Re: Fixes to ls-mat5.cc and ov-class.cc

[Robert T. Short]

This is just a report concerning load/save of objects.

I have tested this with a collection of objects, some very simple, some much more complex.

As you can see, there are bugs, but I am not sure the bugs are in my code.  It will take a while but I will track them down and squash them wherever they are.


--
If I create the save files from MATLAB using the -v6 option, octave loads them perfectly.

If I create the save files from octave using the -v6 option, octave loads the simple ones perfectly but the complex object fails.  I think the format of the .mat file is incorrect, but not sure.  I just need to "hd" the file and walk through it.  Tedious, but simple.

If I create the save files from octave using the octave binary format, octave loads them perfectly.

If I create the save files from octave using the octave -text format, octave loads them perfectly.

--
I have not tested the "loadobj" function, and I have not really worked out some issues with error handling.


Bob
--
Robert T. Short
PhaseLocked Systems

Robert T. Short wrote:

John W. Eaton wrote:

Could you also look at the code in src/ls-mat5.cc that loads objects
from MAT files and make similar changes there?  Also, since the
structure of the code that loads MAT files is different, you can
easily convert an object to a struct if there is a problem with the
exemplar or the call to the constructor fails.

 

Attached is a changeset that does this.  Also a couple of minor changes to ov-class.cc; a spelling error and methods(obj) should return a cell array.

But now I have a question about the whole load/save thing.

I have not fixed the return struct thing yet because I am a little confused.  How are load_binary and the code in ls-mat5 related?  Specifically, I created .mat files from MATLAB and read them into octave.  It seems that the code in load_binary is called, not the mat5 code.

Bob
--
Robert T. Short
PhaseLocked Systems