GeezerVison

Coverage decreased (-0.01%) to 94.828% when pulling 16d9e6a on sawyerbfuller:named-signals into 2746ce1 on python-control:main.

I suggest following the conventions used in linearize():

• Use copy keyword to control whether signal names are copied (default can be True, although for linearize it is False for some reason).
• Use name keyword to set the system name. If name is not specified and if copy is False, a generic name <sys[id]> is generated with a unique integer id. If copy is True, the new system name is determined by adding the prefix and suffix strings in config.defaults['iosys.linearized_system_name_prefix'] and config.defaults['iosys.linearized_system_name_suffix'], with the default being to add the suffix '$linearized'.

Doing this for sample(), the default name of the new system could be the old system with $sampled appended (and overridable via something like iosys.sample_system_name_prefix and iosys.sample_system_name_suffix.

Also, we might want to override the sample() method in LinearIOSystem so that sampling a linear input/output system gives a linear input/output system (rather than just a StateSpace system). This won't matter much since StateSpace is converted to LinearIOSystem if it is connected to anything, but would be a bit cleaner (and points to the fact that we should probably merge StateSpace and LinearIOSystem at some point, as pointed out already in issue #786).

@murrayrm thanks for the pointer. I refactored the code used in iosys.linearize to be a general-use method _copy_names that now lives in namedio that does the copying of various dicts correctly over to a new system.

Questions:

1. I moved naming defaults associated with linearization and sampling to namedio because they are related to naming. OK?
2. I changed the keyword to copy names to copy_names from copy in linearize and sample because I thought copy might somehow suggest copying the whole system somehow and this better represented what it should do. This does (slightly) break backward compatibility. OK?
3. the new sample allows copying the names but also supplying new keyword arguments that additionally allow those signal labels to be overridden. This is suggested for how it works in the linearize docstrings but what actually happens is that copy ignored any keyword-supplied names. I changed this so that supplied names override copied names. OK?
4. sample copies the names over by default because that is how they are likely to be used, but currently linearize has the old behavior of not copying over the names by default, but I am not sure if this is the best. sound reasonable?

I like the revisions and pretty much agree with them all, especially moving the functionality up to namedio.py (which is where it belongs). Some additional thoughts:

• One way to preserve backward compatibility would be to use kwargs to allow copy to still be used, but with a deprecation warning.
• Rather than $copy, it seems like $sampled would be a better choice for sample_system().
• We might want to create a copy method for NamedIOSystem and use the $copy suffix there.

Great. New commit:

• simplifies _copy_names to remove unnecessary functionality
• sets default behavior to copy_names=True in iosys.linearize. happy to revert it if you think it's better not to.
• fix docstring errors
• more unit tests

Following @murrayrm's comment above:

• linearize now accepts copy keyword, with a deprecation warning
• namedio.copy now does a deepcopy instead of a shallow copy, addressing InputOutputSystem.copy makes shallow copies #728. Before, all of its dicts and arrays were just references to objects that could be modified. The new system has $copy appended to its name, as before.
• dtime.sample_system now gets all of the naming functionality that is now present in sys.sample
• sample and sample_system append $sampled to the system name by default

This all looks good to me. Will let it sit for another day or two just in case someone else wants to take a look, but I think it is ready to merge.