Closed SnakeDoc closed 7 months ago
Welcome to AsyncAPI. Thanks a lot for reporting your first issue. Please check out our contributors guide and the instructions about a basic recommended setup useful for opening a pull request.
Keep in mind there are also other channels you can use to interact with AsyncAPI community. For more details check out this issue.
I am on this issue, but at the same time, the validate()
function of parser-js gives on validation of both 'expected' and 'actual' output error:
[
{
code: 'asyncapi3-operation-messages-from-referred-channel',
message: 'Operation message does not belong to the specified channel.',
path: [ 'operations', 'UserSignedUp', 'messages', '0' ],
severity: 0,
range: { start: [Object], end: [Object] }
},
{
code: 'asyncapi3-operation-messages-from-referred-channel',
message: 'Operation message does not belong to the specified channel.',
path: [ 'operations', 'TestOpp', 'messages', '0' ],
severity: 0,
range: { start: [Object], end: [Object] }
}
]
So, while I'm checking from my side, I would please ask you to simultaneously check the AsyncAPI Document's format from your side.
@SnakeDoc
Can I please ask you to be a beta-tester of my PR a bit?
Do
$ mkdir test
$ cd test/
$ git clone https://github.com/aeworxet/asyncapi-bundler.git .
$ git switch feat-add-x-origin-property # <-- essential, because "@apidevtools/json-schema-ref-parser" has another version in this branch
$ npm install
$ tsc
$ cd example/
$ npm install
# add your testing YAMLs
$ node bundle-cjs.cjs
and tell if the PR's code outputs to asyncapi.yaml
what you need.
If the resulting (dereferenced) AsyncAPI Document doesn't pass validation, it outputs an error, similar to
Validation of the resulting AsyncAPI Document failed.
List of remarks:
[
{
code: 'asyncapi3-operation-messages-from-referred-channel',
message: 'Operation message does not belong to the specified channel.',
path: [ 'operations', 'UserSignedUp', 'messages', '0' ],
severity: 0,
range: { start: [Object], end: [Object] }
},
{
code: 'asyncapi3-operation-messages-from-referred-channel',
message: 'Operation message does not belong to the specified channel.',
path: [ 'operations', 'TestOpp', 'messages', '0' ],
severity: 0,
range: { start: [Object], end: [Object] }
}
]
and exits without writing to asyncapi.yaml
.
Currently, bundler is only resolving the references in channels and operations, I think that is why info
references are not getting resolved.
Thank you @aeworxet - great work on this feature so far. I've been playing with your fork and find it works well for me.
I did have a few errors in my original files that caused the validation issues you pointed out.
I'm attaching new files that pass validation, and resolve correctly using your beta bundler.
main.yaml:
asyncapi: 3.0.0
info:
$ref: './info.yaml#/info'
channels:
userSignedup:
address: 'user/signedup'
messages:
userSignedUpMessage:
$ref: './messages.yaml#/messages/UserSignedUp'
test:
address: '/test'
messages:
testMessage:
$ref: '#/components/messages/TestMessage'
operations:
UserSignedUp:
action: send
channel:
$ref: '#/channels/userSignedup'
messages:
- $ref: '#/channels/userSignedup/messages/userSignedUpMessage'
TestOpp:
action: send
channel:
$ref: '#/channels/test'
messages:
- $ref: '#/channels/test/messages/testMessage'
components:
messages:
TestMessage:
payload:
type: string
info.yaml:
info:
title: Account Service
version: 1.0.0
description: This service is in charge of processing user signups
messages.yaml:
messages:
UserSignedUp:
payload:
type: object
properties:
displayName:
type: string
description: Name of the user
email:
type: string
format: email
description: Email of the user
UserLoggedIn:
payload:
type: object
properties:
id:
type: string
I have only been validating my files after bundling, which is why I didn't notice the validation errors before posting. My view has been the multi-file spec is treated as "source" and the bundler produces the final output. I can see flaws in this logic though, and do agree it is better to have validation at each step. I assume asyncapi is promoting a src -> bundler -> optimizer workflow, with validation each step of the way.
