6:51 a.m.
Dear Sven
I simply deleted a couple
of lines with the data file
Everybody have already seen
the answer coincided
So, only simulated data are left
No it's gfn rather than zip
Oleh
2:49 p.m.
Am 02.04.2016 um 12:51 schrieb oleg_komashko(a)ukr.net:
Dear Sven
I simply deleted a couple
of lines with the data file
Everybody have already seen
the answer coincided
So, only simulated data are left
No it's gfn rather than zip
Thanks Oleh, I have approved it and also removed the old .zip version.
Nevertheless, I have some comments, in case you're interested:
- The date of the package still says "2015-12-22".
- There is still some misspelled "heterockedasticity" in the output.
- In the help text the function "GUI_qgtest()" is mentioned ("qg"
instead of "gq").
- And more substantially: I don't quite understand why "sby" is a list
and at the same time you explain it should only be a single variable. It
seems to work, otherwise I wouldn't have approved it, but it feels strange.
cheers,
sven
6:06 p.m.
Thank you,
1) I'm sorry some times ago
I copypasted "ck" from
the package description to
all printing inside functions
I copy-pasted code to
a text editor and checked
I hope, there are no more "ck"
2) > I don't quite understand why "sby"
The default!
Compare
function void dfdf (scalar x, list y[null])
print x
end function
dfdf(1)
AND
function void dfdf2 (scalar x, series y[null])
print x
end function
dfdf2(1)
3) The date (according to my time zone)
Oleh
3 квітня 2016, 21:49:52, від "Sven Schreiber" <svetosch(a)gmx.net>:
Am 02.04.2016 um 12:51 schrieb oleg_komashko(a)ukr.net:
> Dear Sven
> I simply deleted a couple
> of lines with the data file
> Everybody have already seen
> the answer coincided
> So, only simulated data are left
> No it's gfn rather than zip
>
Thanks Oleh, I have approved it and also removed the old .zip version.
Nevertheless, I have some comments, in case you're interested:
- The date of the package still says "2015-12-22".
- There is still some misspelled "heterockedasticity" in the output.
- In the help text the function "GUI_qgtest()" is mentioned ("qg"
instead of "gq").
- And more substantially: I don't quite understand why "sby" is a list
and at the same time you explain it should only be a single variable. It
seems to work, otherwise I wouldn't have approved it, but it feels strange.
cheers,
sven
3:28 a.m.
Am 04.04.2016 um 00:06 schrieb oleg_komashko(a)ukr.net:
2) > I don't quite understand why "sby"
The default!
Ok, I see. In that case you could use the pointer form
"series *sby[null]", as described in the manual. But since I assume that
you know that, I guess you don't like the pointer form?
cheers,
sven
3:36 a.m.
Dear Sven,
Thanks!
It is my gap in hansl!
I have an excuse: I'm with
hansl for less than a year!
I think, I'll try to change,
but not right now!
Getting approving packages inspires!
Yesterday I make a function for
Nelder - Mead. Hope, it will help
to make some models, which
require non-smooth optimization
Oleh
4 квітня 2016, 10:28:03, від "Sven Schreiber" <svetosch(a)gmx.net>:
Am 04.04.2016 um 00:06 schrieb oleg_komashko(a)ukr.net:
>
> 2) > I don't quite understand why "sby"
> The default!
>
Ok, I see. In that case you could use the pointer form
"series *sby[null]", as described in the manual. But since I assume that
you know that, I guess you don't like the pointer form?
cheers,
sven
1:04 p.m.
New subject: unintended consequences of .zip -> .gfn transition
Dear Sven,
I tried to install gqtest at the department and
GUI doesn't worked before I saved null
change to the spec file: It used to say
"The package should be in separate folder"
(sic, it was Windows)
Now I'll try to make some corrections
and try at another OS
Oleh
4 квітня 2016, 10:28:03, від "Sven Schreiber" <svetosch(a)gmx.net>:
Am 04.04.2016 um 00:06 schrieb oleg_komashko(a)ukr.net:
>
> 2) > I don't quite understand why "sby"
> The default!
>
Ok, I see. In that case you could use the pointer form
"series *sby[null]", as described in the manual. But since I assume that
you know that, I guess you don't like the pointer form?
cheers,
sven
4:45 p.m.
