10:24 a.m.
Hi,
simple example:
<hansl>
set verbose off
n = 3
# Works
m = unvech(seq(1,n)')
eval m[end,end]
# Works
m = seq(1,n)'
eval m[end,end]
# Does not work
m = seq(1,n)
eval m[end,end]
</hansl>
Is it expected?
Marcin
--
Marcin Błażejowski
12:12 p.m.
Am 29.03.2022 12:24 schrieb Marcin Błażejowski:
Hi,
simple example:
<hansl>
set verbose off
n = 3
# Works
m = unvech(seq(1,n)')
eval m[end,end]
# Works
m = seq(1,n)'
eval m[end,end]
# Does not work
m = seq(1,n)
eval m[end,end]
</hansl>
The weird thing is that it works for a column vector:
<hansl>
n = 3
m = seq(1,n)'
eval m[end,end]
</hansl>
Artur
5:52 p.m.
On Tue, 29 Mar 2022, atecon wrote:
Am 29.03.2022 12:24 schrieb Marcin Błażejowski:
> Hi,
>
> simple example:
>
> <hansl>
> set verbose off
> n = 3
>
> # Works
> m = unvech(seq(1,n)')
> eval m[end,end]
>
> # Works
> m = seq(1,n)'
> eval m[end,end]
>
> # Does not work
> m = seq(1,n)
> eval m[end,end]
> </hansl>
The weird thing is that it works for a column vector:
<hansl>
n = 3
m = seq(1,n)'
eval m[end,end]
</hansl>
Weird or not, it's now fixed in git.
Allin
6:14 p.m.
Am 29.03.22 um 19:52 schrieb Allin Cottrell:
On Tue, 29 Mar 2022, atecon wrote:
> Am 29.03.2022 12:24 schrieb Marcin Błażejowski:
Weird or not, it's now fixed in git.
Hi Allin,
I updated our test suit on this topic and got the following interesting
cases:
<hansl>
matrix m = {1, 4; 2, 5; 3, 6}
eval m[end:end] # "Ambiguous matrix index"
</hansl>
I guess that makes sense and is expected.
</hansl>
set verbose off
n = 3
# Works after fix
m = seq(1,n)
eval m[end,end]
# Works after fix
m = seq(1,n)'
eval m[end,end]
# Fails
m = seq(1,n)
eval m[(end - 1)]
# Works -- BUT WHY?
m = seq(1,n)
eval m[(end + 1)]
# Works
m = seq(1,n)'
eval m[(end-1)]
# Works -- BUT WHY?
m = seq(1,n)'
eval m[(end+1)]
</hansl>
Artur
5:52 p.m.
On Wed, 30 Mar 2022, Artur T. wrote:
Am 29.03.22 um 19:52 schrieb Allin Cottrell:
> On Tue, 29 Mar 2022, atecon wrote:
>
>> Am 29.03.2022 12:24 schrieb Marcin Błażejowski:
>
> Weird or not, it's now fixed in git.
Hi Allin,
I updated our test suit on this topic and got the following interesting
cases:
<hansl>
matrix m = {1, 4; 2, 5; 3, 6}
eval m[end:end] # "Ambiguous matrix index"
</hansl>
I guess that makes sense and is expected.
Yes, that's a true error, giving a one-dimensional indexation range
for a two-dimensional object.
As for your other cases, good catch! These are now fixed in git: the
ones that should work do work, and the strange ones that shouldn't
work now flag errors.
But I plan to take a deeper look into the relevant code to ensure
that there aren't any other bad cases lurking somewhere. The tricky
thing here is determining unambigously whether an occurrence of
'end' pertains to the row dimension or the column dimension of a
matrix -- even when it's embedded in arithmetic such as "end-1"
and/or wrapped in redundant parentheses as in "(end-1)".
Allin
10:08 p.m.
On Thu, 31 Mar 2022, Allin Cottrell wrote:
On Wed, 30 Mar 2022, Artur T. wrote:
> Am 29.03.22 um 19:52 schrieb Allin Cottrell:
>> On Tue, 29 Mar 2022, atecon wrote:
>>
>>> Am 29.03.2022 12:24 schrieb Marcin Błażejowski:
>>
>> Weird or not, it's now fixed in git.
>
> Hi Allin,
>
> I updated our test suit on this topic and got the following interesting
> cases:
>
> <hansl>
> matrix m = {1, 4; 2, 5; 3, 6}
> eval m[end:end] # "Ambiguous matrix index"
> </hansl>
>
> I guess that makes sense and is expected.
Yes, that's a true error, giving a one-dimensional indexation range for a
two-dimensional object.
As for your other cases, good catch! These are now fixed in git: the ones
that should work do work, and the strange ones that shouldn't work now flag
errors.
But I plan to take a deeper look into the relevant code to ensure that there
aren't any other bad cases lurking somewhere.
I've been stumbling around on this point but now I think I've really
got it, with a new approach. The fix is in git.
Allin
1:09 p.m.
Am 01.04.22 um 00:08 schrieb Allin Cottrell:
On Thu, 31 Mar 2022, Allin Cottrell wrote:
> On Wed, 30 Mar 2022, Artur T. wrote:
>
>> Am 29.03.22 um 19:52 schrieb Allin Cottrell:
>>> On Tue, 29 Mar 2022, atecon wrote:
>>>
>>>> Am 29.03.2022 12:24 schrieb Marcin Błażejowski:
>>>
>>> Weird or not, it's now fixed in git.
>>
>> Hi Allin,
>>
>> I updated our test suit on this topic and got the following
>> interesting cases:
>>
>> <hansl>
>> matrix m = {1, 4; 2, 5; 3, 6}
>> eval m[end:end] # "Ambiguous matrix index"
>> </hansl>
>>
>> I guess that makes sense and is expected.
>
> Yes, that's a true error, giving a one-dimensional indexation range
> for a two-dimensional object.
>
> As for your other cases, good catch! These are now fixed in git: the
> ones that should work do work, and the strange ones that shouldn't
> work now flag errors.
>
> But I plan to take a deeper look into the relevant code to ensure that
> there aren't any other bad cases lurking somewhere.
I've been stumbling around on this point but now I think I've really got
it, with a new approach. The fix is in git.
Great news, Allin. Thank you!
Artur
983
days inactive
986
days old
7 comments
5 participants
participants (5)
-
Allin Cottrell -
Artur T. -
atecon
-
Marcin Błażejowski -
Sven Schreiber