Add GetterDict example

[nuno-andre]

Here is a very simple example of how to subclass GetterDict to process a class with no mapping protocol while dealing with the impedance mismatch between models (DOM's attributes and children vs OOP's properties).

I have added three optional fields to show "missing attribute", "missing children", and "missing children attribute".

closes #1745

[codecov]

Codecov Report

Merging #2463 (cb0f115) into master (429b439) will not change coverage.
The diff coverage is n/a.

❗ Current head cb0f115 differs from pull request most recent head 010ce73. Consider uploading reports for the commit 010ce73 to get more accurate results

@@            Coverage Diff            @@
##            master     #2463   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files           25        25           
  Lines         5099      5100    +1     
  Branches      1048      1048           
=========================================
+ Hits          5099      5100    +1     

Impacted Files	Coverage Δ
pydantic/main.py	100.00% <0.00%> (ø)
pydantic/generics.py	100.00% <0.00%> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 429b439...010ce73. Read the comment docs.

[nuno-andre]

I've also documented the _missing sentinel behavior. But I'm not quite sure that's the intended way to raise missing field errors, as the GetterDict code seems to suggest that KeyError should do so (which would make a lot of sense IMO). However, this is not the case.

Could you please confirm this assumption?

[PrettyWood]

@nuno-andre Thanks a lot for this doc, it was needed!
And good remark for the KeyError thing, which would indeed make sense.
But for now better stick with what you did and maybe open a discussion to suggest a refacto?

[nuno-andre]

Of course, just wanted to confirm that it wasn't an unintended behavior or a regression.

[PrettyWood]

@nuno-andre can you add a change file please? :)

[nuno-andre]

Done! I've just realized I'd completely ignored the PR template. Sorry for that.

[samuelcolvin]

otherwise looks good.

[samuelcolvin]

please update, otherwise I'll accept that one tiny change some time soon and merge.

[nuno-andre]

I guess this CI error isn't related to the PR, right?

[samuelcolvin]

no, that was just a problem with that CI script.