Convert to use pydantic

[GDYendell]

Fixes #65
Closes #9 as not needed

[codecov-commenter]

Codecov Report

Attention: 14 lines in your changes are missing coverage. Please review.

Comparison is base (3e0640a) 88.55% compared to head (aaaca3f) 88.55%.
Report is 3 commits behind head on main.

Files	Patch %	Lines
src/pvi/_yaml_utils.py	82.05%	7 Missing ⚠️
src/pvi/_pv_group.py	40.00%	3 Missing ⚠️
src/pvi/__main__.py	87.50%	1 Missing ⚠️
src/pvi/_convert/_asyn_convert.py	83.33%	1 Missing ⚠️
src/pvi/_format/screen.py	96.42%	1 Missing ⚠️
src/pvi/device.py	99.00%	1 Missing ⚠️

❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #70      +/-   ##
==========================================
- Coverage   88.55%   88.55%   -0.01%     
==========================================
  Files          22       23       +1     
  Lines        1363     1328      -35     
==========================================
- Hits         1207     1176      -31     
+ Misses        156      152       -4     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[GDYendell]

Here it is @gilesknap

[GDYendell]

One thing that I don't like about this is that the type discriminator field of the models is not considered special and that means serialization cannot use exclude_default=True, because the Literal strings are the default and only possible value for the field. It can only use exclude_none=True. This means that any field of a model that shouldn't appear in the converted output cannot use default values, it must use None and manually convert to the default value in a getter or in the caller code.

I did quite a lot of searching and I think this is currently an unsolved problem in pydantic v2. It is quite a lot of extra work for the user to make sure the discriminator is included in the serialized output so that the code knows what specific type to instantiate when deserializing. Maybe there will be a better way in future.

[gilesknap]

One thing that I don't like about this is that the type discriminator field of the models is not considered special and that means serialization cannot use exclude_default=True, because the Literal strings are the default and only possible value for the field. It can only use exclude_none=True. This means that any field of a model that shouldn't appear in the converted output cannot use default values, it must use None and manually convert to the default value in a getter or in the caller code.

I did quite a lot of searching and I think this is currently an unsolved problem in pydantic v2. It is quite a lot of extra work for the user to make sure the discriminator is included in the serialized output so that the code knows what specific type to instantiate when deserializing. Maybe there will be a better way in future.

I also see this problem in ibek. It makes the yaml more verbose that necessary but is not an absolute disaster.

[gilesknap]

I'm not sure I can do a meaningful review of this code without spending a week or so getting my head into it. I will instead, try out this version against ibek and see if the example usage that I have still works.

Will try to get that done this afternoon 15 Jan.

[gilesknap]

I tried this out in an ioc-adsimdetector example. I noticed that ibek's call to the constructor of IndexEntry needed to have its positional arguments made into keyword.

I then verified that it ran OK and generated the same bob files as the previous version.

So I think it's good to go.