build directory to change this: depending on where the scons script is called in the tree, I change the source and build directories accordingly. But for this to work, you have to change how to refer to source and targets in builders (scons objects to build things like object code and libraries). Again, unacceptable, so I did remove most of those in the scons scripts using severals 'tricks' in numscons. But the more I did those tricks, the more I realized it was fragile, and could break in obscure cases. The recommended way to do it in scons is to use a second scons script: http://www.scons.org/wiki/UnderstandingBuildDir. The 'problem' is also linked to the fact that instead of building everything with one scons project, I have one scons project per package (because that's how distutils worked, in particular, and for other reasons I can develop if you want). I asked about doing the same thing as scons does without two files, but this seems difficult (I was told it is 'magic'). http://scons.tigris.org/servlets/ReadMsg?list=usersmsgNo=13642 cheers, David ___ Numpy-discussion mailing list Numpy-discussion@scipy.org http://projects.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] [numscons] Would having two mandatory scons script per package be acceptable ?
On Thu, Jun 12, 2008 at 00:08, David Cournapeau [EMAIL PROTECTED] wrote: Robert Kern wrote: Can you give me a longer story? What are the three lines? Why are they necessary in a separate, boilerplate file? Sure: scons builds everything from the sources in the source directory, and build everything in the build directory. By default, both directories are the same as the directory where the scons script is. Now, this is of course unacceptable to use with distutils, and scons has this notion of build directory to change this: depending on where the scons script is called in the tree, I change the source and build directories accordingly. But for this to work, you have to change how to refer to source and targets in builders (scons objects to build things like object code and libraries). Again, unacceptable, so I did remove most of those in the scons scripts using severals 'tricks' in numscons. But the more I did those tricks, the more I realized it was fragile, and could break in obscure cases. The recommended way to do it in scons is to use a second scons script: http://www.scons.org/wiki/UnderstandingBuildDir. The 'problem' is also linked to the fact that instead of building everything with one scons project, I have one scons project per package (because that's how distutils worked, in particular, and for other reasons I can develop if you want). I asked about doing the same thing as scons does without two files, but this seems difficult (I was told it is 'magic'). http://scons.tigris.org/servlets/ReadMsg?list=usersmsgNo=13642 Works for me. -- Robert Kern I have come to believe that the whole world is an enigma, a harmless enigma that is made terrible by our own mad attempt to interpret it as though it had an underlying truth. -- Umberto Eco ___ Numpy-discussion mailing list Numpy-discussion@scipy.org http://projects.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] [numscons] Would having two mandatory scons script per package be acceptable ?
Robert Kern wrote: Works for me. Do you mean having two files/packages is ok ? cheers, David ___ Numpy-discussion mailing list Numpy-discussion@scipy.org http://projects.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] [numscons] Would having two mandatory scons script per package be acceptable ?
On Thu, Jun 12, 2008 at 00:28, David Cournapeau [EMAIL PROTECTED] wrote: Robert Kern wrote: Works for me. Do you mean having two files/packages is ok ? Yes. You had me at The recommended way to do it in scons -- Robert Kern I have come to believe that the whole world is an enigma, a harmless enigma that is made terrible by our own mad attempt to interpret it as though it had an underlying truth. -- Umberto Eco ___ Numpy-discussion mailing list Numpy-discussion@scipy.org http://projects.scipy.org/mailman/listinfo/numpy-discussion
[Numpy-discussion] [numscons] numscons build for numpy will be broken for a few hours
Hi, Just to mention that building numpy and scipy with numscons will be broken for a few hours, the time to change some organizations. Sorry for the trouble, David ___ Numpy-discussion mailing list Numpy-discussion@scipy.org http://projects.scipy.org/mailman/listinfo/numpy-discussion