5:21 p.m.
Sorry, I didn't reply to the list.
--------------------------- Original Message ----------------------------
Subject: Re: [Gretl-users] vertical matrix concatenation (was matrix
declaration error)
From: "Riccardo (Jack) Lucchetti" <r.lucchetti(a)univpm.it>
Date: Tue, January 30, 2007 23:19
To: "Allin Cottrell" <cottrell(a)wfu.edu>
-------------------------------------------------------------------------
On Tue, January 30, 2007 23:11, Allin Cottrell wrote:
That does make sense, but I'm a bit reluctant to change the
meaning of logical "&" and "|" right now -- though we could do
that later with fair warning.
What we could do is: introduce now "&&" and "||" as synonyms
of "&" and
"|" but issue a warning everytime the 1-char versions are found,
notifying users that they are deprecated (big red letters, fire alarm,
buzzers, flashing lights, whatever). In a year's time or so, introduce
"|".
Riccardo (Jack) Lucchetti
Dipartimento di Economia
Facoltà di Economia "G. Fuà"
Ancona
5:41 p.m.
New subject: vertical matrix concatenation (was matrix declaration error)
On Tue, 30 Jan 2007, Riccardo (Jack) Lucchetti wrote:
On Tue, January 30, 2007 23:11, Allin Cottrell wrote:
> That does make sense, but I'm a bit reluctant to change the
> meaning of logical "&" and "|" right now -- though we could
do
> that later with fair warning.
What we could do is: introduce now "&&" and "||" as synonyms
of "&" and
"|" but issue a warning everytime the 1-char versions are found,
notifying users that they are deprecated (big red letters, fire alarm,
buzzers, flashing lights, whatever). In a year's time or so, introduce
"|".
Yes, I think that's a good idea. Meanwhile, do we want a
temporary symbol for vertical concatenation, or just forget it for
now?
Allin.
5:49 p.m.
New subject: vertical matrix concatenation (was matrix declaration error)
On Tue, January 30, 2007 23:41, Allin Cottrell wrote:
On Tue, 30 Jan 2007, Riccardo (Jack) Lucchetti wrote:
> On Tue, January 30, 2007 23:11, Allin Cottrell wrote:
>
>> That does make sense, but I'm a bit reluctant to change the
>> meaning of logical "&" and "|" right now -- though we
could do
>> that later with fair warning.
>
> What we could do is: introduce now "&&" and "||" as
synonyms of "&" and
> "|" but issue a warning everytime the 1-char versions are found,
> notifying users that they are deprecated (big red letters, fire alarm,
> buzzers, flashing lights, whatever). In a year's time or so, introduce
> "|".
Yes, I think that's a good idea. Meanwhile, do we want a
temporary symbol for vertical concatenation, or just forget it for
now?
IMHO there's no point in introducing an element of syntax that is going
to be dropped in a year. As far as I'm concerned, I can live with
X = ( a' ~ b' )'
for now, but I can see the point of those who disagee.
Riccardo (Jack) Lucchetti
Dipartimento di Economia
Facoltà di Economia "G. Fuà"
Ancona
6:25 p.m.
New subject: vertical matrix concatenation (was matrix declaration error)
On Tue, 30 Jan 2007, Allin Cottrell wrote:
On Tue, 30 Jan 2007, Riccardo (Jack) Lucchetti wrote:
> On Tue, January 30, 2007 23:11, Allin Cottrell wrote:
>
>> That does make sense, but I'm a bit reluctant to change the
>> meaning of logical "&" and "|" right now -- though we
could do
>> that later with fair warning.
>
> What we could do is: introduce now "&&" and "||" as
synonyms of "&" and
> "|" but issue a warning everytime the 1-char versions are found,
> notifying users that they are deprecated (big red letters, fire alarm,
> buzzers, flashing lights, whatever). In a year's time or so, introduce
> "|".
One more take on this (still trying to square the circle here!).
It strikes me that logical OR as applied to matrices, is, while
potentially meaningful, quite arcane. I'd be surprised if there
are any gretl scripts in circulation that involve OR'ing matrices.
So, could we perhaps do this (as Jack suggests, with one
amendment):
* Amend the gretl manual to make "&&" and "||" the preferred
forms
of logical AND and OR.
* Continue to support "&" as AND and "|" as OR for the near
future, but with a "deprecation" message...
* ...with the exception that "A|B" in matrix-generating
expressions means vertical concatenation (and if you really want
to OR matrices you must use the new "||" form).
I'm not pushing this idea all that hard, just wondering. It
carries some potential for confusion. The main problem I see is
identifying a "matrix-generating expression" for the purpose of
interpreting "|". My notion is that this would be the right-hand
side of a "matrix M = ..." command, or the right-hand side of
commands on the pattern
genr M = ...
M = ...
where "M" can be identified as a pre-existing matrix.
Allin.
4:34 a.m.
New subject: vertical matrix concatenation (was matrix declaration error)
Allin Cottrell schrieb:
One more take on this (still trying to square the circle here!).
It strikes me that logical OR as applied to matrices, is, while
potentially meaningful, quite arcane. I'd be surprised if there are any
gretl scripts in circulation that involve OR'ing matrices.
