Enhance split words

[ramnes]

Keep "Response200" as "response_200", but don't make "s3config" become "s_3_config". :)

Related to the comments here: #375 (comment)

[codecov]

Codecov Report

Merging #413 (d6cce56) into main (a731477) will not change coverage.
The diff coverage is 100.00%.

❗ Current head d6cce56 differs from pull request most recent head 999a81d. Consider uploading reports for the commit 999a81d to get more accurate results

@@            Coverage Diff            @@
##              main      #413   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files           47        47           
  Lines         1549      1550    +1     
=========================================
+ Hits          1549      1550    +1     

Impacted Files	Coverage Δ
openapi_python_client/utils.py	100.00% <100.00%> (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 a731477...999a81d. Read the comment docs.

[dbanty]

I'm not totally convinced and I'm worried that this could get bikesheddy and keep becoming more specialized. Like, what if someone wanted HTTP200 to be http_200? Or in the golden records examples, type1 really is type_1 (meaning "the first type").

Happy to get input from others (e.g. @emann, @bowenwr) but I'm tempted to call it good enough and stop making changes to names unless something is really broken. Maybe allowing a manual override of specific values would be good though (e.g. s_3_config -> s3_config), I think we could add that through the normal config process with minor effort.

[ramnes]

I think we don't want to guess the words if there is no capital letter (i.e. if we're not seeing a PascalCase or camelCase string) because that means we would only split on numbers and will never be able to guess any kind of other word. Plus OpenAPI specifications authors will use some form of delimiters if there really is multiple words that need to be separated for readability anyway.

As to the type_1 being changed to type1 in the golden records, I couldn't find were it came from and assumed it was something internal to openapi-python-client as to how it handles lists, or something in these lines. Am I wrong? If not, we could probably add an underscore or something at the source, before split_words is called.

[dbanty]

Oh I see, I misunderstood the change. For some reason I thought that it was making the snake case version of S3Config "s3_config" instead of "s_3_config" but I see now that it was already "s3_config" and the change here is to make the lowercase variant ("s3config") not be split at all. So my note about HTTP200 is irrelevant, both in main and in this branch it will be http200.

I agree with detecting the use of case and splitting based on that, sounds good. Let's just change build_union_property to put and underscore between type and {i} so the generated code doesn't flip around again 😄.

[ramnes]

Ah, done, thanks for the hint!

[dbanty]

Latest on main breaks e2e 😬. I made a branch enhance-split-words in this repo with the update (or you can just task re).

[dbanty]

Good to go as soon as the golden record conflict is resolved!

[ramnes]

Rebase went fine on my side, not sure what you meant. :)

[ramnes]

(and yeah, I did the task re)