There is a bundle of different issues here. Let me
try to clarify, especially for anyone trying to
follow along.
1. The guessing is about -outreg2- because I haven't
tested it. I don't use it; I just looked at the code.
2. There is no guessing about -index()-. This
is documented. Richard wants a list of old stuff:
the list under the help for -version- is long
and detailed. In about 15 years of Stata programming
I doubt that I have looked at it, or its predecessors,
more than a very few times. If you use -version-,
you usually don't need to. That doesn't provide all
he wants, but I imagine the explanation really is as simple as
the fact that the manuals are already pretty long
and only a very few people want to know about all
the old stuff too. As each new set of manuals arrives,
I always keep the old ones.
3. Let me emphasise here that the bugs being talked about
are those of user-programmmers,
Roy Wada for -outreg2- and my sparring partner Thomas
J. Steichen and myself for -concord-. Bugs bite democratically,
the just and the unjust, the young and the ancient, and so
forth. And StataCorp do warn, in every manual, to go back and
put -version <whatever>- at the top. It's true.
4. What happens to -version- when one program calls another?
I used the term "subroutine" informally (or sloppily, if you prefer)
but Stata, unlike say Fortran, has no formal term
"subroutine". The situation is this: one program calls
another. Nothing else is important here.
Whether they are in the same file, or whether
you, the Stata user, think about them as being
bound together in a family relationship, is not
important. That's a matter of your convenience or
your way of thinking about it, but your Stata executable
is, as you will agree, essentially unaware of any of that.)
Richard's guess, as I understand it, was that when a program
calls another, the -version- set is passed on (unless it is
was explicitly set to something else inside).
In the case of -concord-, it is the main program that
lacks a -version- statement. So, the program runs under
version 10 in Stata 10.
In the case of -outreg2-, something else is going
on, as the main program is -version 8.2- and
calls subroutines, one of which lacks a -version-
statement. Richard raises a good point
and I don't quite understand this. The only oddity
that springs to mind is that -program outreg2-
is _not_ the first program defined in outreg2.ado.
At one time, if not now, there could be puzzling
consequences if that were so. Perhaps StataCorp
can elucidate.
Nick
[email protected]
Richard Williams
> At 12:13 PM 8/1/2007, Mentzakis, Emmanouil wrote:
> >As from Stata 10, -index()- is only
> >available under version control.
> >
> >It so happens that at least one subroutine of -outreg2- lacks a
> >-version- statement by a small accident. Thus the program is
> being run
> >under Stata 10. My guess is that it needs a -version 8.2- statement.
>
> That is my guess too. As I understand it, -index- got renamed in
> Stata 9. Stata policy is to keep the old names working for one
> version, and then they stop working the version after that, unless
> you use version control. So, that is why -index- still worked in
> Stata 9 but not in 10.
>
> To clarify though: does the -version- command only apply to the
> subroutine that contains it? i.e. you can't just say version 8.2 in
> the first routine, it needs to be in every subroutine as well? I
> thought you just had to do it the first time, but apparently that is
> not the case, you need to do it in every subroutine.
>
> As a sidelight, I wish the old names just kept working, and/or that
> there was some nice master listing of all the old names and their
> modern counterparts. When going through older programs, it can be
> difficult to figure out what a command does because, while it works
> under version control, it is no longer documented.
*
* 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/
