begin quoting Andrew Lentvorski as of Fri, Mar 28, 2008 at 06:01:14AM -0700: > [EMAIL PROTECTED] wrote: > >At an old job the resident programming guru was said to be amazingly > >prolific. It was said he could work over a weekend and come back > >with 6000 lines of code.
Measured by wc -l? Easy. Measured by counting semicolons? Open/close braces? Method calls? Assignment statements? Possible. > >All my code has always been slow and steady. Just the design phase takes > >lots of calm meditation to find the right way to do things. It is *never* > >just a crazed ejaculation of 100s of lines of working code. > > > >Was this a myth? I'm skeptical anyone could even type 6000 lines in > >a weekend even if they had the source print out in front of them to > >enter?!?! Use a better editor. > It's probably not a myth. > > The issue, however, is that new code creation generates far more lines > of code than any other task. Yah, is it 6k lines of new code, 6k lines of tested code, or 6k lines of debugged code? > Taking a look at some of my projects, I seem to be able to generate > 1500-2000 LOC in a weekend. I can easily see 3000 with unit tests which > tend to be cut'n'paste. > > Then, I start debugging and my LOC productivity drops like a stone. Ain't that the truth. I often end the day with a negative LOC count. And LOC is a lousy measure anyway, because cut-and-paste is a REALLY easy way to get that number up there. (As is code generation -- CORBA, RMI, lexers/parsers, etc., all can really inflate this sort of thing.) One of my most "productive" days ended up with ~15 lines of new code, ~30 lines of modified code, and ~2000 lines of deleted code. (The code I was fixing was the result of progressed being measured by the number of new lines of code added by each programmer.) Ideally, we'd like to compare "completed features". -- If you impose a metric, programmers will code to fulfill the metric. Stewart Stremler -- [email protected] http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-lpsg
