Re: xlsread in Octave 3.6.4

[Philip Nienhuis]

<moved from help-octave to octave-maintainers ML>

Markus Bergholz wrote:
> On Mon, Sep 2, 2013 at 11:38 AM, Markus Bergholz <address@hidden
> <mailto:address@hidden>> wrote:
>     On Mon, Sep 2, 2013 at 12:10 AM, Markus Bergholz <address@hidden
>     <mailto:address@hidden>> wrote:
>         On Sun, Sep 1, 2013 at 11:42 PM, PhilipNienhuis
>         <address@hidden <mailto:address@hidden>> wrote:
>             Markus Bergholz wrote

<snip>

>              >>>>     Markus Bergholz wrote
>              >>>> > I haven't follow this thread and it's issue, but
>             i've wrote a
>              >>>>     xlsxread
>              >>>> > function whitch don't need java.
>              >>>> > but it's very very rudimentary, works just with
>             linux and is a
>              >>>>     quick&dirty
>              >>>> > write-down.
>              >>>> > furthermore, you have to remove the string-analyse
>             part, if your
>              >>>>     sheet
>              >>>> > don't contain strings.
>              >>>> > but maybe it helps someone else or someone want to
>             improve it or
>              >>>>     someone
>              >>>> > rewrite it in c/c++ as oct file, to get it even
>             faster than
>              >>>>     matlab (for me
>              >>>> > it's still faster than the java stuff atm).

<snip, see thread on help-octave ML>

>     i've made a few quick and dirty changes, change to gpl licence and
>     commit the broken range part too.
>
>     https://github.com/markuman/xlsxread
>
>     it's now plattform indepentend and - once again - faster than before
>     (~58 seconds). now it's nearly twice as fast as matlab (~110 seconds).
>     enough time to waste it for ranges, strings etc in future.
>
>
>
> here comes version 0.6 - https://github.com/markuman/xlsxread
>
> * strings and calculations are now replaced with NaN (without any speed
> losses!)
>    * testet with a excel 2007 and excel for mac 2011 file (example files
> are added)
> * it's using now nested functions. this should be easier to ingetrate it
> in octave-io
>
> ranges and empty columns still don't work!

Good work Markus.

Anyway, sorry to come up with a few more potential gotchas:

- Interesting would be if your code properly handles merged and hidden 
cells. I don't know what they look like in raw OOXML.

- Does OOXML have repeated-rows and repeated-columns "folding"?
E.g., ODS1.2 has the table:TableNumberRowsRepeated and 
table:TableNumberColumnsRepeated tags.

It would be really good to have a Java-free (and ActiveX-free) 
spreadsheet reading capability in Octave, even if only a basic one. 
Sergei suggested a Perl-based solution; but Perl would still be a 
dependency, not all systems have Perl installed (e.g., Windows).
You've made a first try for OOXML; I have a basis for decoding ODS lying 
around, it doesn't work at all yet but might not need undue amounts of 
attention.
You made the vital piece: unzipping the spreadsheet file to disk.

For inclusion in the OF io package (in a later stage, first try to get 
your version robust and fail-safe) I'd suggest to see how the various 
"interfaces" are built and called in the OF io package.

For follow-up I'd suggest to move this discussion to the maintainers 
list. I've swapped help-octave into octave-maintainers.


Oh BTW another idea (that I explored in 2009 but couldn't get to work at 
the time):
There is a binary (compiled) xmlread function, currently it is in the io 
package. Maybe with a proper "template" it could just read the 
worksheets into a struct in RAM, faster than regexp can decode it. The 
missing piece is the "template". (sorry for my lack of XML proficiency & 
lingo) Unfortunately that xmlread is very tersely, if not badly documented.
However there are xml toolboxes around that could be gotten to work in 
Octave.

Philip