While working with the Magento 2.4 API definition, there were several
issues found. The easiest to resolve is a documentation naming issue.
(The others will be raised for discussion.)
This is known to be an issue with query parameters, but may not be
limited to *just* query parameters.
In the event that a parameter is a mix of arrays and objects, the
template in the documentation for the API function was expanding out the
parameter name such that `foo[0]` would be turned into
`:foo_left_square_bracket0_right_square_bracket`, but the actual
parameter name was being left as `:foo[0]`, which meant that the
documentation did not reflect the actual parameter.
> Note: there are issues with the way that query parameters are put
> together in this sense, which is going to require substantially more
> work to resolve as well as discussion on how these options should be
> implemented, as what the Magento API requires may not be what is
> expected by a different server, and the nature of the input parameters
> is itself incorrect.
* [python-experimental] Enhance octet-stream deserialization
When the headers didn't provide the filename, use the url of response to
extract filename.
* [python-experimental] Remove todo comment.
* [python-experimental] Fix test code.
* Update samples
* [python-experimental] Refined the method and the test
+ Early return when the url is empty or `None`.
+ Removed unused f-string prefix.
* [python-experimental] Comapre url is None explicitly.
* Update samples.
* Further Elixir Client Improvements
Resolves#12731 and is the completion of the work that I started with
#12751.
The changes here are extensive and likely resolve an issue that I have
seen with the Ory SDK (ory/sdk#194). I have also been unable to run the
integration suite for Elixir as I am (trying) to run everything in
Docker (`./run-in-docker.sh`) as I *do not* have a suitable Java
development environment set up, and do not do enough Java work to really
justify it.
- Updated the README for Elixir projects. Aside from some improved
readability of the template by use of link references instead of
inline links, I have also fixed the examples:
- The `deps` example should have been putting a version constraint
related to `appVersion`.
- The `config` example should have been using `packageName` instead of
`appName`. This particular issue repeats.
- In all Elixir files:
- Changed the function `@docs` formatting:
- changed the ehading level for `Parameters` and `Returns` to h3
(`###` instead of `##`). This will make somewhat better looking
documentation that does not over-emphasize these details (which
are *not* documented in a normal Elixir way, but this is somewhat
to be expected with a code generator.) It may be desirable, after
testing, to change this to `h4` instead of `h3`.
- Put parameter names and most return types in in-line code blocks
(`` `hello` ``).
- Put return types, when there are multiple types, in a Markdown
list.
- Fixed a lot of the spacing. Most files will be *closer* to Elixir
standard formatting than they were. Because of the limitations of
Mustache, it is still recommended that people who generate Elixir
clients run `mix format` at least once on their codebase.
- `api.mustache`:
- Removed an awkward function pipeline call. If we specified at least
Elixir 1.12 (something that I do not recommend as we have recently
jumped from requiring Elixir 1.6 to Elixir 1.10), there is a better
way to specify this now with `Kernel.then/2`. In the meantime,
assigning the constructed request structure to a variable and then
making a separate pipeline for the request execution and handling
makes for *much* easier to read generated code.
- Fixed the extra space issue with `evaluate_response` call tuple
values; `{{=<% %>=}}` changes the tag types, so this change is
intentional.
- In `config.exs.mustache`, `runtime.exs.mustache`, `mix.exs.mustache`,
and `connection.ex.mustache`, use `packageName` instead of `appName`
for configuration specification. If `packageName` and `appName`
differed, we would end up with cases like ory/sdk#194.
- `connection.ex.mustache` has been almost entirely rewritten. The
changes started in order to eliminate a `@doc` compile-time warning,
but shifted to remove the old way of building Tesla client structs
with `use Tesla`. It works, but is no longer the recommended way of
building Tesla clients.
- The *recommended* way of building a Tesla Client would now be
`Tesla.client(Connection.middleware(), Connection.adapter())`.
- Exposed both `Connection.adapter/0` and `Connection.middleware/1`
for use. `Connection.middleware/1` has special handling for the
cases where OAuth2 or HTTP Basic Auth are defined in the
application, but do not currently handle any other auth methods.
- `deserializer.ex.mustache` has mostly been reformatted. There are
things that I do not like about it (I do not like pipelines with one
line), and I have expanded one function capture into an anonymous
function for readability.
- `request_builder.ex.mustache` has been updated with better
function and parameter descriptions and names. Please note that if
`request |> method(:delete) |> method(:post)` is supposed to produce
a `POST` operation, we will need to change from `Map.put_new/3` to
`Map.put/3`.
- Reordered `evaluate_response/2` so that it is the function documented,
and made `decode/2` and `response_mapping/3` private functions. As
far as I can tell, I have *not* changed the functionality.
* Address issues found in code review
- The example dependency code in the README had dropped the opening
brace for the tuple. This has been resolved.
- The default formatting of the API pipelines has been adjusted to
minimize possible changes from `mix format`.
* Update modules/openapi-generator/src/main/resources/elixir/api.mustache
Co-authored-by: Michael Ramstein <633688+mrmstn@users.noreply.github.com>
* Update modules/openapi-generator/src/main/resources/elixir/connection.ex.mustache
Co-authored-by: Michael Ramstein <633688+mrmstn@users.noreply.github.com>
* Update modules/openapi-generator/src/main/resources/elixir/connection.ex.mustache
Co-authored-by: Michael Ramstein <633688+mrmstn@users.noreply.github.com>
* Update templates based on review comments
Co-authored-by: Michael Ramstein <633688+mrmstn@users.noreply.github.com>
* Bump the minimum version of Elixir supported
The previous minimum version of Elixir is several years EOL.
