Re: AW: AW: AW: AW: Re: AW: textscan (Contd.)

[Philip Nienhuis]

(Maintainers list:
Alexander Graf and I had a private e-mail exchange about textscan bugs 
that yesterday finally have been fixed (see bugs #36356 and #36392) but 
not pushed yet.
From this point on I think the discussion may be useful for a somewhat 
broader audience as it displays more ML textscan corner case 
idiosyncrasies - see previous dicussions, e.g. here: 
https://mailman.cae.wisc.edu/pipermail/octave-maintainers/2011-October/025443.html 
)

==================

Thanks, Alexander, for your prompt reply.

Your ML results show that there's another bug in textscan and one in 
strread.m while processing litterals, see below:

Graf, Alexander wrote:
> Philip Nienhuis wrote:
>
> > Could you please tell me what ML returns for the following
> > statements (I'm not at work until Tuesday 15 May, so I can't
> > check myself):
> Hi,
>
> > cell (1, 3)
>
> ans =
>
>      []    []    []

octave:2> cell (1, 3)
ans =
{
  [1,1] = [](0x0)
  [1,2] = [](0x0)
  [1,3] = [](0x0)
}

> > cell (0, 3)
>
> ans =
>
>     Empty cell array: 0-by-3

octave:1> cell (0, 3)
ans = {}(0x3)

My conclusion:
ML and Octave do largely the same but display it differently. Which 
confirms your conclusion that for your scripts, Octave's textscan now 
performs similar to ML's textscan.
That means: 2 bugs squashed. Good.

> > C = textscan (' ', '%f %f')
>
> ??? Error using ==>  textscan
> First input can not be empty.

octave:3> C = textscan (' ', '%f %f')
C =
{
  [1,1] = [](0x1)
  [1,2] = [](0x1)
}

Apparantly a ML bug. Or more politely put: undocumented behaviour.
The leading whitespace should have been simply ignored.

> > D = textscan ([char(10)], '%f %f')
>
> D =
>
>      [0x1 double]    [0x1 double]

octave:4> D = textscan ([char(10)], '%f %f')
error: textscan: subscript indices must be either positive integers or 
logicals
error: called from:
error: 
C:\Programs\Octave\Octave3.6.1_gcc4.6.2\share\octave\3.6.1\m\io\textscan.m 
at line 256, column 12

*** Wow, another bug in textscan.

> > E = textscan (['Empty'], 'Empty%f %f')
>
> E =
>
>      [NaN]    [NaN]

Probably another ML bug.
The command doesn't really differ from the previous one except for a 
leading litteral, so IMO it should return the same output.

octave:4> E = textscan (['Empty'], 'Empty%f %f')
warning: strread: unable to parse text or file with given format string
error: some elements undefined in return list
error: called from:
error: 
C:\Programs\Octave\Octave3.6.1_gcc4.6.2\share\octave\3.6.1\m\io\textscan.m 
at line 266, column 9

Ah... another strread.m bug.

> > E = textscan (['Empty' char(10)], 'Empty%f %f')
>
> E =
>
>      [NaN]    [NaN]

Octave:
same as above.

So, some more home work for me......

BTW what ML version do you use?

Thanks very much for your cooperation.
I have the fixes ready, hopefully I can upload them this evening, if not 
it'll be two weeks from now (as from early tomorrow I'll be away until 
mid May).

Philip