feat(dynamite)!: add experimental standalone JsonSchema generation

[Leptopoda]

implements: #933

[provokateurin]

First commit message has a typo and should be feat(dynamite): do not rely on the openapi spec for type resolving

[provokateurin]

Quite a bit of this code is just copied from the existing builder and should be factored out into common methods.

[Leptopoda]

I'm not so sure how much we can deduplicate this without making a mess of the code.
I'll take another look on what makes sense.

[provokateurin]

This is not a schema definition: https://json-schema.org/learn/getting-started-step-by-step#create
We should really implement the whole thing and not just the schema part itself.

[Leptopoda]

I don't understand? what is still missing?

[provokateurin]

The metadata that can be inside a schema definition is not considered in the generated code. For OpenAPI specs we add title, description, license etc. to the top of the file while this doesn't seem to happen for JSON schema definitions afaict.

[provokateurin]

Sorry my initial message was a bit confusing, I'd simply like to have proper support for JSON schema definitions.

[Leptopoda]

Can you please say what JsonSchema keywords you'd like to support?

It appears that the only user facing information is the title and I don't think that it should be set as a library level doc comment. I'd rather add it to the generated class itself similar to how the dart doc style expects the documentation:

/// {title}.
///
/// {description}.
class MyObject {}

[provokateurin]

Yes the comment should be on the generated class. IMO the $id can also be relevant for users, so I'd like to have that linked as well.

[provokateurin]

Since there is $defs it is possible to have multiple object inside a single schema. I don't know if the title and description should be interpreted as the title and description for the main object or the whole schema.

[Leptopoda]

In theory we already support $defs. our reference resolving is independent from the dart schema object and directly searches in the json data (see: #1953).

Therefore only the schemas referenced in the main schema will be generated. Or do you mean to generate all schemas defined in the definitions just in case 🤷‍♀️

[provokateurin]

No this was about the other fields, not specifically $defs. I was just saying that title and description might not be for the main schema but for the whole schema which would make them library level.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 29.31%. Comparing base (729e5ae) to head (44e594b).
Report is 5 commits behind head on main.

❗ Current head 44e594b differs from pull request most recent head a1705e0. Consider uploading reports for the commit a1705e0 to get more accurate results

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1964      +/-   ##
==========================================
+ Coverage   28.43%   29.31%   +0.87%     
==========================================
  Files         245      245              
  Lines       81449    76519    -4930     
==========================================
- Hits        23161    22430     -731     
+ Misses      58288    54089    -4199     

Flag	Coverage Δ		*Carryforward flag
dynamite	22.46% <ø> (-0.26%)	⬇️
dynamite_end_to_end_test	62.98% <ø> (+0.35%)	⬆️
dynamite_runtime	82.65% <ø> (ø)		Carriedforward from a3c55c1
neon_dashboard	92.56% <ø> (ø)
neon_framework	∅ <ø> (∅)		Carriedforward from a3c55c1
neon_talk	∅ <ø> (∅)		Carriedforward from a3c55c1
nextcloud	∅ <ø> (∅)		Carriedforward from a3c55c1
sort_box	90.90% <ø> (ø)		Carriedforward from a3c55c1

*This pull request uses carry forward flags. Click here to find out more.

see 160 files with indirect coverage changes

[Leptopoda]

Making a draft for now until we have merged all the depending changes.

[provokateurin]

$defs should also be supported as it acts like the components.schemas

[provokateurin]

@Leptopoda can we move this forward and include it in the next batch of releases or is there something holding it back?

[Leptopoda]

I don't think this is anything we can immediately implement.

$defs should also be supported as it acts like the components.schemas

This requires a complete rework of how our $ref resolving works and this will take quite some time ...