Potential bug: "The option '--plot' requires a parameter"
by A. T.
Dear all,
I obtain an error when doing the following steps:
Open some dataset
mark two variables, right click "correlation matrix"
tick the box "ensure uniform sample size" and click "ok"
and you'll receive error "The option '--plot' requires a parameter"
This is tested on the current git from yesterday on Ubuntu.
Best,
Artur
8 years, 1 month
Testing Gretl
by Hélio Guilherme
This is a kind of roadmap and status of the tests (Allin's):
- There is the Test project provided by Allin it lives at
https://github.com/HelioGuilherme66/gretl
(anyone can `git clone https://github.com/HelioGuilherme66/gretl.git`
<https://github.com/HelioGuilherme66/gretl.git> but you only need
`gretl-tests/test-gretl` ) (sorry I have messed up the repository)
- We run `make` to generate the newout directory with the new installed
Gretl, this is then compared with the output (from the baseline Gretl one
year ago).
- Attached is the generated diff file.
- We can do some raw inspection (no statistical tests on it ;) ). For
example I find really different the following result:
output/MK.out vs newout/MK.out:
46c46
< R-squared 0.557876 Adjusted R-squared 0.551707
---
> Uncentered R-squared 0.557876 Centered R-squared 0.154190
- The structure is only planned for Linuxes systems.
- There was no updates to the gretl scripts files,
- After the new release, we could do some maintenance of the gretl scripts,
and update the baseline. I will also do a release of the testspackage, and
so we can keep a progress of them (I am thinking in one year period).
- It is possible to do automated GUI tests, but please think well on what
tests are needed, because they are "costly" to do.
If you have concrete proposals, you can create an Issue or a Pull Request
(PR).
Or, just comment on this mailing list.
Thank you so much for this project!
(PS: Resending with CC just in case)
*My Favorite Open Source Projects awsome-lists
<https://github.com/sindresorhus/awesome> gretl
<http://gretl.sourceforge.net/> robotframework <http://robotframework.org/>
*(*sponsored/patrocinado*) Recomendo servidores e alojamento Web em:
http://www.proalojamento.pt/
<http://www.proalojamento.com/clientes/aff.php?aff=258>
8 years, 1 month
why no null default arg (non-pointerized) for matrices, strings...
by Sven Schreiber
Hi again,
so I've revisited the whole story about default function arguments. Some
time ago it was decided that for 'string' and 'matrix' types it is
possible to mark them as optional by appending '[null]', without using
pointers (which is still possible).
The same trick doesn't work for the array types 'strings' or 'matrices'
(or 'bundles', for that matter). Is there a specific reason?
To be clear, this does NOT work:
<hansl-fail>
function void ch2( strings ss[null])
if exists(ss)
print "exists"
elif !isnull(ss)
print "isn't null"
endif
end function
ch2()
</hansl-fail>
It would sometimes come in handy for optional array arguments.
thanks,
sven
8 years, 1 month
forthcoming release
by Allin Cottrell
Hello all,
In the run up to the 2017a release I ran my usual set of regression
tests and they turned up a few issues that needed attention, so I've
made quite a few small changes over the last several days.
Today I ran all my tests again (including testing all the function
packages and addons). Things look OK, and I've put up a new snapshot
which is in effect the release candidate. But I'd feel more
comfortable if other people could do some independent testing to
check that my "fixes" haven't broken anything.
I'm therefore proposing to wait until this Friday for the release.
If you can beat on the snapshot and report any problems before then
that would be much appreciated.
Allin
8 years, 1 month
function arg defaults reloaded (and other doc issues)
by Sven Schreiber
Hi,
several documentation-related issues, perhaps in time for the release?
First, relating to the User Guide section 13.4 (function programming
details):
- An old sin of mine: Here in this thread
http://lists.wfu.edu/pipermail/gretl-devel/2016-January/006447.html I
sort of promised to document the fact that you can give a plain
(non-pointerized) null default for a matrix argument to a function, but
I never did that. So the documentation is still not reflecting this. Or
more generally: A paragraph about matrix arguments is missing in that
section.
- It should be explained how to specify default arguments for int and
bool arguments (especially without the need for pointers).
- In section 10.8 there is a paragraph "Arrays as function arguments"
which IMHO should be moved into section 13.4, with only a reference
remaining in section 10.8. (In that paragraph there is a footnote about
arrays of lists which doesn't even exist in hansl, so I think that
footnote just causes confusion and should be simply deleted.)
- Similarly in section 10.7 there is an analogous paragraph "What are
bundles good for?", and I think that also belongs better in the function
programming section 13.4.
Then I guess section 13.6 about old-style functions can be removed by now.
Next, two issues with respect to the "Hansl primer" chapter 9
(user-written functions -- caveat: I'm looking at the June 2016 version,
maybe not up to date):
- Both the listings of possible return types and of possible arg types
are missing the array types "matrices", "strings", "bundles".
- There's a "Beware" pointing hand explaining that there are no multiple
returned objects in hansl. Here it probably would be good to repeat the
hint that pointerized arguments are another possibility to retrieve more
stuff.
Ok, so much for now. And no, I don't really have time to apply these
changes myself right now, because properly finding and disentangling
these issues was already quite time-consuming. Sorry.
cheers,
sven
8 years, 1 month
for translators
by Allin Cottrell
Late-breaking: I've been working to simplify and rationalize the
output from the "hausman" command (panel-data diagnostics following
estimation of pooled OLS) and I've changed several translatable
strings. The changes are just simplifications (dropping some
clauses, removing newlines) so hopefully it should be trivial to
update translations.
I mention this now because we're planning to put out a new release
within about a week.
Allin
8 years, 1 month
syntax coloring
by Sven Schreiber
Hi,
trivial glitches for the upcoming release:
- 'exists' as func name isn't syntax colored
- I think there was some other stuff as well, but can't remember right
now (perhaps somebody else has other things to add?)
Apart from that, I have an updated de.po ready, but have to remember the
right git commands, and also am waiting for an answer to my other
question about "spread (MIDAS)".
thanks,
sven
8 years, 2 months
Small session problem
by Marcin Błażejowski
Hi,
I've noticed that when I add bundle to session rename option doesn't
work (I mean the option works but when I close and open 'icon view'
window there is still old name).
And another thing: Allin, I know that sessions are GUI specific feature
but maybe we could _just_ save a session from the hansl level? Why?
Suppose I run big simulation and I do it not on my PC but remote
machine/server with a lot of CPUs, RAM and so on. And the only way to
access the server is SSH and '-X' doesn't work. Or '-X' works but
simulation lasts for hours...
Do you know what I mean?
Regards,
Marcin
--
Marcin Błażejowski
8 years, 2 months
VAR documentation
by Allin Cottrell
Following a suggestion from Sven not so long ago I've added a couple
of things to the VAR chapter of the User's Guide: an explicit
reference to the SVAR addon, and an account of the current state of
the diagnostic tests for VARs. If anyone would like to check them
over that would be good.
There's still a couple of "todo" sections in that chapter (as in
several other chapters, though the Guide is now 387 pages). If
anyone felt like having a go at filling in the blanks that would be
good too!
Allin
8 years, 2 months
