[Introspector-developers] Fwd: [intentionalprogramming] What IP is.

[James M. D. <mdu...@ya...>]

Here is a nice article that might be of interest to you 

--- techtrader_bob <f5w...@sn...> wrote:
> To: int...@ya...
> Mailing-List: list int...@ya...; contact
> int...@ya...
> Date: Wed, 04 Dec 2002 18:15:55 -0000
> Subject: [intentionalprogramming] What IP is.
> 
> My take on this:
> IP is infrastructure on which you build code generators.  Period.
> 
> Please read the following paper NOW! I added the homesite of this 
> paper to the 'Links' section of the group..
> 
> ftp://ftp.cs.utexas.edu/pub/predator/distil.pdf
> 
> Ok, now that you are back..
> 
> I think of IP as a glorified Question/Answer system along with a 
> database of code fragments.  The code fragments have attributes.
> It also has some stuff to visualize code (graphically, math symbols, 
> whatever) via attributes on an intention telling the IP system how to
> 
> display it.  You could display any code as Java, C, Math, arbitrary 
> bitmap, etc).  IP is infrastructure, not a language.  
> 
> Depending on the attributes, different code would be generated.
> IE for Sorts, and less than 1000 items, use a quick sort, for 1 
> billion items, generate code for a disk based B-Tree.
> 
> You need to state your INTENT to the IDE.  When it can not resolve 
> what you mean, you would actually type some code (c, java, etc) that 
> you want pasted in.
> 
> If you find a better way to do something, type in the code, then add 
> attributes to that code that would allow the IP system to determine 
> which code (old or new) that should be used.  With the above example,
> 
> say your program got a ton of new data.  You would code up the b-
> tree, and say for LARGE sets, use this.  Then in your Sort intention,
> 
> you would add the "LARGE" attribute/requirement.
> 
> Your code would then be transformed to use the new algorythm.
> 
> These code fragments could be entire components, subroutines, or even
> 
> lines of code.
> 
> The magic of IP is it knows how to weave it all together into a 
> cohesive functioning program.  This allows huge amounts of re-use 
> since you just have to describe the desired output.  If no code 
> matches, create some new attributes, and some new code, allowing it 
> to be wove in to the new enhanced codebase.
> 
> With regard to the sort, your code would be independant of the data 
> types.  Sorting integers, characters or names would be the same.  The
> 
> generator would look to the layer above to determine the datatype and
> 
> then generate code specific to that type.  
> 
> This quickly leads into a need for Refactoring, and code aspects, 
> hence simonyi's new company.
> 
> These ideas are shown in:
>  ftp://ftp.cs.utexas.edu/pub/predator/tr-95-26.pdf
>  ftp://ftp.cs.utexas.edu/pub/predator/tr-95-04.pdf
>  ftp://ftp.cs.utexas.edu/pub/predator/sigsoft-93.pdf  REALLY GOOD!!
>  ftp://ftp.cs.utexas.edu/pub/predator/wisr-93-p2.pdf
>  ftp://ftp.cs.utexas.edu/pub/predator/sigsoft-94.pdf  REALLY GOOD!!
> 
> 
> Hope this makes sense..
> 
> 


=====
James Michael DuPont
http://introspector.sourceforge.net/

__________________________________________________
Do you Yahoo!?
Yahoo! Mail Plus - Powerful. Affordable. Sign up now.
http://mailplus.yahoo.com