Codegen: Add new useHeadTagForObjectNames flag to permit splitting generated endpoint objects by tag
#3594
Conversation
|
Thanks! Could you add documentation for this new flag to |
|
Sure. Will do that in a few mins & push up a test, too |
|
Done! (In case you can't tell, this is a feature I really want 😂) |
|
Yes, I can tell 😁 Sorry to push back some more, but could you also explain in the docs what the "first declared tag" mean? This could be added outside of the table with flags. |
|
@kciesielski I've updated the copy a little more, let me know what you think Edit: Happy to make changes, please do comment on anything you're not happy with. Meantime I've brewed another pr for ya 😄 |
|
Thanks! Looks good to me :) |
hughsimpson
force-pushed
the
separate_endpoint_objects_by_tag
branch
from
d1c1f2d to
5a047c3
Compare
|
Pushed rebase to fix merge conflicts |
| .map { case (name, definition) => | ||
| s"""|lazy val $name = | ||
| val geMap = | ||
| doc.paths.flatMap(generatedEndpoints(components, useHeadTagForObjectNames)).groupBy(_._1).mapValues(_.map(_._2).reduce(_ ++ _)) |
There was a problem hiding this comment.
This would've been so much cleaner with groupMapReduce(_._1)(_._2)(_ ++ _) 😂 but, alas, scala 2.12...
I've been thinking about this for ages and finally got around to it. If autogenerating files for a very large spec, it's possible to hit the JVM limits on class size since all definitions go into the same object; so from a purely technical perspective, this does enable working with certain specs that would otherwise require some fudging (such as calling the codegen cli on multiple input files); although I confess to also just finding it a bit tidier. This pr adds a new flag
useHeadTagForObjectNames(defaulting to false, to preserve existing behaviour). If this flag is true then the generated endpoints are grouped by the first tag, and each group put into a new file+object with the same name as that tag. All types/enums/classes generated from schemas remain in the default object model as before, and if an endpoint has no tags at all, it is placed into the default object (i.e. a spec with no tags would see no change as a result of this pr, regardless of whether the flag were set to true or not)