[BUG][R] Generated samples/client/echo_api/r cannot be run

[jwimberl]

Bug Report Checklist

• Have you provided a full/minimal spec to reproduce the issue?
• Have you validated the input using an OpenAPI validator?
• Have you tested with the latest master to confirm the issue still exists?
• Have you searched for related issues/PRs?
• What's the actual output vs expected output?
• [Optional] Sponsorship to speed up the bug fix or feature request (example)

Description

The generated echo_api client for R breaks out of the box because of invalid R syntax:

openapi-generator/samples/client/echo_api/r/R/data_query.R

Line 37 in 4242641

initialize = function(`id` = NULL, `outcomes` = [SUCCESS, FAILURE], `suffix` = NULL, `text` = NULL, `date` = NULL, ...) {

The = [SUCCESS, FAILURE] has the appearance of Python code that has snuck in.

openapi-generator version

This is found in the master branch and is an issue in generated integration test code, not an issue found actually running the generator.

OpenAPI declaration file content or url

NA; this is an integration test bug

Generation Details

NA; this is an integration test bug

Steps to reproduce

Navigating to samples/client/echo_api/r, trying to load the package or run tests fails because of the syntax error:

r % Rscript -e 'devtools::test()'
ℹ Testing openapi
Error in `load_all()`:
! Failed to load R/data_query.R
Caused by error in `parse()`:
! At R/data_query.R:37:53: unexpected '['
36:     #' @param ... Other optional arguments.
37:     initialize = function(`id` = NULL, `outcomes` = [

Related issues/PRs

Not found

Suggest a fix

I am not quite sure what template source is responsible for this errant R code. According to

openapi-generator/bin/configs/r-echo-api.yaml

Line 4 in 4242641

templateDir: modules/openapi-generator/src/main/resources/r

this uses the same template as the working petstore R client tests; however, the latter three all use settings

  exceptionPackage: rlang
  useRlangExceptionHandling: true
  returnExceptionOnFailure: true

do to which they are following a quite different code generation path that might bypass wherever the issue is. I don't see SUCCESS or FAILURE anywhere in that template directory so I couldn't trace it back any further than that.

[jwimberl]

Also relevant is that in the corresponding query.R file, the SUCCESS and FAILURE values are quoted, but the array object is still created using [] instead of c().

Some LLM-assisted investigation has pointed to this issue arising from the underlying RClientCodegen.java. The LLM was able to fix the generated r-echo-api client (to the extent that the syntax is correct and it can be installed), but it took it fairly substantial changes in that file as well as several mustache files to do so, and I don't have the familiarity with the codebase to confirm if these are all necessary and minimal changes. It claimed that the atomic bugs in the R client generation that it needed to fix were

• arrays emitted as bracketed lists instead of c(), as noted in the issue description
• the generated validation for array properites all use is.character, even for non-character arrays
• multiline strings emitted as '''...''' instead of R syntax
• assorted issues regarding string escaping in generated code and docs

The fixes it added were primarily in the toDefaultValue method in RClientCodegen.java and some helper methods, branching on ModelUtils.isArraySchema(p) to begin with.