Mon 15 Feb 2016 06:47:56 PM UTC, comment #11:
I committed the patch here.
I changed
to
because it seems that any empty matrix should be suppressed and that
It also created a failing test in copyobj.m. That seems to have been potentially unclear coding that was vulnerable to the change in definition. I fixed that too.
|
Mon 08 Feb 2016 06:53:04 PM UTC, comment #10:
@Lachlan
I've tested it and reported / edited the bug overview you and Avinoam made in the wiki some day ago. AFAIR it worked well.
I didn't push yet due to lack of time (testing is one thing, can be done in between, but with pushing I always try to be extra cautious)
[OT]
Did I understand well that you've been given push credentials? If so, congratulations! very well.
(and if not so, it would be about time you would get it IMO)
|
Tue 26 Jan 2016 04:54:49 AM UTC, comment #9:
The attached patch fixes the problem reported.
It makes constructs such as [ [], {1}; {2}, {3} ] illegal, since the rows have different numbers of elements. I assume that is the desired behaviour.
(file #36166)
|
Fri 20 Nov 2015 01:47:49 PM UTC, comment #8:
This also occurs in the released Octave 4.0.0
|
Thu 19 Mar 2015 10:01:13 PM UTC, comment #7:
Matlab r2015a behaves the same as r2012a.
|
Sat 08 Nov 2014 10:36:00 PM UTC, comment #6:
Sure, here is the result:
I've marked it as a regression :-)
|
Fri 07 Nov 2014 03:57:53 PM UTC, comment #5:
Well, at least the behavior seems to be generalized to concatenation with any cell.
@Philip: Could you make one more test? I don't think the first entry is special, but I want to verify that Matlab removes all null matrices wherever they appear in the list.
The following example in the development version of Octave produces:
Under Octave 3.4.3 it produces
If Matlab is the same as 3.4.3 then I think we should mark this bug as a regression.
|
Fri 07 Nov 2014 12:13:33 PM UTC, comment #4:
Forgot (comment #3): r2012a 32-bit
|
Fri 07 Nov 2014 12:13:06 PM UTC, comment #3:
Yes it does strip it out.
(I can't copy results verbatim, as my ML prerelease timed out and I ran the commands in passing while being logged in remotely at work.)
|
Mon 03 Nov 2014 04:28:22 AM UTC, comment #2:
I don't think there is any difference between cellstr and regular cells arrays, but could someone check the following concatenation in Matlab?
Does this also strip out the null matrix b?
|
Fri 31 Oct 2014 10:11:41 PM UTC, comment #1:
Matlab r2014b prerelease does (empty lines removed):
...so yes, Octave (just checked in 4.1.0+ too) behaves at least ML-incompatible.
Release => dev, OS => any, title (Summary) adapted a little
|
Fri 31 Oct 2014 10:02:14 PM UTC, original submission:
Version 3.4.3 and older matlab versions would return
|