Sorry for (huge) delay, by the paper
for JSS consumed almost all my time. But for two weeks the paper
is "conditionally accepted" so it's high time to do some funny
work.
The process of package verification is very complex. I was
trying to start this work twice on summer and I found a lot of
things (beside the list I've posted to the Gretl-devel in the
email "Comprehensive quality control for function packages").
Let me explain just two main topics:
1. Tests
Here we have the easiest work: since certain test tests certain
thing, we can even run the script with procedure for testing the
tests. This is not very complicated and possible to workout in
few days.
2. Redundancy
Let me give an example: HEGY and DHF tests: do we really have a
redundancy here? Rather not, but in case of Jack's "Hegy" and
Ignacio's "GHegy" rather yes. So we need procedure for deciding
which one should stay or how to join them into just one package.
This is (theoretical) easy, but...
2* Some packages consist on (or in more precise word: use)
another package(s). So in my opinion in case of wasting the
redundancy (what is basic thing in informatics) we MUST write
wrappers.
What I (now) want to do: this is may last really busy week, so
up to 20th November I prepare draft of such procedure with my
proposals. In case of the Team I think that we could split the
work according to type of the package: tests, estimators and so
on.
Best Regards,
Marcin
W dniu 06.11.2014 o 17:52, Riccardo (Jack) Lucchetti pisze:
On Thu, 6 Nov 2014, Henrique Andrade wrote:
Em 5 de novembro de 2005, Allin
escreveu:
(...)
I won't go into all the specifics -- although I think you're
right -- but
rather I'll address the general issue
here.
We could really do with some editoral control over the
function packages.
Up till now we have been totally "laissez faire", but if the
packages on
the server are supposed to be a showcase for adding to
gretl's built-in
functionality they need to be pruned and consolidated. I
wonder if we could
assemble a committee of 2 or 3 members to work on this.
I would like to be one of them. I really like to test that
kind of stuff :-)
I believe that the Polish team has already done some preparatory
work in that direction and it would be a shame it that gets
wasted, so in my opinion the best thing would be to join forces.
If nobody minds, I'd like to put forward a proposal: Henrique,
Sven and someone else (nominated by Tadeusz, Marcin and Paweł)
become the package maintenance team, with the following
responsibilities:
* short run: classify packages with a set of criteria like the
ones we've been discussing so far, and make proposals about them
(also in the light of the expanding set of features that Hansl
has been gaining over the years). For example, Claudia's
clustered_ols is IMO quite valuable in terms of example script,
while for example I wonder whether we really need "cnumber" any
longer, since a possible Hansl equivalent implementation may
reduce to the 7-liner
<hansl>
function scalar cnumber (matrix X "set a variables matrix")
matrix A = X ./ sqrt(sumc(X.^2))
matrix En = eigensym(A'A) # compute the roots of AA
scalar cn = sqrt(maxc(En)/minc(En))
printf "Condition Number = %12.2f\n",cn
return cn
end function
</hansl>
On the other hand, more complex stuff like HIP may become an
add-on. Sven's previous message was a very valuable start at
this.
* medium run: verify that all the packages install and run the
provided example script on the widest possible assortment of
OSes (various flavours of Windows/Linux/OSX): perhaps, it'd be
nice to have someone with some OSX competence in the team, since
I assume the people I mentioned can patrol the Linux/Win
territory quite well.
Of course, I don't mean to force anybody to do things they don't
have time for or don't feel like doing, so I don't want anybody
to feel obliged.
-------------------------------------------------------
Riccardo (Jack) Lucchetti
Dipartimento di Scienze Economiche e Sociali (DiSES)
Università Politecnica delle Marche
(formerly known as Università di Ancona)
r.lucchetti@univpm.it
http://www2.econ.univpm.it/servizi/hpp/lucchetti
-------------------------------------------------------
_______________________________________________
Gretl-users mailing list
Gretl-users@lists.wfu.edu
http://lists.wfu.edu/mailman/listinfo/gretl-users
--
Marcin Błażejowski
GG: 203127