floating point number

[mattn]

How about support 0.9 to 1.0 with increment?

[tpope]

Sounds pretty complicated, once you take it to its logical conclusion: -1.00000 => -0.99999, for example (and that's not even counting the inevitable feature request for an option to omit the leading zero there). Plus it conflicts with versioning schemes (2.9 => 2.10), and needs to tread lightly around certain date formats (03.09.2012). And what happens if we invoke it before the decimal point? What happens if we invoke it on the 2 in 1.20?

The short of it is this feels like a lot of complexity for very little gain.

[mattn]

Sounds pretty complicated

Yes I think so. I could't write patch easily.
Currently speeddating has d<c-a> like key mappings to differentiate something. How about f<c-a> treat floating point value?