With:
```
sampleObject:
type: object
additionalProperties:
$ref: '#/components/schemas/SampleData'
minProperties: 1
description: Map of Charging data policy decisions.
```
and
```
SampleData:
type: object
...
nullable: true
```
We currently generate: HashMap<String, SampleData>, which doesn't allow
null charging data entries. This MR changes this to be
`HashMap<String, swagger::Nullable<SampleData>>`, which thus will allow null data entries.
We do this by moving null-handling to the Java code - primarily `getTypeDeclaration()`.
Note, to some extent this is wrong. In this MR (and previously) we are treating
`nullable: true` as an extrinsic property (like required), whereas it should be an
intrinsic property (and thus `HashMap<String, SampleData>` is correct, but `SampleData`
absorbs the nullability.
This would be possible with this code:
```
enum ChargingData = {
Null,
Present {
...
}
}
```
Which would remove the usage of https://docs.rs/swagger/2.0.2/swagger/nullable_format/enum.Nullable.html.
I haven't resolved this - and have instead done a more targeted fix.
This, along with some other crude code, creates a scenario where we need to
unpick whether something is null. I've left that, though flagged a TODO to tidy it up at some point.
* [Rust Server] Fix up model generation
- Correctly generate anyOf/oneOf models
- Fix up ToString / FromStr support
- Disable PartialOrd generation for anyOf/oneOf models
- Generate models for inline enums
- Support enums in headers, and vectors of models in headers
* [Rust Server] Add test for anyOf with additional properties
* Update samples
* [Rust Server] Tidy up logging
* [Rust Server] Allow configuration of multipart/form attachment size limit
multipart 0.14+ imposes a 8MB size limit on multipart/form bodies.
This allows that limit to be configured. The default is left as is.
This also improves error messages produced when handling multipart/form bodies.
* Update samples
* Added authentication via Bearer-token api_key and basic for server and client
* Improved errorhandling
* Added check on oAuth audience to example
* Updates of the petstore files for Rust-server
* Moved module import to prevent issue in callbacks
* updated samples
* Fix for unused-qualifications issue
* updated sampmles
* refactor: move closure definition to own statement
A clippy lint recommends not using a closure inside of a statement.
The code generation used to produce code that triggered this warning,
which caused the rust-server integration test to fail because clippy is
configured to return lint violations as errors.
For details for the lint see:
https://rust-lang.github.io/rust-clippy/master/index.html#blocks_in_conditions
* refactor: Remove unnecessary qualification
A GitHub action test used to fail because of a clippy warning that was
emitted due to not using an imported symbol but qualifying its use.
A failed test can be seen here:
https://github.com/OpenAPITools/openapi-generator/actions/runs/8958257509/job/24603984954?pr=18563
With the relevant error being:
error: unnecessary qualification
--> output/openapi-v3/src/client/callbacks.rs:88:9
|
88 | futures::future::ok(Service::new(
This commit simply removes the qualification.
* test: Update examples and run integration test.
Updated examples by running
`./bin/generate-samples.sh ./bin/configs/rust-server-*`
The integration test with the following command passes:
`mvn integration-test -f samples/server/petstore/rust-server/pom.xml`
* Fix clippy errors (rustc 1.73.0)
* Add feature docker-in-docker
* Fix mapping of "date"
See issue #9769
The type
type: string
format: date
was mapped to DateTime<Utc> which violates the OpenAPI spec
see https://swagger.io/docs/specification/data-models/data-types/
* [Rust][Server] Upgrade hyper from 0.13 to 0.14, swagger-rs from 5.0.2 to 6.1.0
* Also upgrade tokio from 0.2 to 1.14
* Re-add JavaClientCodegenTest.java
* Fix incorrect slashes due to wrong generation
* Spelling fix: failuare -> failure
* Upgrade more packages
* Commit generated code
* Fix typo
Co-authored-by: Foorack / Max Faxälv <max@foorack.com>
* [Core] Inline Model Resolution of Enums
Enums need to be named types, so handle them as part of inline model resolution
* [Rust Server] Handle models starting with a number correctly
* [Rust Server] Additional trace
* [Rust Server] Add support for oneOf/anyOf
* [Rust Server] Update supported features
* [Rust Server] General template tidy up
* [Rust Server] Implement IntoHeaderValue for wrapped data types
* [Rust Server] Convert from string correctly
* [Rust Server] Test for anyOf/oneOf
* Update samples
* Update docs
* [Rust Server] Pass context to client middleware
When creating a client with middleware, pass the context.
This allows users to pass contextual data about the requests through to the
client middleware.
* Update samples
* [Rust Server] Support complex query parameters
* [Rust Server] Remove unused extern crate
* [Rust Server] tokio should be a dev-dependency
* [Rust Server] Add test for complex query params
* Update samples
Co-authored-by: Matt Williams
* Remove test file erroneously checked in
* [Rust Server] Fix no features build
- Need serde_json for no-features build for undefined value structures
- Don't include `IntoHeader` type and implementations if we aren't including the client/server features
- Don't export the `IntoHeader` type at all - it's internal
* Update samples
* [Rust Server] Don't use structs in models
This avoids namespace clashes between model names and types used.
* [Rust Server] Handle models named after results
* [Rust Server] Add test for result models
* Update samples
* [Rust Server] Make parse error displayable
Change error type to be displayable to prevent compile errors
* [Rust Server] Add test for enum in path
* Update samples
[Rust Server] Hyper 0.12 Support
- Hyper upgraded to Hyper 0.12.
- NewService becomes MakeService
- Request on MakeContext is not parameterised - instead, ReqBody, and ResBody are parameterized, and must implement hyper::body::Payload.
- This means that our existing tuples (i.e. (Body, Context)) don't work - instead we have a type ContextualPayload, which implements Payload, and derefs to an inner payload (the body), and contains a context.
- This is handled by the work done in Metaswitch/swagger-rs#63 but has some fall out here as well.
- tokio-proto/tokio-core is no longer supported, and hyper instead depends on tokio.
- Hyper depends on mime 0.3, but multipart depends on mime 0.2, so we now import both
- Hyper TLS 0.2 and native-tls 0.1
- Use Swagger Support library 3.x
- Futures are now Send. This forces ApiImpl to be need to Send as a result, and all forces Clients to be Send + Sync.
- Ensure response IDs are unique, even if the first sentence is identical
- Handle responses where different responses produce a different type of data
- Use the correct mime type for plain text and byte streams
- Tidy up whitespace in client-mod.mustache
- Specify locale for String.format
- Add test for multiple response types
- Update samples
[Rust Server] Support multipart/form_data
- Support multipart/form_data in the Rust Server
- Add a new test API to test the change.
- Update the examples to match
Richard Whitehouse
Richard Whitehouse
Nathaniel Cook
cvkem
Max Nachlinger
myz-dev
Markus Lenger
Erik Wegner
Naoki Ikeguchi
Nathan Baulch
Jacob Halsey
Naoki Ikeguchi
William Cheng
Samodya Abey
Jérémie Bresson
Richard Whitehouse