New subject: gqtest 0.17: 0.19
Dear Sven
The GUI bug is corrected:
I rebuilt the package from
the script with no changes w.r.t. 0.18
If you do not object,
I am not very inclined to
change to *series, since
the behavior of &const
is to special, and practically
the half of the package
has to be rewritten without
no visible external changes.
It consists of 3 lines of formulas
and 300 lines of "if"s, which
have to be remastered
Oleh
4 квітня 2016, 10:28:03, від "Sven Schreiber" <svetosch(a)gmx.net>:
Am 04.04.2016 um 00:06 schrieb oleg_komashko(a)ukr.net:
>
> 2) > I don't quite understand why "sby"
> The default!
>
Ok, I see. In that case you could use the pointer form
"series *sby[null]", as described in the manual. But since I assume that
you know that, I guess you don't like the pointer form?
cheers,
sven
6:57 p.m.
New subject: gqtest 0.17: 0.19
Am 04.04.2016 um 22:45 schrieb oleg_komashko(a)ukr.net:
Dear Sven
The GUI bug is corrected:
I rebuilt the package from
the script with no changes w.r.t. 0.18
If you do not object,
I am not very inclined to
change to *series, since
the behavior of &const
is to special,
It's your choice as the package author -- and I have to admit/agree that
having to use the pointer form just to make it possible to set a default
value of no argument doesn't feel very logical. (Cf. the situation with
string arguments where it's possible without pointers.)
The other thing, what you call GUI bug, I haven't really understood, but
it doesn't matter.
I'm going to look at it tomorrow (I hope).
thanks,
sven
1:36 a.m.
New subject: gqtest 0.17: 0.19
Dear Sven,
Since gretl have no named parameters,
I do not like things like ffff(x,null,null,null,1,1,1,1,null,y)
That's why, I tried to as littele arguments as
possible, to this end I used variant with "const"
with a special meaning. I know how to
make the same with *series argument
but it requires too mach to change inside.
I recall, I tortured myself for several
hours to make all ifs to work properly
As for bug: when the package was .zip,
it have specification to look for directory
called gqtest. It retained in 0.18.
I cant detect it before trying inslalling
the package on PC where there are
no previous versions. Obviously
rebuilding anew excluded any
mentions to the directory name.
To make: abstract:
the code was non changed,
only typos, sample script, and spec
compared to the first uploaded version
Oleh
5 квітня 2016, 01:57:51, від "Sven Schreiber" <svetosch(a)gmx.net>:
Am 04.04.2016 um 22:45 schrieb oleg_komashko(a)ukr.net:
> Dear Sven
> The GUI bug is corrected:
> I rebuilt the package from
> the script with no changes w.r.t. 0.18
> If you do not object,
> I am not very inclined to
> change to *series, since
> the behavior of &const
> is to special,
It's your choice as the package author -- and I have to admit/agree that
having to use the pointer form just to make it possible to set a default
value of no argument doesn't feel very logical. (Cf. the situation with
string arguments where it's possible without pointers.)
The other thing, what you call GUI bug, I haven't really understood, but
it doesn't matter.
I'm going to look at it tomorrow (I hope).
thanks,
sven
2:57 a.m.
New subject: gqtest 0.17: 0.19
Am 05.04.2016 um 07:36 schrieb oleg_komashko(a)ukr.net:
Dear Sven,
Since gretl have no named parameters,
I do not like things like ffff(x,null,null,null,1,1,1,1,null,y)
Yes, I agree (named args are nice!), but...
That's why, I tried to as littele arguments as
possible, to this end I used variant with "const"
with a special meaning. I know how to
make the same with *series argument
but it requires too mach to change inside.
this is a totally different argument. But as I said, it's your choice.
As for bug: when the package was .zip,
it have specification to look for directory
called gqtest. It retained in 0.18.
Ok, I see.
One other thing: Now I saw in your code that you use the variable 'ay'
for two different things, either varname(sby) or argname(y). Are you
sure it's always printing out the information that you actually want to
print in the end?
cheers,
sven
3147
days inactive
3152
days old
10 comments
2 participants
participants (2)
-
oleg_komashko＠ukr.net -
Sven Schreiber