Re: [SED-ML-discuss] SED-ML reports

[David N. <dav...@gm...>]

> The other issue here is that any implementation must provide the output in
> *some* order, and if you don't tell it what order you want, it will pick
> something, and there will be, generally no reason to pick one order over
> another (this is not true of the other things you've compared order to,
> like font and color).  Why not tell it what you want, so it can give it to
> you?  Why explicitly deny it that information?

While it makes sense for the SBML test suite to describe the way
simulation results are stored in order for testing, isn't that a
feature of the test suite? For example, can't the test suite simply
state that when reports are generated with the data ordered
alphabetically by label you can compare them against the "correct"
original data.

Having said that, in my experience a simple text-based diff of
simulation data is rarely sufficient to determine if the results are
"correct". And once you go beyond a simple text-based diff, as Nicolas
points out, there is very little extra effort to work out what is what
using the label. Any required ordering for specific applications can
then be specified by the application based on the label. After all,
there is nothing wrong with labeling your data references "1", "2",
"3".... :)


Cheers,
Andre.