[SciPy-user] Scipy / matplotlib to replace matlab and indexes

[John Pye]

So, what will 'they' say in return?

    * Where's the manual?
    * What is the one single file that I have to install?
    * Why do I have to pay to get the numpy manual?

I use python/matplotlib and to a lesser extent scipy very frequently but
I have to say that I don't think the user experience is quite smooth
enough for foisting on those poor undergrads just yet. I think that the
fragmentation of the documentation is just too much, as well as the fact
that a really important part of the documentation is non-free. I would
like to see a one-file installer for Windows (AFAIK this doesn't exist)
and big searchable .CHM file with everything you need. You have to think
about all those people who struggle through their first programming
course...

I'm curious as to whether SciLab was an option you've considered?

Cheers
JP


Nicolas Pettiaux wrote:
> 2007/5/6, Stef Mientki <s.mientki at ru.nl>:
>
> Thank you very much Stef
>
>   
>> I'm in a similar situation (Radboud university, Nijmegen),
>> and even worse some people want to change from MatLab to LabView,
>> which, for as far as I can see, is very bad for educational purposes
>> (and even more expensive).
>>     
>
> Indeed. Good luck.
>
> Here, I first introduced the idea to go from matlab to octave, keeping
> much of the syntax and programs but going in a direction where the
> students would not have to copy illegally the software to run it on
> their own computers, but then someone came with the idea that
> Mathematica should be tought instead. So I want to answer that if we
> consider to switch to something else than octave / matlab, python with
> its extension is the way to go.
>
>   
>> To make them like SciPy more than MatLab,
>> I'm writing a few shell programs,
>> that will make it more attractive.
>>     
>
> this is a vey good idea. I would very much like ot contribute (when I
> have the time and opportinity, that is not now unfortunately).
>
>   
>> I've made a first version of a general data acquisition / realtime
>> analysis tool (which btw also supports MatLab ;-)
>> http://oase.uci.kun.nl/~mientki/data_www/pic/jalcc/help/jalcc_data_acquisition.html
>> <http://oase.uci.kun.nl/%7Emientki/data_www/pic/jalcc/help/jalcc_data_acquisition.html>
>>     
>
> this is very nice? Is it a portable application that runs on Windows ,
> Mac but also GNU/linux ? (as python does)
>
>   
>> you can see a movie here
>> http://oase.uci.ru.nl/~mientki/download/medilab_tot.htm
>> <http://oase.uci.ru.nl/%7Emientki/download/medilab_tot.htm>
>>     
> Thanks. I'll do as soon as I have the time to do so
>
>   
>> And now I'm working on a Signal WorkBench,
>> a first demo (including comparison with MatLab) is described here
>> http://oase.uci.kun.nl/~mientki/data_www/pic/jalcc/help/jallcc_swb_filters.html
>> <http://oase.uci.kun.nl/%7Emientki/data_www/pic/jalcc/help/jallcc_swb_filters.html>
>>     
>
> a signal toolbox for scipy / numpy is a good idea
>
>
>   
>> Yes, indeed, starting with zero is inhuman ;-)
>> but,
>> - working with zero-based index, makes cicular arrays more easy to handle
>> - it's very easy to get the third element from the end   a[-3]
>> so not very strong arguments ;-)
>> And even harder to get used, is the end-index is not included, like in
>> MatLab.
>>
>> So I think you should use other arguments, like
>>     
>
> yes, I will, but the more and best I have, the better I'll feel !
>
>   
>> - it's easy to vectorize any function (to gain speed)
>>     
> is it different / better / simpler than in Matlab ? How do you proceed ?
>
>   
>> - all code is avaliable and can be changed (In MatLab, I often need to
>> run test series, to see what a function exactly did, because the
>> description was inadaequat, and MathWorks didn't want to tell me what's
>> inside, btw the standard answer from MathWorks "only available if you
>> manage to get a job at MathWorks" ;-)
>>     
>
> indeed. (as FLOSS)
>
>   
>> - SciPy is free, and therefor every student can take it home and study
>> it when it fits him/her
>>     
>
> indeed. (as FLOSS)
>
>   
>> - Python can be used to perform any computer task (general programming),
>> while MatLab is mainly pointed at math
>>     
>
> indeed. (as FLOSS)
>
>   
>> - graphical user interface with feedback are much easier to create than
>> in MatLab (ever designed an interactive plot in MatLab ?)
>>     
>
> very good point . I did not try and will try asap. Have you got a
> simple example accessible ?
>
>   
>> - Python can be used fully Object Oriented (MatLab not), which makes
>> extending code (in the future) extremly easy
>>     
>
> yes
>   
>> - in MatLab every function must in a seperate file (do you have also
>> directories with a few hundred very small files ;-)
>>     
>
> very good point
>
>   
>> - Memory footprint MatLab : 110 MB, Scipy 10 MB
>>     
>
> good point but I suppose they don't care.
>
>   
>> - Embedding and encapsultion of MatLab is very limites, Python is made
>> for it
>>     
>
> very good point
>
>   
>> - although more scientific problems are discussed in MatLab newsgroups,
>> the answers often only consists of the right direction. In Python
>> newsgroups (all that I know), give you support (and often more than 1
>> solution), until  your problem is solved (Thanks newsgroups !!)
>>     
>
> very good point
>
>   
>> - although the Python documentation is sometimes fragmented, the
>> available sources on the web and the available books are much better,
>> just compare the ratio of contents and number of pages of "Using MatLab,
>> version 5" and a standard Pythion manula like "Learning Python, O'Reilly"
>>     
>
> very good point
>
>   
>> - New versions of MatLab are full of bugs and often doesn't support
>> newer platforms, I've the feeling Python doesn't suffer from that.
>>     
>
> very good point
>
>   
>> - Function calls with many (named) parameters is much easier in Python
>> than in MatLab
>>
>> So aren't there any disadvantages ?,
>> Yes there are a few:
>>     
>
>   
>> - Python (Scipy) is fully unknown (so you must be a Don QuiChotte)
>>     
>
> one point is that a teacher of computing science is considering python
> to teach the basis of  programming instead pf the C++ that is
> currently tought in the first year ... and that is a real pain for the
> students, who do hardly remember anything when I see then in the
> second year where I teach.
>
>   
>> - the MatLab workspace is missing in Python
>>     
>
> yes, but I suppose that can be replaced with well configured editors
>
>   
>> succes with your attempt,
>>     
>
> thanks
>
> Nicolas
>