external data-handling

[Daniel Heiserer]

Hi,

an additional question to the handling of
external data with octave.

I have many huge matrices I cannot keep in core.
But I want to have them available 
THE SAME WAY AS THEY ARE IN CORE

What I think how it works:
----------------------
Everytime octave accesses data, like
passing it to a function or whatever
at the very end it somehow has
to be passed to a build in function.
At that moment, the function
has to know, where the data starts
and what kind of data it is.
e.g. pointer in core, some info about
the datablock .......
RIGHT?

What I think it could work with less coding effort:
--------------------------------------------------
Wouldn't it be quite easy to lets
say define some scratch directory
/tmp/octave-2132312
were octave keeps big data.
The only thing it does 
then, except passing a memory pointer 
to a function is, that it 
passes a file-identifier 
to the function. After
the function is finished
the file is closed again
and the result(s) of the 
function are stored in the
scratch directory.
Or octave reads the data,
passes it to the function 
in the same way it does right
now and saves the result on disk.
Which means, every time 
octave is not inside a build-in function
the memory is almost empty.
Of course to make this attempt shine
one should be able to have some parameters
to define how much is put on the
disk and how much is in memory and ....


The whole file-system could work as a sort of a database....
Wouldn't it be just another "method"
of passing data to a function?

Thanks 

Daniel

daniel.heiserer.vcf
Description: Card for Daniel Heiserer