On Wed, Jul 14, 2010 at 1:41 PM, Maurizio <maurizio.gran...@gmail.com> wrote: > how do you get the new canvas with matplotlib 1.0? I installed it > today, but from matplotlib.rcsetup.all_backends I don't get any item > which looks like html5 canvas
http://trac.sagemath.org/sage_trac/ticket/9471 > > Maurizio > > On 14 Lug, 10:03, William Stein <wst...@gmail.com> wrote: >> On Wed, Jul 14, 2010 at 9:42 AM, Maurizio <maurizio.gran...@gmail.com> wrote: >> > Hi >> >> >> Very interesting. >> >> >> 1. How does the speed of the Sage notebook running locally on your >> >> computer compare to Spyder locally on your computer? >> >> > I don't think they can be comparable, doing so different functions... >> > Anyway I always use notebook from a server in the local network (which >> > is anyway pretty fast). And I still fear that my python installation >> > in the windows virtual machine is somehow broken because the ipython >> > console is much slower than the plain python console. By the way, some >> > time ago we managed to run spyder letting it use the sage console as >> > interpreter, which was kind of fun, but never used it extensively. >> > Basically what we did was to install spyder on the server hosting >> > SAGE, and then running Spyder logging on that machine from our >> > workstation and exporting the display. That has been a very >> > interesting experiments, but later I had little occasion to use it, >> > and the notebook is still more comfortable for touch and go. >> >> >> 2. Are the plotting issues you mention the result of Spyder embedding >> >> static png images (like the sage notebook does) or something more >> >> subtle. The sage notebook might switch to HTML5 canvas rendering >> >> soon.... I say might, because after having tried it a bunch, I'm >> >> seriously concerned that HTML5 canvas matplotlib is slow -- >> >> surprisingly, maybe much slower than using png's and image maps, which >> >> we should have at least enabled long ago. >> >> > Yes, they embed static png images with image maps (if I understand >> > what you refer to), which is standard matplotlib output nicely put >> > inside a Spyder subwindow. The problem is that I strongly dislike this >> > output form, I consider it like a fake form of interaction. Some time >> > ago, we were considering to spend a little amount of time in writing a >> > HTML5 canvas for matplotlib, but we stopped because of (apart not >> > knowing how to interface with a server) doubts about matplotlib >> > structure. >> >> As of matplotlib-1.0, there is now an HTML5 canvas for matplotlib. I >> played around with it a bunch on Sunday. It'll be in Sage soon enough. >> >> >> >> > What I mean is that matplotlib is designed so that its >> > canvas is just translating a bunch of lines and points and other >> > graphical objects into something that is understood by the target >> > viewer. To enhance real interaction, IMHO, the best way would be to >> > pass to the viewer also an idea of the hierarchical structure of the >> > plot, so that the viewer by itself is capable of changing basic >> > properties like "axis visibiliy", "plot line colour", etc. At this >> > point, I hope we were wrong and that HTML5 canvas that has been >> > developed can overcome these problems. Anyway, I think that doing >> > everything on the server side and letting the client only plot the >> > received data may be too much communication overhead, while there are >> > a number of different javascript viewers which are pretty powerful and >> > fast. >> >> >> 3. I have talked with people about making a Matlab-clone-ish version >> >> of the Sage notebook. This would be web-based, but instead of feeling >> >> Mathematica-like, it would feel much more Matlab-like. Thoughts? >> >> > I think that SAGE-python can be easily accepted by Matlab users >> > because of intrinsic similarity of scripting language structure, >> > console interaction, and stuff like that. The problem is that Matlab >> > is very reliable for operations like vector manipulation (which >> > require additional interaction with numpy in SAGE), data analysis >> > (there are many potential toolboxes in scipy) and symbolic analysis >> > (for which SAGE is growing, but still very far from industrial-level >> > reliability). >> > On the filter design side, I agree that is very useful and often used, >> > and I can tell that scipy has the signal toolbox which incorporates >> > some functions to do this. I think that most of the engineering >> > appealing that SAGE can show is currently strongly supported by numpy/ >> > scipy power: if SAGE can be better integrated with them, and if we can >> > improve their functions, things will improve for engineers. >> >> > At the moment, there may be very little advantage of using SAGE >> > instead of plain python to interface with numpy/scipy, which are >> > anyway the core toolboxes needed. >> >> Yes, that's what engineers think. >> >> -- >> William Stein >> Professor of Mathematics >> University of Washingtonhttp://wstein.org > > -- > To post to this group, send an email to sage-devel@googlegroups.com > To unsubscribe from this group, send an email to > sage-devel+unsubscr...@googlegroups.com > For more options, visit this group at > http://groups.google.com/group/sage-devel > URL: http://www.sagemath.org > -- William Stein Professor of Mathematics University of Washington http://wstein.org -- To post to this group, send an email to sage-devel@googlegroups.com To unsubscribe from this group, send an email to sage-devel+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/sage-devel URL: http://www.sagemath.org
