[Scilab-Dev] Scilab Improvement Proposal

[Serge Steer]

Claude Gomez a écrit :
>
> It would be good that the output of dae "y" has the same semantics as
> ode's ie a matrix n X T where T is the size of t and each column j of
> y is the value of y at time t(j).
>
> Claude
>

Currently with dassl "y" is a  (1+2*n) X T matrix each column icontains
[tj;xj;xdj].

For basic use tj values are redundant because they are a copy of the t
input argument, but if info(2) ==1 the  time value associated to xj  is
no more known ( intermediate computed values ). In ode "y" is a  n X T
matrix each column  contains [xj] only if odeoptions(1)<>1 and [tj;xj]
if odeoptions(1)==2. The dassl interface choice seems better for me, but
it is not possible today to change the ode semantics. So may be we have
to make dae behave as ode depending on info(2)

Moreover the xdj values cannot be recomputed easily given tj and xj. So
i do not think it is a good idea not to return the xdj values.

Serge