Yicong-Huang opened a new issue, #8757:
URL: https://github.com/apache/texera/issues/8757
## What happened?
Exporting a workflow containing one of the four advanced sklearn trainers
(KNN classifier/regressor, SVC, SVR) via `POST /workflow-to-python` produces a
Python script that cannot be parsed when the operator's parameter table
(`paraList`) contains two rows naming the same parameter. The emitted estimator
call repeats the keyword, e.g. `SVC(C = float ('1.0'),C = float ('2.0'),)`, and
CPython raises `SyntaxError: keyword argument repeated` when the script is run.
The export itself reports success (`WorkflowToPythonSuccess`); the failure only
appears when the user runs the script.
What I expected: the export path refuses the duplicate the way the native
path does. `SklearnAdvancedBaseDesc.getParameter` (used by the in-engine
`generatePythonCode`) starts with `requireDistinctParameters(paraList)`, whose
own comment states the reason: "two rows naming one parameter emit that keyword
twice and Python rejects the operator with a repeated-keyword SyntaxError.
Refused here instead, where the workflow fails to compile with the parameter
named." `generateStandaloneCode()` calls `getParameterStandalone(paraList)`
directly and never invokes that guard, so the new export path lacks the
backstop the PR itself added to the native path.
## How to reproduce?
1. Build a workflow JSON whose advanced trainer operator (e.g.
`SklearnAdvancedSVCTrainerOpDesc`) has two `paraList` rows with the same
`parameter` (e.g. two rows naming `C`). Note the form-level `uniqueAmongRows`
validator blocks this interactively, so construct the plan by hand or via the
API; stored workflows are not schema-validated on load, and the export endpoint
takes the `LogicalPlanPojo` with no operator-property validation.
2. `POST` the plan to `/workflow-to-python`.
3. Run the returned script with Python: it fails with `SyntaxError: keyword
argument repeated`.
For comparison, running the same operator natively fails at compile time
with the parameter named (the `require` throws inside `generatePythonCode`, and
`getPhysicalOp` embeds it as `#EXCEPTION DURING CODE GENERATION:`).
## Version/Branch
1.4.0-incubating-SNAPSHOT (main)
## Commit Hash (Optional)
Verified still present at `acc6dd1139d936d849e2aaa6abb288f8ef3e98ad` (main
tip, 2026-09-29): `requireDistinctParameters` is defined at
`SklearnAdvancedBaseDesc.scala:266` and called only from `getParameter`
(`:286`); `generateStandaloneCode` (`:447-450`) still bypasses it. Introduced
by #8348 (merge `5127b78bc5dfd48a904894792c299ec72ba07ab9`).
## What browsers are you seeing the problem on?
n/a (backend export endpoint)
## Relevant log output
n/a — the exporter returns success; the SyntaxError surfaces when running
the emitted script.
## Suggested fix
Call `requireDistinctParameters(paraList)` at the top of
`generateStandaloneCode()` (or inside `getParameterStandalone`), and add a
standalone regression test mirroring `SklearnAdvancedBaseDescSpec.scala:359`
("refuse two rows setting one parameter"), which currently exercises only the
native `getParameter`.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