Right, especially when the truth value of a matrix seems undefined?
So, could we perhaps do this (as Jack suggests, with one amendment):
* Amend the gretl manual to make "&&" and "||" the preferred
forms of
logical AND and OR.
Perfectly fine with me. Just to add another perspective: in python it's
just the keywords 'and' and 'or', would that be an option? (python is
quite intuitive, you know ;-)
* Continue to support "&" as AND and "|" as OR for the near
future, but
with a "deprecation" message...
* ...with the exception that "A|B" in matrix-generating expressions
means vertical concatenation (and if you really want to OR matrices you
must use the new "||" form).
Fine again. Another possibility would be to allow a and b to be matrices
in expressions like {a;b}, giving implicit concatenation. Though I guess
it's more difficult to implement...
I'm not pushing this idea all that hard, just wondering. It carries
some potential for confusion. The main problem I see is identifying a
"matrix-generating expression" for the purpose of interpreting "|".
My
notion is that this would be the right-hand side of a "matrix M = ..."
command, or the right-hand side of commands on the pattern
genr M = ...
M = ...
where "M" can be identified as a pre-existing matrix.
Has the syntax changed here? I thought the 'matrix' keyword always has
to precede matrix-related statements?
thanks,
Sven
9:52 a.m.
New subject: vertical matrix concatenation (was matrix declaration error)
Me:
> The main problem I see is identifying a "matrix-generating
> expression" for the purpose of interpreting "|". My notion is
> that this would be the right-hand side of a "matrix M = ..."
> command, or the right-hand side of commands on the pattern
>
> genr M = ...
> M = ...
>
> where "M" can be identified as a pre-existing matrix.
Sven:
Has the syntax changed here? I thought the 'matrix' keyword
always has to precede matrix-related statements?
Yes, it has changed.
Gretl now has "static typing". This means that once a given
variable is defined as having a certain type, you can't change
its type on the fly. For example, this will produce an error:
scalar x = 2
x = log(price)
if "price" is a series. On the other hand it is OK to "cast" a
variable to a different type, as in
scalar x = 2
series y = x
which produces a series each of whose observations has value 2.
One advantage of static typing is that you don't have to keep
repeating type identifiers such as "matrix": if M is initially
defined as a matrix, you can thereafter use plain "M" on the
left-hand side of genr-type commands: gretl will know that it's
supposed to be generating a matrix.
Allin.
5:32 a.m.
New subject: vertical matrix concatenation (was matrix declaration error)
On Wed, January 31, 2007 00:25, Allin Cottrell wrote:
It strikes me that logical OR as applied to matrices, is, while
potentially meaningful, quite arcane. I'd be surprised if there
are any gretl scripts in circulation that involve OR'ing matrices.
...
* ...with the exception that "A|B" in matrix-generating
expressions means vertical concatenation (and if you really want
to OR matrices you must use the new "||" form).
I'm not pushing this idea all that hard, just wondering. It
carries some potential for confusion. The main problem I see is
identifying a "matrix-generating expression" for the purpose of
interpreting "|". My notion is that this would be the right-hand
side of a "matrix M = ..." command, or the right-hand side of
commands on the pattern
genr M = ...
M = ...
Yes, this makes sense to me. The only thing is that we need to warn users
that
c = 2 | 1
will _not_ return a 2x1 vector, as one might expect, but
c = {2} | {1}
will.
Riccardo (Jack) Lucchetti
Dipartimento di Economia
Facoltà di Economia "G. Fuà"
Ancona
Riccardo (Jack) Lucchetti
Dipartimento di Economia
Facoltà di Economia "G. Fuà"
Ancona
3:53 a.m.
New subject: vertical matrix concatenation (was matrix declaration error)
El Martes, 30 de Enero de 2007 23:41, Allin Cottrell escribió:
Yes, I think that's a good idea. Meanwhile, do we want a
temporary symbol for vertical concatenation, or just forget it for
now?
My vote is for leaving the things as they are. I think it is not so
problematic to use X = ( a' ~ b' )'
I would not define another symbol for this function. But probably this should
be clearly indicated in the manual.
--
Ignacio Díaz-Emparanza
Dpto. de Economía Aplicada III (Econometría y Estadística)
UPV-EHU
9:42 a.m.
New subject: vertical matrix concatenation (was matrix declaration error)
On Wed, 31 Jan 2007, Ignacio Díaz-Emparanza wrote:
El Martes, 30 de Enero de 2007 23:41, Allin Cottrell escribió:
> Yes, I think that's a good idea. Meanwhile, do we want a
> temporary symbol for vertical concatenation, or just forget it for
> now?
My vote is for leaving the things as they are. I think it is not so
problematic to use X = ( a' ~ b' )'
It's true, that does the job. But a direct "X = a | b" would be
more computationally efficient. The version above involves n+1
"unnecessary" transpositions when concatenating n matrices. With
small matrices and a simple script that hardly matters, but it
could make an appreciable difference with large matrices in the
context of an iteration.
Allin.
6459
days inactive
6460
days old
8 comments
4 participants
participants (4)
-
Allin Cottrell -
Ignacio Díaz-Emparanza -
Riccardo (Jack) Lucchetti -
Sven Schreiber