Re: Fwd: Test files [WAS: Re: xlsread in Octave 3.6.4]

[Philip Nienhuis]

Markus wrote:
> > =============================================================
> > >> xlsfinfo ("Mergedranges.xlsx", "uno")
> >   1: Sheet1                           (Used range ~ J1:AMJ15)
> > ans = Microsoft Excel Spreadsheet
> > =============================================================
> > Hey.....?? ^^^^^^^^^^^^^^^^^^^ ??
>
>
> It's fixed now. Can i push your attached files too?

You mean: to your repo?

Sure, that's fine with me.

As to your other reply: No, I'm not so interested in formula evaluation 
support; that would get extremely complicated. I merely wanted to draw 
attention to this (little) pitfall - I've encountered it a few times 
myself. Matlab itself has the same problem on systems w/o MS-Excel 
installed (Linux, Mac OSX, ....) but hasn't got it documented.

> Matlab mat file: http://p.osuv.de/index.php/PIO/
> Octave mat file: http://p.osuv.de/index.php/aiS/
>
> the last thing is that matlab ignored unused cells in the given
> dimension around the values.
> matlab size 8x1015
> octave size 15x1024
> next step is to get rid of them too.
>
> And xlsxread can already handle folded cells.
> Take a look at the 3 screenshots. It was just one more line to do this job.
>
> Test1.xlsx http://p.osuv.de/index.php/RyD/
> Octave http://p.osuv.de/index.php/rXK/
> Matlab http://p.osuv.de/index.php/Zc3/
> Excel Screenshot: http://p.osuv.de/index.php/hLtY/

Next Friday I'll have more time, now only a few minutes. Sorry for that.

If xlsxread.m works OK to your satisfaction, we'll have to think on:
- Do you want it to remain Matlab-compatible, or Octave style
- Do you want in the OF-io package
- If yes (I suppose so), well: how then?
  * do we split it up cf. the other interfaces, or
  * keep it in one file as just xlsxread
  * do you want to add more options, like
    > specifying a range to read
    > specifying sheets by name and/or by index number
etc etc.

Splitting up would impose a little speed penalty, but add a lot of 
flexibility.
You could have a look on how things work for xlsread.m in the io package 
and see if you like the philosophy there:

xlsread.m is a wrapper for xlsopen-xls2oct-[xls2oct-][...-]xlsclose 
fuctions, to allow reading successive worksheets w/o repeated opening & 
closing of a complete spreadsheet file (one of Matlab's more braindead 
implementation details).

xlsopen.m is a wrapper for interface-specific functions in ./private/
Same for xls2oct.m and xlsclose.m. The interesting interface-specific 
part is at the bottom of those files.

Interface-specific functions would be needed for:
1. Opening the spreadsheet file (unzipping it to a temp dir);
   (This has already been done (for ODS) using your solution)
2. Detecting the occupied cell range in a worksheet;
3. Reading the data;
4. Closing the spreadsheet file;
   (Extremely easy)
and when that's in place it's a small step to add a function to:
5. Get worksheet names & cell ranges for xlsfinfo.m support.
   (Not so hard I think)

So for this sequence your xlsxread would merely have to be split into 
steps 2 and 3, if at all possible.

Please think about it. I'll come back to it next Friday.

Thanks,

Philip