Tidy up how FitPage parameter table reflects array distributions

[smk78]

When polydispersity is enabled in a model, selected parameters can take on distributions of values. Array distributions (basically a simple file of weight values as a function of the chosen parameter, such as a radius) allow the User greater flexibility over the form of the distribution to be applied than simply using the built-in dispersity functions (Gaussian, Lognormal, etc), without the complexity of custom-defined distribution functions which require some coding.

The caveat is that, unlike the built-in or custom distributions, array distributions are utilised 'as supplied'; sasmodels does not shift or scale the distribution, meaning it cannot be fit/optimised. This is stated in the help docs. However, this is not reflected in the FitPage.

The unobservant User could easily miss that the parameter using an array distribution is not changing and, worse, believe that that parameter displayed actually has some meaning. It does not! This needs rectifying.

Discussions with @butlerpd and @pkienzle have suggested two options:

• the quick fix would be to simply clear any values and 'grey out' the parameter that is using an array distribution.

• a better fix would be to return some meaningful value characterising the chosen array distribution to the parameter table (though it would not be a fittable parameter for the reason above); the mean of the distribution would be one such parameter.

On this latter point @pkienzle comments:

The help docs for array distributions would also need updating to reflect any change made in the process of dealing with this issue.

[smk78]

@pkienzle comments:
Maybe add something like the follow to the docs (suitably translated to English):

Consider a sample which is a mixture of monodisperse spheres of radius 40 with an equal mass of spheres of radius 80, then in terms of number density you have 1 part 80 for 8 parts 40, giving an array distribution of [40, 8], [80, 1]. This is what sasmodels uses when combining the two I(q).

For continuous distributions, you can use a simple rectangle rule integration with equally spaced x and weight w(x) chosen at the center of each interval. If you use unequally spaced intervals then choose the weight value as the entire density within each interval, and set the the center value to the weighted mean of the interval.

This applies equally to user defined distributions and to array distributions.

[smk78]

@pkienzle also comments:

[smk78]

Also see #1213

[smk78]

Doc updates are complete, but the wider issue mentioned remains. Removing the SasView Documentation label but leaving the issue open.