I wrote a simple text parser which understands semi-free text like

2013-10-17

10:00

$P_1$ +30m reading lesswrong

$A_1$ +1h project A: implementing feature 1

includes discussion with Pete

12:00

-15m lunch break

+15m dicussion about new feature 2

This allows me to take notes with extremely low overhead. I just switch to my programmers editor and write down a few more lines. No opening window, not clicking or tabbing thru redundant fields.

And I can do so incrementally. The $P_1: token assigns the cost center and I usually assign these at some later time batching related items quickly and possibly creating the cost centers after the fact - which is usual at the beginning of a project. The cost token is looked up via a comparable file describing the hierarchical structure of cost centers.

When I started this it was intended to just track time for myself, but it developed into a tool which can push the data into Jira (via REST) and Kimai (via SQL) and export selected sub hierarchies as TeX reports for invoices.

But it is still a simplistic undocumented CLI tool run from my IDE with lots of idiosyncrasies, so I wouldn't recommend it for anybody else.

I believe that it (i.e. the use of plaintext) is a lot more efficient and powerful than a spreadsheet.