Better interface for setParameters

[SengerM]

Purpose

setParameters(Name=value)
setParameters(Name1=value1, Name2=value2)

are more clean and natural than

setParameters("Name=value")
setParameters(["Name1=value1","Name2=value2"])

It should be backwards compatible, meaning that it still accepts the strings.

[CLAassistant]

All committers have signed the CLA.

[syntron]

@SengerM would this work on top of PR #314? it creates the option to use dictionaries for all set*() functions ...

[SengerM]

@syntron I think PR #314 is an improvement with respect to the current interface, but the interface from this PR is superior and I would suggest to implement it like this in all the set methods. It is the most natural way.

For quick comparison:

model.setParameters(['Name1=Value1', 'Name2=Value2', ...]) # Current interface (cumbersome, error prone, have to deal with conversion to strings, etc).
model.SetParameters(dict(Name1=Value1, Name2=Value2, ...)) # PR #314 (boilerplate `dict()`, other than that it is acceptable).
model.SetParameters(Name1=Value1, Name2=Value2, ...) # This PR (as good as it gets).

I would also only keep the current interface with strings for backwards compatibility, and add a warning that it will be deprecated in the future.

[syntron]

I see the dict based interface as preferable if lots of options are handled which are defined before as dictionary. My idea is to add an additional function which just translates the argument key based variant to the dict based version.

I would also only keep the current interface with strings for backwards compatibility, and add a warning that it will be deprecated in the future.

This is done in PR #314; I also prepared a commit which would remove the depreciated / old way of setting the options.

[syntron]

Please see commit 6cd65ea of PR #314 ...

[SengerM]

I see the dict based interface as preferable if lots of options are handled which are defined before as dictionary.

The "normal Python interface" from this PR can naturally do that (if I understood correctly what you mean):

parameters = {f'Name{i}': i**2 for i in range(999)} # 999 different parameters.
model.setParameters(**parameters)

My advice is to keep it simple and to offer the user an interface he would expect in Python. Less maintenance and more user friendly.

This is also the way it is done in many other packages, e.g. matplotlyb.pyplot.plot, which is defined as matplotlib.pyplot.plot(*args, scalex=True, scaley=True, data=None, **kwargs) instead of matplotlib.pyplot.plot(*args, scalex=True, scaley=True, data=None, kwargs). This allows you to do plot(x, y, color='green', marker='o', linestyle='dashed', linewidth=2, markersize=12) instead of plot(x, y, dict(color='green', marker='o', linestyle='dashed', linewidth=2, markersize=12)), and you can of course do

plot_settings = dict(color='green', marker='o', linestyle='dashed', linewidth=2, markersize=12)
plot(x, y, **plot_settings)

[syntron]

@SengerM I learned something and will check this out - thanks for the explanation!