Services & Resources / Wolfram Forums / MathGroup Archive
-----

MathGroup Archive 2008

[Date Index] [Thread Index] [Author Index]

Search the Archive

Packages without packages

  • To: mathgroup at smc.vnet.net
  • Subject: [mg86836] Packages without packages
  • From: AES <siegman at stanford.edu>
  • Date: Sun, 23 Mar 2008 01:00:38 -0500 (EST)
  • Organization: Stanford University
  • References: <fro3a6$i8l$1@smc.vnet.net> <frqpnm$505$1@smc.vnet.net> <fs26nq$6r$1@smc.vnet.net>

In article <fs26nq$6r$1 at smc.vnet.net>, in the "Saving Packages" thread,
 "alexxx.magni at gmail.com" <alexxx.magni at gmail.com> wrote:

> Yet I'd like to explain better what is my larger problem is:
> to break down a big project in pieces, it is correct then to save big
> chunks of code definitions as packages?

In my musing on packages, I've been thinking that I tend to have a fair 
number of individual projects on different physical topics, which I keep 
in individual folders or directories, and each of which tends to have 
its own set of project-specific modules or packages.

And, my packages in fact consist almost totally of just Module 
definitions, sometimes lengthy, which implement some complicated 
computational algorithm or create some plot or table generally specific 
to that project.

So, why futz with the complexity of Mathematica's global package 
mechanism at all?  How about a "packages without packages" approach.

For each project, I typically have one or a few project notebooks which, 
essentially, explore physics questions and prepare reports or memos or 
documents that I file or send to others (and in which I may want to 
present a lot of graphic or tabular results, with a lot of associated 
commentary, but very little detailed code). 

So, why not associate with each project a ProjectModules.nb notebook 
which contains nothing but the definitions of any major  or lengthy 
Modules or := formula definitions tht I use in that project's notebooks. 

This ProjectModules.nb notebook **can reside right in the folder for 
that project** -- in fact, it can be open all the time I'm working on 
any of the notebooks for that project, with its window minimized.  

If I decide I want to reformat a graphic that is created in one of those 
project notebooks by a certain Module, I can open ProjectModules.nb 
window, do the editing of that Module, and re-execute the whole notebook 
very quickly (it doesn't _do_ any work when re=executed; just redefines 
stuff).

Shifting over to work on a different project?  Just open the folder for 
that project; restart Mathematica to clear out all the sludge from the 
previous project; execute the ProjectModules.nb in that folder; and 
start on whatever other notebook(s) that use those modules.

I'm not knocking that basic packages concept, for those who may need it.  
But for me, "packages without packages" may offer:

1)  Modularity
2)  Simplicity
3)  I always know what's going on.
4)  Freedom from cross-project confusion
5)  Freedom from the complexities of the whole package system
(and from Wolfram's compulsion to keep changing it)
(and from their inability to document these changes usefully)


  • Prev by Date: Re: ListPlot Joined/Filling bug?
  • Next by Date: Re: Problem involving NDSolve
  • Previous by thread: Re: Grabbing a value from a Dynamic Object
  • Next by thread: Mathlink: How do I pass arbitrary data from Mathematica to C?