Funny that. Was thinking smth along those lines the other day, as I was wrestling with my own spreadsheets, I was considering how ML-type languages would be ideal for spreadsheets.
A column could be of type:
type itemCode = string type itemName = string type itemMaker = string type itemPrice = int type itemPurchase = int type itemcolumnType = (itemPurchase * (itemCode * itemName * itemMaker) * itemPrice) list
Types can add dimensions to spreadsheets. The path not taken saves you from spreadsheet hell. You can only look at so many little rectangles at a time... Data in a spreadsheet should fit into your visual system. Otherwise, it requires magic. And, if you do magic -> magic -> magic, which is what statisticians and data mining specialist do, you end up with a simple graph.
Thinking about it, you can see that Google Docs kinda sucks...Can I tell my spreadsheet to sort every other row in a column, according to a regex, fuse it with data from another one and make it generate me a report in Teh Eevil .doc format? Oh, the Python API you say? Hmmm. I thought Google was all about clickety click and a "simple interface". Yahoo Pipes fares scores higher here, except for the fact that they like to be stuck in a corner wearing the funny pointy hat.
Where's the innovation? With all the money Google's got, you'd expect a little more.