15 minutes ago, Neil Van Dyke wrote:
> But I usually avoid SLOC metrics. Inferences are easily off by an
> order of magnitude. And that's before you account for behavior
> modified to play to the metrics (Hawthorne effect). But PHBs will
> draw fine conclusions from SLOC because, presumably, exaggerating
> the value of poor information is better than admitting that you
> really don't know.
>
> Quoting a line from "https://www.ohloh.net/p/racket/estimated_cost":
I think that they did a reasonable job of putting the disclaimers in
the right places. It seems clear that the intention is just a
semi-amusing show-off number.
After all, SLOC counts, or any other amount-of-code counts have
inherent bogosities, like the fact that Matthew invested a ton of work
for the 5.1 release which resulted in a significant LOC lossage...
And if you consider line changes or instead, then I have a few giant
commits which are almost straight regexp-replace scans, yet still
would be estimated by a good amount.
Speaking about code graphs, I think that the language graph has a few
interesting points:
* You can how in 2009 Matthew removed a good chunk of C++ code which
was refactored as Scheme code. You can also see that the amount of
added Scheme code is much smaller.
* Towards the end of the year there's a small step down in the Scheme
counts, which is probably when profj was removed.
* Next, there's the switch to Racket, and it's nice to see that the
new LOC continutes at the same count.
* And then there's the 5.1 jump, with a huge jump down in C/C++ code,
and a much smaller jump up in Racket code. This is much more
drastic than the 2009 rafactor, which probably corresponds to mostly
translating higher level C++ code to Racket in the former, and a
complete re-write in this one.
--
((lambda (x) (x x)) (lambda (x) (x x))) Eli Barzilay:
http://barzilay.org/ Maze is Life!
_________________________________________________
For list-related administrative tasks:
http://lists.racket-lang.org/listinfo/users
