[Opencfd-develop] scripting vs library

[Jinming X. <jin...@ms...>]

>From: Anton Kulchitsky <an...@ku...>
>To: Jinming Xu <jin...@ms...>
>Subject: mail list etc.
>Date: Tue, 24 Aug 2004 13:15:39 -0800
>
>Jinming,
>
>>(1) Regarding the GUI, you can discuss with dano, one of the developers of 
>>OpenCFD. You two can take charge of it. I have no strong opinion of any 
>>particular GUI. What I said may not be correct and I don't stick with it. 
>>Let the hands-on developers decide.
>
>Do you have a mail list? Probably, we could discuss it together.
>
>>
>>(2) Regarding the stand-alone and library, you are only partially right. 
>>The stand-alone you are talking about is not programmable. The new version 
>>of OpenCFD will be programmable, or say scriptable. So it's not merely a 
>>stand-alone, it's also a server, or a library. You can use simple commands 
>>(macro) to control it. All the commercial CFD software is like this.
>
>I mostly agree. However, it is hard to develop a good script (or find a 
>good one, that is IMHO better). Even more however ;), we are very different 
>from commercial software. They chose this way (of stand-alone monsters with 
>OS like behaviour) mostly because they are "close-source" and it is much 
>more convenient for them to give some kind of enviroment for users. For an 
>open-source projects it is very big advantage if the project is a library 
>(open!) plus some gui scriptable interface above the library, some kind of 
>front-end. Thus, we leave a choice for users what to use.

Why do you think there is no good script to choose? From Tcl, Python to 
Scheme, these scripting languages are widely used in commercial as well as 
open-source codes. I once saw some open-source CFD codes use scripting 
languages too.

The big advantage of scripting is the easy of use, both in terms of the 
labor involved and the skills required. In the sense of skills involved, 
library approach will be very hard for users who are not developers. Without 
scripting, you have to compile each time for each job, but with script, all 
this can be done in a single script without compilation.

Here is a sample script for OpenCFD 0.3 to generate a grid:
# --------------- Tutorial Script ------------------------
# A very preliminary script for OpenCFD to generate a grid
# For more info, please refer to the doc or visit http://opencfd.sf.net
# All Rights Reserved!  Jinming Xu jx...@op...
import geom
geom.get("rub.dxf")
geom.shrink([6.0])
import grid
grid.cart(30,25)
grid.addExtrLayer(3,0.8)
grid.save("2d-bas.cgns")
#--------------------end---------------------------------------------

>That is why I am talking about free library rather than an application like 
>Matlab or Mathematica. I think Octave is not succeeded very much because 
>they tried to be similar to Matlab. But GSL (GNU scientific library) 
>succeeded much better due to they are different but still cover same tasks.
>
>
>I do not mean that my opinion is better. It is my point of view as a user 
>of a library and model developer. I think this is a kind of "UNIX-way": we 
>have components (libraries) and we have fron-ends and UI if users need 
>them. This is something different from commercial way, but it looks more 
>attractive for me. Finally, this software is for advanced users or even for 
>programmers (script or FORTRAN programmers at least), so the library could 
>be very attractive for them still.
>
>>
>>(3) Yes Latex files are much better than .docs. We can made this change. 
>>Actually all the doc included in the tar ball is in latex.
>
>Sounds just fine.
>
>Anton

_________________________________________________________________
Express yourself instantly with MSN Messenger! Download today - it's FREE! 
http://messenger.msn.click-url.com/go/onm00200471ave/direct/01/