I, however, was not able to resolve references to entire channel and operation objects, due to how the bundler resolves objects in-line instead of moving them to the components section for reuse. It would be very convenient to be able to do something like:
main.yaml:
asyncapi: 3.0.0
info:
$ref: './info.yaml#/info'
channels:
userLoggedIn:
$ref: './channels.yaml#/channels/userLoggedIn'
operations:
UserLoggedIn:
$ref: './operations.yaml#/operations/UserLoggedIn'
info.yaml:
info:
title: Account Service
version: 1.0.0
description: This service is in charge of processing user signups
messages.yaml:
messages:
UserLoggedIn:
payload:
type: object
properties:
id:
type: string
channels.yaml:
channels:
userLoggedIn:
address: 'user/loggedin'
messages:
userLoggedInMessage:
$ref: './messages.yaml#/messages/UserLoggedIn'
operations.yaml
operations:
UserLoggedIn:
action: send
channel:
$ref: './channels.yaml#/channels/userLoggedIn'
messages:
- $ref: './messages.yaml#/messages/UserLoggedIn'
The above gets a validation error:
Validation of the resulting AsyncAPI Document failed.
List of remarks:
[
{
code: 'asyncapi3-operation-messages-from-referred-channel',
message: 'Operation message does not belong to the specified channel.',
path: [ 'operations', 'UserLoggedIn', 'messages', '0' ],
severity: 0,
range: { start: [Object], end: [Object] }
}
]
Failed Document:
{
"asyncapi": "3.0.0",
"info": {
"x-origin": "./info.yaml#/info",
"title": "Account Service",
"version": "1.0.0",
"description": "This service is in charge of processing user signups"
},
"channels": {
"userLoggedIn": {
"x-origin": "./channels.yaml#/channels/userLoggedIn",
"address": "user/loggedin",
"messages": {
"userLoggedInMessage": {
"payload": {
"type": "object",
"properties": {
"id": {
"type": "string"
}
}
}
}
}
}
},
"operations": {
"UserLoggedIn": {
"x-origin": "./operations.yaml#/operations/UserLoggedIn",
"action": "send",
"channel": {
"$ref": "./channels.yaml#/channels/userLoggedIn"
},
"messages": [
{
"$ref": "./messages.yaml#/messages/UserLoggedIn"
}
]
}
}
}
This error seems to be due to bundler resolving objects in-line instead of moving them to the components section for reuse. You can see how the channel message gets resolved in-line, and the operation fails to resolve entirely.
Maybe optimizer takes care of this, but it would be nice to have it happen at bundling time so the source documents can be written more efficiently. Maybe I missed something though?
Might be wishful thinking, but it would be even more convenient to bulk reference, like:
main.yaml:
asyncapi: 3.0.0
info:
$ref: './info.yaml#/info'
channels:
$ref: './channels.yaml#/channels'
operations:
$ref: './operations.yaml#/operations'
Overall, the beta bundler clearly seems headed in a great direction.
After completion of https://github.com/asyncapi/bundler/issues/141, the process Bundler
->Optimizer
, to obtain an AsyncAPI Document with components moved to components
, will become mandatory in CI/code (about May 2024.)
Meanwhile, I'm gathering feedback during the development.
@SnakeDoc
Can you please take a look at https://github.com/asyncapi/bundler/issues/141#issuecomment-2017469779 and tell which version of the AsyncAPI Document after Optimizer
would serve your needs better, what behavior of Optimizer
on schemas
do you think should be default, and in what way would you prefer to toggle it?
(I'm leaning to a command-line switch that will be parsed in code, assigned to a property of some options
object, which then will be passed around the code to different functions)
@aeworxet Thanks for the follow-up. I've reviewed #141 and can offer a few thoughts:
1) Perhaps an issue for Optimizer, but I do not see value in retaining the x-origin
property post optimization. It seems to be an artifact left over from previous "stages" of generating the final document. From my observation and limited perspective, it appears the output of bundler will not be readily useful for most people, and they will want to run their bundled document through the optimizer. x-origin
may leak internal details about how the source of the document is structured, and doesn't seem provide useful information for codegen/end-users later on. This also probably means the CLI
should automatically do bundle -> optimize unless the user opts-out with a flag or something, ie. --disable-optimization
. Therefore my opinion is optimizer should consume/strip x-origin
before output of the final document.
