here's some suggestions for how to improve (IMHO) system estimation in gretl (or hansl for that matter): 1) The basic system spec syntax is a bit awkward: 'system ... end system' should be like a parenthesis, as in 'if ... endif', 'loop ... endloop' etc. Hence the keyword 'system' should be at the beginning of the line. So instead of: mysysname <- system ... end system it would be better to have: system -> mysysname ... end system or something like that. (Of course such a syntax reform would be something for gretl 2.0.)

2) It would be nice if the system spec could be built programatically in a script, e.g. something like: <hansl> mysysname <- system loop foreach i myendoglist equation $i myregressorlist endloop end system </hansl>

3) And finally a general (and old) topic: I thought variable identifiers could now be longer than 15 chars in a script, but I get an error with list names that become longer than that. As I keep repeating every once a while (now's a good opportunity) I still find that restrictive, makes scripts harder to read because we have to use less descriptive, self-explanatory names.

Well I think that for the length of var names, you could allow 31 characters, but when displaying them in reports or printouts requiring column formatting, you would truncate them adding ... to the fixed length. So we would have: IntLoopCountIndexWithAVeryLongName printed IntLoopCountIn... It's the user choice to make variables names distinct at the beginning.

On Fri, 9 Sep 2011, Riccardo (Jack) Lucchetti wrote: > On Fri, 9 Sep 2011, Hélio Guilherme wrote: > >> Well I think that for the length of var names, you could allow 31 >> characters, but when displaying them in reports >> or printouts requiring column formatting, you would truncate them >> adding ... to the fixed length. >> So we would have: >> IntLoopCountIndexWithAVeryLongName >> >> printed >> >> IntLoopCountIn... >> >> It's the user choice to make variables names distinct at the beginning. > > This was discussed at Toruń, and the general consensus seems to be > that we _may_ adopt some Stata-like policy for truncating overly long > variable names at output time. > > I'm not against this, but I'd rather put it off after release. Me too ;-) Let's think some more about this after 1.9.6.

Am 09.09.2011 03:32, schrieb Allin Cottrell: > On Wed, 7 Sep 2011, Sven Schreiber wrote: > >> here's some suggestions for how to improve (IMHO) system estimation in >> gretl (or hansl for that matter): >> >> 1) The basic system spec syntax is a bit awkward: >> 'system ... end system' >> should be like a parenthesis, as in 'if ... endif', 'loop ... endloop' >> etc. Hence the keyword 'system' should be at the beginning of the line. >> So instead of: >> >> mysysname <- system >> ... >> end system >> >> it would be better to have: >> >> system -> mysysname >> ... >> end system >> >> or something like that. (Of course such a syntax reform would be >> something for gretl 2.0.) > > Sorry, Sven, but I don't really get that. We don't have much of the > "pointy arrow" business in gretl (and maybe we should have none of > it) but its role in gretl syntax is to handle a command (or command > block) that is acting something like a function, in that it offers > some sort of assignment. In all the computer science contexts that > I've come across (and notably in R, with which at least some gretl > users may be familiar), the idiom is > > assignee <- <thing offering assignment> > > I'm not sure if you're proposing that all such arrows be reversed, > or just the one for "system", but we need consistency, and I don't > see how reversing all the arrows (mle, gmm, plain single-equation > models, graphs) would improve/clarify matters. > > Maybe I'm missing your point. If so, please elaborate. > My point was only that the keyword 'system' should be at the beginning of the line, because then it would --together with 'end system'-- form a code block also in a visual sense which makes it easier to see the structure of the code. I did not really want to make a definite suggestion on how the rest of the syntax should be changed. However, your response indeed was illuminating: As a matter of fact, I remember that to me the pointy arrows have always felt somewhat "alien" in the context of the rest of what is now known as hansl. So maybe it would indeed be a good idea to ditch them. I mean it's fine for R, where assignments are done with such arrows, but in hansl assignments are generally done via the = sign. Just to give another example of a syntax possibility, instead of: 'mysysname <- system' one could have: 'system --name="mysysname"' which would look and feel much more gretl/hansl-like IMHO. (And analogously for the other pointy arrow cases.) But well, of course that would be a rather drastic change, and the net payoff might be questionable.