transparency (just thinking out loud)

[Alan G Isaac <aisaac american edu>]

My experience with spreadsheets is very out of date (I see
them mostly when my wife brings home some of her work or
students bring in their data) but I am wondering what steps
have been taken to produce more transparency.  My interest
has increased because of the rumored imminent Windows binary
of gnumeric, which will remove my last excuse not to teach
my students about spreadsheets.

This particular question came up because I had just finished
writing a gawk script to manage my student grades when my
wife asked me to look at a spreadsheet she was working on.
It was a startling reminder of how concealed the relations
are in spreadsheets, at least in my limited knowledge of
them.  (I know there is a literature on how this contributes
to human error.)

Naturally I wished for something like the following, I
wanted to open a page in the spreadsheet software (call it a
"report preparation page" or RPP) that is just space to
*visibly* define functions and variables that could then be either
i. referenced from a sheet, 
or possibly,
ii. directed from the RPP to a cell range in a sheet
By *visibly* define, I mean that the RPP is essentially 
a text editor window (*not* another sheet).

The difference between a RPP and say a VB program to
manipulate an Excel sheet would be
i. integration (the RPP would be part of the spreadsheet),
and
ii. specialization and simplicity of the syntax (which I
imagine to be roughly "matrix" oriented using very natural
syntax---say more like the commercial package GAUSS than
the commercial package MatLab or, to get closer to the
project, than Dadisp's Series Processing Language).

E.g., I imagine assigning to a variable a "matrix" based on
a rectangle in a sheet (possibly transposed), doing
operations on this "matrix", and assigning results
either 
i. to variables that could be then accessed from the sheet,
or even 
ii. making assignments from the RPP to actual locations in
the sheet

Does this functionality already exist?  If not, I (no doubt
naïvely) think it is a way that gnumeric could truly
distinguish itself.

Alan Isaac