I just did some timings on a trivial program
wrapped around
di 42
and the same program with 10,000 comment lines
* a comment
added. The time taken by the two programs
is indistinguishable. Running the two 10,000
times each produced the interesting result
that the program with comments ran faster...
I don't believe this, but I do assert that
-- whatever Bill said, and I'd like to hear
the context -- worries about
speed should _not_, usually, inhibit users from
adding comments to taste. That doesn't rule
out minor but possibly useful tricks such
as not putting comments within loops, but
outside loops. (Thinks: I probably did that
somewhere....)
Nick
[email protected]
David Airey
> Yes, I remember that too. I think it was the case that comments after
> the program did not slow it down, because nothing was
> interpreted after
> end. I think perhaps this means the run copy could be up top and the
> commented (documented) version down below or something like
> that. There
> are numerous texts (one good one by Microsoft) on the value of code
> documentation. For the in house purposes of StataCorp, their level of
> ado documentation must be adequate, otherwise they would be
> inefficiently having to figure out the code and the fix. My guess is
> that documentation would serve the user community more than StataCorp
> by improving external programming skills. But, that's why we
> pay for a product. Maybe the structure of the ado code should have this
> formalized.
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
