Re: Octave 2.1.55 available for ftp

[Alois Schloegl]

John W. Eaton wrote:

> Octave 2.1.55 is now available for ftp from ftp.octave.org in the
> directory /pub/octave/bleeding-edge:
>
>  -rw-r--r--  1 103  5323000 Feb 20 21:21 octave-2.1.55.tar.gz
>  -rw-r--r--  1 103  4197965 Feb 20 21:21 octave-2.1.55.tar.bz2
>  -rw-r--r--  1 103    23802 Feb 20 21:43 octave-2.1.54-2.1.55.patch.gz
>  -rw-r--r--  1 103    22930 Feb 20 21:43 octave-2.1.54-2.1.55.patch.bz2
>
> This version should fix most of the bugs reported since 2.1.54.
>
> There are still some significant features related to N-d arrays that
> need to be implemented, but this version does provide a cat function
> and also allows you to use the [] to concatenate N-d numeric arrays.
> This still needs to be extended to handle cell and structure arrays
> and user-defined types.  We hope to fix this for 2.1.56.  Indexing and
> indexed assignment for N-d arrays should also be much better in
> 2.1.55.  If you find that another of your favorite N-d array
> operations doesn't work, please mention it so we can have some idea of
> what features are missed the most.  Patches are also welcome.
>
> As always, if your favorite bug is still not fixed, please report it.
>
> Also, 2.1.50 is still the recommended version, but 2.1.55 should be an
> improvement over 2.1.54.  If there are no serious problems reported
> during the next few weeks for for anything that should have worked
> with 2.1.50, then 2.1.55 will probably be promoted to the recommended
> version.
>
> Thanks,
>
> jwe
>  


In 2.1.50 it was possible to do something like the following

fid = fopen(filename);
hdr = fread(fid, ...);       % reading header information, header might
contain complicated constructs.
hdr.Blocksize=fread(fid, 1, 'int32'); % one field in the header contains
e.g. the Blocksize, or the Number_of_blocks
HeaderLength = ftell(fid);
fseek(fid,0,'eof');
FileSize = ftell(fid);
hdr.Blocksize = (FileSize-HeaderLength)/hdr.number_of_blocks;
fseek(fid,HeaderLen,'bof');
for k = 1:hdr.Blocksize,
        d = fread(fid,hdr.number_of_blocks,'int16');
        % do something with d           
end;
fclose;

This is not possible in 2.1.55, because ftell return value is not numeric.

A possible hack (i.e. workaround) for this problem is the patch in
http://www.octave.org/octave-lists/archive/bug-octave.2004/msg00002.html.
At least this seems to do the job.

I was thinking about an alternative solution, too. A possible solution 
could also be separating the internal file-io from the file-io for the 
users. The user file-io could be mapped to the file-io of stdio; the 
internal file-io of Octave could still use the c++ streams. Presumably, 
you know much better than me, if this makes sense.


Alois