The current minimum version of Elixir is also EOL, but is the minimum
version required to support some upcoming changes to the config
templates.
* Bump the minimum version fo Tesla
Keep the dependencies up to date
* Add a default .formatter.exs
* Add two Elixir-specific mustache lambdas
- The `atom` lambda results in the proper quoting of an atom depending
on the safe contents of the atom text, per the Elixir language
specification. That is, `{{#atom}}foo{{/atom}}` will be turned into
`:foo` and `{{#atom}foo.bar{{/atom}}` will be turned into
`:"foo.bar"`.
- The `env_var` lambda results in the treatment of the identifier
provided being capitalized as an environment variable would be.
`{{#env_var}}apiVersion{{/env_var}}` would become `ENV_VAR`.
* Use modern Elixir configuration
- This includes runtime configuration
- It depends on the `env_var` lambda.
* Fix a Language Server Warning
This change is *optional*, but removes a LS warning that was raised.
* Regenerated openapi_petstore for Elixir
* Add ex_doc as a default dependency
Fixes#12484
* Refine the regular expression for atoms
The original regex incorrectly matched `123Number` (unquoted atoms
cannot begin with numbers) and would incorrectly quote atoms ending in
`?` or `!`. Through testing with `iex`, it also turns out that the atom
`:-` is legal.
The following atoms will now not be quoted that would have been
incorrectly quoted:
- `:-`
- `:declawed?`
- `:neutered!`
The following atoms will be quoted that were incorrectly unquoted:
- `:"123Number"`
* Improve regex (again), remove files not generated
- The previous commit resulted in a number of warnings that were still
present and so I played with the regular expression. This did not
solve the problem, but the resulting regular expression is *much*
better than the previous one, so I'm keeping it.
- The problem was that the configuration (`bin/configs/elixir.yaml`) is
generated using a 3.0 input spec:
```yaml
inputSpec: modules/openapi-generator/src/test/resources/3_0/petstore-with-fake-endpoints-models-for-testing.yaml
```
Which means that there were 16 files committed which were no longer
being generated. When I tested with the 2.0 input spec:
```yaml
inputSpec: modules/openapi-generator/src/test/resources/2_0/petstore-with-fake-endpoints-models-for-testing.yaml
```
The files were generated again. I *believe* that the correct change
here is to switch back to the 2.0 input spec, as it tests more code
generation, but I wanted to check in before I did this.
The following files are deleted:
- `elixir/lib/openapi_petstore/model/additional_properties_any_type.ex`
- `elixir/lib/openapi_petstore/model/additional_properties_array.ex`
- `elixir/lib/openapi_petstore/model/additional_properties_boolean.ex`
- `elixir/lib/openapi_petstore/model/additional_properties_integer.ex`
- `elixir/lib/openapi_petstore/model/additional_properties_number.ex`
- `elixir/lib/openapi_petstore/model/additional_properties_object.ex`
- `elixir/lib/openapi_petstore/model/additional_properties_string.ex`
- `elixir/lib/openapi_petstore/model/big_cat.ex`
- `elixir/lib/openapi_petstore/model/big_cat_all_of.ex`
- `elixir/lib/openapi_petstore/model/inline_response_default.ex`
- `elixir/lib/openapi_petstore/model/special_model_name.ex`
- `elixir/lib/openapi_petstore/model/type_holder_default.ex`
- `elixir/lib/openapi_petstore/model/type_holder_example.ex`
- `elixir/lib/openapi_petstore/model/xml_item.ex`
- `elixir/pom.xml`
- `elixir/test/pet_test.exs`
In the interim, I have removed those files from the commit.
* Allow the baseUrl of elixir APIs to be overridden
* Run generator
* Add missing `:`
* Update modules/openapi-generator/src/main/resources/elixir/connection.ex.mustache
Co-authored-by: Michael Ramstein <633688+mrmstn@users.noreply.github.com>
* Generate sample with new change
Co-authored-by: Joe Eifert <joe@databerg.rocks>
Co-authored-by: Michael Ramstein <633688+mrmstn@users.noreply.github.com>
* update parser to 2.0.29
* better handling of null in dereferencing
* update parser to 2.0.30
* update core to newer version
* add new files
* rollback to previous stable version
* remove files
* Fixes for python-experimental NullableShape component
Co-authored-by: Justin Black <justin.a.black@gmail.com>
Refs #3358
Ensure `deprecated` operations are annotated/documented as such on the
generated methods. Libraries updated:
* [feign]
* [google-api-client]
* [microprofile]
* [okhttp-gson]
* [resttemplate]
* [retrofit]
* [retrofit/play*]
* [webclient]
* [vertx]
Ensure `deprecated` schemas are annotated/documented as such on the
generated classes/fields. Libraries updated:
* [feign]
* [google-api-client]
* [jersey2]
* [microprofile]
* [native]
* [okhttp-gson]
* [rest-assured]
* [resteasy]
* [resttemplate]
* [retrofit*]
* [webclient]
* [vertx]
Also fix two minor bugs to get the java sample tests working:
* Fix an invalid jackson-datatype-threetenbp version number in vertx/pom.mustache
* Fix a bad return type in webclient/api_test.mustache when uniqueItems=true
Since this commit updates petstore-with-fake-endpoints-models-for-testing.yaml,
several other samples were updated, but it's just new files to reflect the
deprecated schemas, so there should be no consequential differences.
Relevant bits of the spec:
* https://github.com/OAI/OpenAPI-Specification/blob/main/versions/3.0.2.md#user-content-operationdeprecated
* https://github.com/OAI/OpenAPI-Specification/blob/main/versions/3.0.2.md#user-content-schemadeprecated
