feat(http): use smarter bundling #15
Conversation
P0lip
force-pushed
the
jrozek/dev/reparent-target
branch
from
489c973 to
706bf40
Compare
P0lip
changed the title
P0lip
force-pushed
the
jrozek/dev/reparent-target
branch
from
706bf40 to
ff615cd
Compare
P0lip
force-pushed
the
jrozek/dev/reparent-target
branch
from
ff615cd to
19f9188
Compare
| const $ref = schema.$ref; | ||
| schema.$ref = rewriteRef($ref); | ||
| if (schema.$ref === $ref) return; | ||
| if (processedRefs.has($ref)) return; |
There was a problem hiding this comment.
Jakub - is this where you're short-circuiting so we don't try to rebundle already bundled refs? I guess I'm trying to work out why you need both 138 and 139 🤔
| function rewriteRef(value: string) { | ||
| return `#/${ROOT_KEY}${value.slice(1).replace(`/${ROOT_KEY}/`, '/')}`; | ||
| } |
There was a problem hiding this comment.
I am positive this works and makes sense man but I'm struggling a bit - prolly just too early for me - why are you doing this rewriteRef, what's the overall interplay between traversed, seen, and processedRefs collection?
Ahhh maybe... yeah ok I think I'm seeing it - the rewrite points the ref from some external reference to the new, bundled reference. Which you only need to add to the document once, the first time you rewrite a ref to it... but you still need to rewrite each existing ref to point to the now bundled schema - yeah ok, think I've got it sussed a bit, let me know if I'm way off lol
Summary
This pull request includes several changes to the
packages/httpmodule to support more performant bundling.It exploits the today's behavior of http-spec, and the tooling used today, and is strictly tailored to them.
Some changes were needed to generators in
packages/http/src/mocker/generator/JSONSchema.tsto properly persist the non-enumerable property.Checklist