[AstroPy] Books on pylab, scipy, ...?

Johann Cohen-Tanugi cohen@lpta.in2p3...
Fri Feb 19 17:36:16 CST 2010

Wayne Watson wrote:
> Thanks very much for your description.  From what I can pull out of 
> the book via Amazon makes me think it would be more useful than 
> anything I've found on the web.  I'm very familiar with math and stat 
> methods, so that's of minimal interest to understanding how the 
> packages plan together.
> To tighten up my knowledge on these subjects, I plan to go to a 
> monthly PUG meeting San Francisco Bay Area next week to talk to live 
> people about these matters and others. It's 180 miles from here, but I 
> still have family and friends there, so  problem with distance.
> What prompted me to post this message is matplotlib.  I began 
> earnestly considering its plot facility, plain plot (line plot) in 
> pyplot, about 2 weeks ago. I immediately got stuck on what I would 
> call the weird behavior of show(). In time, I got that under control 
> and discovered some highly usable parts of matplotlib. I am still new 
> to it, but see the benefits.
> About two  days ago, I looked at the plot and decided it needed to 
> have a large circle drawn around it. The plot is the somewhat zig-zag 
> plot of a all-sky camera I use, and the data captured by it, track 
> points. The circle would represent the horizon that camera sees. As 
> far as I know, there's no way to draw a circle, except one. Put a 
> large marker at the center point of the pixels where the zenith would 
> be. It'll work, but left me thinking why I couldn't just draw it in  
> some "normal" way. That made me wonder if I was seeing the whole 
> picture of scipy, pylab, and so on. However, I've begun getting 
> familiar with them now.
matplotlib has patch objects for common geometrical figures. You have to 
pass them to a "axis" object as patch objects. So for instance the 
following should work :
from matplotlib import pylab as plt
from matplotlib.patches import Circle
cir = Circle( (0.5,0.5), radius=0.2)

a bit contrived, so there might be more direct ways to get it....

> Regarding MatLab, I happen to have a 5 year old student copy that I 
> have barely used, since I purchased it. It's given me some insight to 
> MPL, but I have no plans to dig into ML. Python is quite suitable.
> Another tact I plan to use is to review  OOP inheritance in Pyhon. 
> It's been a while since I used OOP, and I think there's a key there to 
> understanding how these facilities work together.
> On 2/19/2010 11:53 AM, Johann Cohen-Tanugi wrote:
>> hi Wayne, what you ask is hard to achieve for any author. First of 
>> all, while matlab is a sort of testbench of what things could look 
>> like and has been at least partially driving some considerations in 
>> matplotlib and numpy/scipy, the latter do not intend to "work like" 
>> matlab, nor to offer the same level of integration. So first of all, 
>> is you want a high level of integration within one open source 
>> product, check out octave and/or scilab.
>> Second, matplotlib is a very large OO libraries, and you can 
>> manipulate the objects in the libraries within python. A subset of 
>> matplotlib then strive to provide high level single-command interface 
>> to this OO... this is what you call pylab I believe... With pylab you 
>> have access to easy commands for all the normal call to plotting, 
>> with a 'look and feel' somewhat equivalent to MATLAB.
>> Third, scipy is a "toolbox" or a "toolstack". It starts with the 
>> numpy core library, and provide a series of libraries for further 
>> scientific computing, alwaus striving to make use of numpy 
>> underneath. These includes integration, ode and pde solver, root 
>> finding, some multivariate analysis tools, etc... What is or should 
>> be in scipy is not clear and is subject to many threads in the 
>> scipy-user and scipy-dev mailing lists. In that respect, the role of 
>> the scikits separate modules is not clear either. In that condition, 
>> you will have a very hard time finding a scipy dedicated book. What 
>> you can find is books on numerical recipes in python, probably using 
>> numpy, and describing implementations that might or might not already 
>> be in scipy. The next stage is then, for a given problem you are 
>> interested in, to ask the scipy mailing list if there are tools in 
>> the toolstack that could help you reach your goal.
>> When scipy matures more, then maybe it will have a shape and a 
>> definiteness that will warrant a dedicated book. I do not think that 
>> we are there yet.
>> HTH,
>> Johann
>> Wayne Watson wrote:
>>> Perhaps there are books on Subject and similar topics? I know 
>>> there's one on matplotlib. "A Primer on Scientific Programming with 
>>> Python" is somewhat along these lines. I decided to use the "Inside 
>>> Search" for books on Amazon. numpy is well represented, and 
>>> apparently quite a few math methods. When I searched for scipy,  it 
>>> found a number of pages, but not one of the pages would be 
>>> displayed. The entirety of books there is limited to maybe 1/2 a 
>>> book. I'm more interested in how pylab, scipy and matlab play together.
>>> -- 
>>>             "There is nothing so annoying as to have two people 
>>>              talking when you're busy interrupting." -- Mark Twain
>>> -- 
>>> This message has been scanned for viruses and
>>> dangerous content by *MailScanner* <http://www.mailscanner.info/>, 
>>> and is
>>> believed to be clean.
>>> ------------------------------------------------------------------------ 
>>> _______________________________________________
>>> AstroPy mailing list
>>> AstroPy@scipy.org
>>> http://mail.scipy.org/mailman/listinfo/astropy

More information about the AstroPy mailing list