Re: Problem loading/saving of lists and structures.

[John W. Eaton]

On 14-Mar-2000, Steven G. Johnson <address@hidden> wrote:

| Are you referring to people creating new data types from C++ plugin files
| (.oct files)?  Surely you can't define new datatypes from .m files?

Right, you have to use C++.

| Do people actually do this?

It's not just whether other people do it (I think they might, if the
interfaces were stable and documented) but even if I decide to add a
new data type to Octave, it is even more painful having to deal with
more formats.

| (With lists and structures, I wouldn't think that people would need
| entirely new data structures very often.  The only basic datatype missing
| from Octave is  the multi-dimensional array.)

And sparse matrices, matrices (multidimensional or not) of various
integer types, Matlab-like cell arrays, Matlab-like structure arrays,
and probably other special-purpose data types that you might not care
about, but that someone else would.

If someone adds a new data type, creates some variables using it, and
then tries to save the workspace, the current implementation of save
skips everything it doesn't know about.  If save and load try to
accomodate a number of different formats (we now have at least five)
then adding support for new data types is even more work, and it is
not obvious that load-save.cc (which is already a mess) needs to be
modified in a number of places.

I'm not opposed to supporting different file formats, but I'm not sure
that it makes sense to overload the load and save functions to handle
every new file format that people might want to have.

jwe