Re: st: simply start over and build something better

[Anders Alexandersson <andersalex@gmail.com>]

Another example besides Mata of an incremental
start-over-and-build-something-better implementation in Stata is the
change
from having common options to prefix commands. For example, the survey
manual was much bigger before the -svy- prefix was added.
I do not recall the exact history of the prefix commands. Did it start
with the -by- prefix and then came the -svy- prefix, and then came the
remaining prefixes?

Sometimes the best thing to do is to simply start over without
necessarily building something better immediately.
An example is the program for detecting multivariate outliers in
Stata, -hadimvo-, which StataCorp got rid of for version 8.
In the latest issue of Stata Journal we now have a user-written
command -mcd- which is written for version 10.

I kindly disagree with Martin's comment:
Don`t get us started on the 3D thing again! It comes up less and less
in "Wishes & Grumbles"...

That is, I think that less frequent wishes and grumbles for official
3D graphics capabilities do not indicate lack of interest on either
side
but rather a mutual understanding that it's difficult for users to
provide constructive suggestions to StataCorp on very large projects.

The distinction between starting over and making major improvements to
existing code can get blurred.
We could continue to try to find interesting historical examples
and/or take a more critical and future looking suggestive approach
like the blog entry on R.

Anders
andersalex@gmail.com

On Thu, Sep 16, 2010 at 2:27 PM, Joseph Coveney <jcoveney@bigplanet.com> wrote:
> Stas Kolenikov wrote:
>
> Came randomly across a blog entry on R:
> http://xianblog.wordpress.com/2010/09/13/simply-start-over-and-build-something-b
> etter/
>
> I wonder if Stata Corp. starts over and builds something better from
> time to time. The two major rearrangements of the code that I am aware
> of are the move towards Mata overt the last five or so years, and code
> multi-threading for MP flavors. Stata does not seem to suffer from the
> problems discussed in the above blog entry: the scope of everything is
> well defined in both Stata and Mata, and Mata calls by reference
> rather than by value (while Stata only deals with the data in memory
> plus a small assortment of locals... at least insofar you don't touch
> classes).
>
> --------------------------------------------------------------------------------
>
> The switch to the new graphics in Release 8 is about the closest thing to a
> start-over-and-build-something-better event that I can recall.
>
> Introduction of Mata seems more incremental.  I'm sure that there are resource
> constraints at play, too, but its seems primarily that StataCorp wants to take
> its time with the development of Mata in order to make sure that they get it
> right.  About the only thing that I hope StataCorp will do in the near term is
> to add two words to help m2_reswords: "interface" and "implements".
>
> Joseph Coveney
>
> P.S. I really like the new factor variables.  Brilliant.  I just wish that
> StataCorp would provide more coaching about how users can implement them in
> user-written modules.
>
>
> *
> *   For searches and help try:
> *   http://www.stata.com/help.cgi?search
> *   http://www.stata.com/support/statalist/faq
> *   http://www.ats.ucla.edu/stat/stata/
>

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/