RE: Unit validation in spreadsheets

["Paul A. Steckler" <steck stecksoft com>]

> Certainly.  As previously mentioned we've got small versions 
> of that sort of functionality for the autoformater.  We also 
> had a recent discussion on how to extend the MS XL style 
> format strings to handle SI (metric) unit prefixes.  We'd 
> need to be able to disable the checking to keep users 
> unfamiliar with the process comfortable, but I doubt that 
> would be onerous.

Sorry, I'm new to the list, so I wasn't aware of that 
work.

The way we designed it, the inferencer works with 
any units at all (physical units, financial units, what have you),
as long as their names are used consistently.  

In our Excel prototype, rather than put the units in format strings, we 
put them in annotations.  That's OK, but then the annotation 
fields aren't available for other duty.  It would be better 
to have separate fields for unit annotations.  Having units 
in format strings might tend to clutter things.   

Also, because some units are inferred, it might be confusing 
to see units in the formats of some numbers, but not others.
For finite formulas, our prototype only required unit annotations 
at the leaf cells.  All other units are inferred.  The main 
technical contribution that we made (IMO) was the technique 
we devised for inferring units for formulas with circular 
dependencies.  Does the system there now handle circularity?

> > What would be the best way to get that done?
> In what sense ?  Having a clear design of how the feature 
> interacts with other features from a user perspective would 
> be a good starting point.  It would greatly decrease the 
> number of development iterations.  Now is a good time 
> development wise to work on this sort of thing, we'll be 
> releasing 1.3.x tarballs for the next few months.  So there 
> is room to develop.

I'd certainly like to participate in the design of a unit 
inferencer; some of the others who worked on our Excel prototype 
-might also want to do so.

Myself, I'm probably unable to devote enough cycles to do 
the bulk of the coding (although I can probably do bits of it).
My hope is that someone in the Gnumeric community might be interested 
enough to take that on.

-- Paul