On Thu, 30 May 2019, Fred Engst wrote:
I love gretl so much, both for my own research and for teaching.
I wish I know how to correct the following minor errors on my own,
but I don’t, so I’m saving them for you. :)
Thanks for the report. The more of these little things we can get
right, the better!
1) Spelling error in file “csvdata.c” at the very end.
When I join a dataset, if there was no change, the returned message
is then: "No changes were nade to the dataset”.
OK, Jack has already handled that one.
2) when I select “help->keyboard shortcut”, an error
error 56 (Failure when receiving data from the peer)”
The "core" gretl PDF docs are included in our packages for Windows and
Mac, but the others get downloaded on demand, and apparently that
failed for you. It worked fine for me this morning; I suspect it was
just a transient internet thing.
3) Once I save and close a model in icon view with a non default
name, the next time when I open it, it doesn’t show the model name
on the tab.
Yes, in the tabbed model window a user-specified model name is shown
in the model's (sub)window, but the label on the tab is just "model
<ID number>". This was a design choice, (a) as a simple means of
ensuring uniqueness and (b) to avoid having overly long model names
mess up the tabbing apparatus.
4) Issue with graphic output from “View->Multiple graphs->X-Y
scatters” or “View->Graph specified vars->X-Y scatter” : Font change
in graph has no effect after copying it to other apps, such as
powerpoint or word.
I see what you mean and have tried to fix that in git.
5) Sorting is not working correctly on some dataset’s variables,
such as Mroz dataset’s wage or lwage. Even $hat from regression that
I saved on this dataset wouldn’t get sorted correctly either. I have
not noticed this issue on other datasets, however.
Like Jack, I can't replicate that. Could you give a very specific
recipe for producing an incorrect sorting (and say in what way it's
6) Some models in the Icons view tends to be hidden if I gave long
model names. When I try to put some models in "Model table”, this
can be a problem.
Long names can be a problem in many contexts[*].
7) Speaking of Model table, can’t we let it take models with
different dependent variables? All we need to do is to list the
dependent variable on top of each column. Am I right?
Yes, it's possible, but I can't rate it as a high priority. The common
use case for a model table is comparing specifications for a given
8) Model table’s outputted asterisk is limited to 2, which can be
different from the original model’s output.
That was intended to conserve horizontal space. But we might as well
allow up to three asterisks for consistency with regular model output;
that's now done in git.
[*] Long names are a real problem for interface design. We can't let
them run into each other on pain of producing incomprehensible output.
So we must either truncate (which doesn't look good and likely
produces non-uniqueness) or separate output elements horizontally by a
large amount, introducing tons of whitespace. This gives output with
poor information-density and, in the GUI, leads to windows that are
too wide and/or have to be scrolled horizontally.