4:55 p.m.
Hi,
I'm relaying a usability issue report which I think could lead to
improvements.
(This is not meant for the release in progress, but in general.)
First, if you type in an expression on the console, gretl will not
automatically evaluate it. Instead we have the 'eval' command. However,
most interpreted programming languages out there will directly spit out
the result on what you type in at the prompt. Why not gretl?
Or to put it more formally: I suggest that whenever gretl's parser
cannot make sense of the console input and would respond with an error
message, it should try to automatically prepend an 'eval' and see if
that works.
Secondly (a minor bug), sometimes you get the error message "incomplete
expression", for example if you type in "y[2]" instead of "eval
y[2]".
But this message is misleading, because as an _expression_ in a computer
science sense y[2] is perfectly valid I think. Instead the error could
read, "incomplete command" or perhaps "incomplete statement".
thanks,
sven
7:13 p.m.
On Mar 12, 2018, at 4:55 PM, Sven Schreiber <svetosch(a)gmx.net>
wrote:
Hi,
I'm relaying a usability issue report which I think could lead to improvements.
(This is not meant for the release in progress, but in general.)
First, if you type in an expression on the console, gretl will not automatically evaluate
it. Instead we have the 'eval' command. However, most interpreted programming
languages out there will directly spit out the result on what you type in at the prompt.
Why not gretl?
Or to put it more formally: I suggest that whenever gretl's parser cannot make sense
of the console input and would respond with an error message, it should try to
automatically prepend an 'eval' and see if that works.
If you mean being able to do something like (assuming m has been defined)
<console>
rows(m)
</console>
instead of
<console>
r = rows(m)
r
</console>
then I'd very much like to see this implemented (also for scripts if possible). Maybe
this just reflects my Matlab background, but it's always struck me as inefficient on
gretl's part.
FWIW,
PS
Sent from my iPad
3:18 a.m.
Am 13.03.2018 um 00:13 schrieb Summers, Peter:
If you mean being able to do something like (assuming m has been
defined) <console> rows(m) </console> instead of <console> r =
rows(m) r </console>
Exactly.
then I'd very much like to see this implemented
(also for scripts if possible).
Hm, don't know about scripts. This would be unusual compared to other
languages, and also would hide programming errors IMHO. I think in
scripts it's OK to use print (for plain objects) and eval and/or printf
(for general expressions).
Maybe this just reflects my Matlab
background, but it's always struck me as inefficient on
gretl's
part.
Not just Matlab, but also R, Python, Julia, you name it.
I haven't used the gretl console much in the past, and I think that
limitation was one of the reasons, yes.
thanks,
sven
2:02 p.m.
On Tue, 13 Mar 2018, Sven Schreiber wrote:
Am 13.03.2018 um 00:13 schrieb Summers, Peter:
> If you mean being able to do something like (assuming m has been
> defined)
>
> <console>
> rows(m)
> </console>
> instead of
> <console>
> r = rows(m)
> r
> </console>
Exactly.
Though of course you don't have to take two lines over it as things
stand: the real comparison is "eval rows(m)" versus plain "rows(m)".
> Maybe this just reflects my Matlab background, but it's
always
> struck me as inefficient on gretl's part.
Not just Matlab, but also R, Python, Julia, you name it.
What you're suggesting (making "eval" implicit for any input line that
doesn't make sense as a command or function call) could probably be
done, but there's a relevant difference between the languages you
mention and hansl. Unlike hansl, none of those languages have
commands, they all work purely via functions. So if a given input line
is not a function call it can immediately be taken as an "eval"
request.
Other languages that have commands -- bash, stata, for example, and I
suspect other econometrics software such as RATS, Eviews, Limdep --
don't do the "automatic-eval" thing. By default they're expecting a
command-word, if not a function call.
At present we print one line of text when the console is opened,
apprising the user of 'help'. Maybe we could add a second line,
something like "You can use 'eval' to evaluate any expression".
Allin
3:16 p.m.
Am 13.03.2018 um 19:02 schrieb Allin Cottrell:
On Tue, 13 Mar 2018, Sven Schreiber wrote:
> Not just Matlab, but also R, Python, Julia, you name it.
What you're suggesting (making "eval" implicit for any input line that
doesn't make sense as a command or function call) could probably be
done, but there's a relevant difference between the languages you
mention and hansl. Unlike hansl, none of those languages have commands,
they all work purely via functions. So if a given input line is not a
function call it can immediately be taken as an "eval" request.
Other languages that have commands -- bash, stata, for example, and I
suspect other econometrics software such as RATS, Eviews, Limdep --
don't do the "automatic-eval" thing. By default they're expecting a
command-word, if not a function call.
Ah, very insightful, yes. Although it has to be said that the ipython
shell does both. (AFAIK its "magic" commands start with a % character,
presumably to make them special and to separate them from the allowed
identifiers in python. But the finite set of gretl commands could
probably be treated in a similar way.)
Also, the gretl console kind of already has a bit of implicit
assumptions built in: For a known object x just typing 'x' is equivalent
to typing 'print x', no?
At present we print one line of text when the console is opened,
apprising the user of 'help'. Maybe we could add a second line,
something like "You can use 'eval' to evaluate any expression".
That would already be helpful I think. Perhaps even include a short
example like 'eval I(2) + 2'.
thanks,
sven
2328
days inactive
2329
days old
4 comments
3 participants
participants (3)
-
Allin Cottrell -
Summers, Peter -
Sven Schreiber