2) I must preface that my experience comes from the OpenAPI side, which may bias my opinion - but I actually like the components/schema
option better. In my opinion, schemas are the smallest, reusable component within the document, and should be used to construct more complex objects, such as messages, with channels/operations built on top of messages, etc.
Two different messages can share the same payload (schema), but have different headers, etc, or even be exactly the same. Duplicating and in-lining the payload definition seems unnecessary, and perhaps incorrectly hints that the two definitions are not the same object.
I think the primary audience for bundled & optimized documents is mostly code/documentation generators, and to a much lesser extent, actual people. Therefore, the verbosity and complex references produced using the schema option probably doesn't matter much in my opinion.
Further - from a document writer perspective (document source, pre-bundler and pre-optimizer), I want to clearly communicate my intentions, and re-use as much as possible wherever possible (DRY). Modularity of various components also greatly enhances the maintainability of larger documents as the api grows.
I would love to do something like:
and produce this after bundler -> optimizer:
@SnakeDoc, thank you for the test case!
I had run this scenario through Bundler
-> Optimizer
process
with newly introduced switches
disableOptimizationFor: { schema: false }
disableOptimizationFor: { schema: true }
and your expectations meet everything except schemas
(this behavior is yet to be changed in the future.)
@KhudaDad414 Can https://github.com/asyncapi/bundler/issues/151#issuecomment-2030436419 be a hint for https://github.com/asyncapi/optimizer/issues/88 ?
@SnakeDoc A bit improved mechanism meets your expectations even better.
@SnakeDoc
x-origin
property originated from this comment in a year-long discussion
The $ref usually also carries a semantical meaning to understand easier what it is (example "$ref : financial-system.yaml#/components/schemas/bankAccountIdentifier"). If the bundling just resolves this ref inline, the semantical meaning of the $ref pointer gets lost and cannot be recovered in later steps.
and is essential to mindful properties' naming during optimization with Optimizer
The optimizer would need to invent an artificial component name for the "bankAccountIdentifier" when moving it to the components section.
So x-origin
needs to be there.
A switch might be introduced to remove it after the optimization or not to add it at all during bundling if a user doesn't need it, though; this point is completely valid.
@SnakeDoc, can you please check if it is understandable how to perform this process to achieve this result, using code from PRs https://github.com/asyncapi/bundler/pull/147 and https://github.com/asyncapi/optimizer/pull/216 ?
(@apidevtools/json-schema-ref-parser
's version was changed in Bundler
again)
@aeworxet excellent work on both #147 and asyncapi/optimizer#216!
With latest aeworxet/asyncapi-bundler:feat-add-x-origin-property, I was able to produce the following using the above example:
With latest aeworxet/asyncapi-optimizer:feat-add-flag-moveAllToComponents, and the below configuration, the following optimized document was produced:
Which makes me quite happy! :+1:
Disabling the schema optimization with "disableOptimizationFor": { "schema": true }
yields:
All of the above validates with @asyncapi/cli@1.8.3.
Coupled with the other newly available optimizer configurations, I imagine this covers just about it all. Very good work!
@SnakeDoc
Thanks for the testing!
The PRs https://github.com/asyncapi/bundler/pull/147 and https://github.com/asyncapi/optimizer/pull/216 are now known as Bundler v0.5.0
and Optimizer v1.0.0
, respectively.
There was a baseDir
option reimplemented in the last minute. You can read about how it works now at https://github.com/asyncapi/bundler/blob/master/README.md#option-basedir and observe it in action in ./example/bundle-cjs.cjs
.
https://github.com/asyncapi/bundler/blob/f4cc6d91820a1e672f7b05df9e8f9473fbc576b9/src/v3/parser.ts#L101
Referencing the linked line - is there a reason the bundler is not being used? As-is, resolving a v3 document only resolves references found within channels and operations. I might be missing something?
ends up outputting
$ref: './info.yaml#/info'
instead of resolving the value of the reference into the final document. The same goes for other parts of the v3 document, including tags, etc.Switching the spec version to 2.6.0 does result in correctly resolved references (ignoring differences between 2.x and 3.x specs).
main.yaml:
messages.yaml:
info.yaml:
expected output:
actual output:
the following test fails when run against the above files: