2:31 p.m.
Dear all,
is there a way to delete a matrix? The "delete" command does not work
for matrices (by the way also "rename" is not considered -- even though
I don't know whether this is a relevant case).
I am asking as I face a case where I would like to drop a huge matrix
"x" demanding a lot of RAM after exploiting its information content. The
only way I see to reduce its size is by running "matrix x = zeros(1,1)"...
Best,
Artur
2:55 p.m.
On Mon, 4 Dec 2017, Sven Schreiber wrote:
Am 04.12.2017 um 20:31 schrieb Artur Tarassow:
> Dear all,
>
> is there a way to delete a matrix? The "delete" command does not work for
> matrices
This works for me:
matrix m = ones(2,1)
delete m
print m
Me too:
matrix M = mnormal(500,500)
eval tr(M)
delete M
What's the context in which "delete" is not working on a matrix?
Allin
3:14 p.m.
> This works for me:
>
> matrix m = ones(2,1)
> delete m
> print m
Me too:
matrix M = mnormal(500,500)
eval tr(M)
delete M
What's the context in which "delete" is not working on a matrix?
Ok, here I am again. The following does not work using 2-days old git on
ubuntu:
<hansl>
set verbose off
clear
# generate artificial data set
nulldata 10
matrix M = mnormal(10,2)
loop i=1..2 -q
series v$i=M[,i]
endloop
store "testdat1.csv" --csv --no-header
# Load in a case matrix
loop lcase=1..1 -q
sprintf f "testdat$lcase.csv"
open "@f" --quiet --preserve
list L = null
list L = v1..v2
list L = dataset - const
matrix case$lcase = {L}
eval case$lcase
delete case$lcase # error: "delete case1: not allowed"
endloop
delete case1 # works
</hansl>
Artur
5:59 p.m.
On Mon, 4 Dec 2017, Artur Tarassow wrote:
> What's the context in which "delete" is not working
on a matrix?
Ok, here I am again. The following does not work using 2-days old git on
ubuntu:
<hansl>
set verbose off
clear
# generate artificial data set
nulldata 10
matrix M = mnormal(10,2)
loop i=1..2 -q
series v$i=M[,i]
endloop
store "testdat1.csv" --csv --no-header
# Load in a case matrix
loop lcase=1..1 -q
sprintf f "testdat$lcase.csv"
open "@f" --quiet --preserve
list L = null
list L = v1..v2
list L = dataset - const
matrix case$lcase = {L}
eval case$lcase
delete case$lcase # error: "delete case1: not allowed"
endloop
delete case1 # works
Ah, then that's the generic ban against deleting objects inside a
loop: this is the price to be paid for certain optimizations of the
gretl loop code. If we're to "quasi-compile" loops we can't have
objects popping out of existence as the iterations proceed.
The nearest acceptable equivalent would be
case1 = {}
to reduce the matrix in question to an empty one (0 x 0), hence
freeing up almost all of the memory allocated to it.
Allin
6:01 p.m.
Ah, then that's the generic ban against deleting objects inside a
loop:
this is the price to be paid for certain optimizations of the gretl loop
code. If we're to "quasi-compile" loops we can't have objects popping
out of existence as the iterations proceed.
Ah, ok! Thanks for clarification, Allin. I am totally fine with this
decision.
The nearest acceptable equivalent would be
case1 = {}
to reduce the matrix in question to an empty one (0 x 0), hence freeing
up almost all of the memory allocated to it.
Yes, indeed.
Best,
Artur
2510
days inactive
2510
days old
7 comments
3 participants
participants (3)
-
Allin Cottrell -
Artur Tarassow -
Sven Schreiber