CausalML is not compatible with Scikit-Learn Pipeline

[PelinV]

Describe the solution you'd like
Swap the arguments treatment and y in the fit function.
To change the order of arguments in the fit function from X, treatment, y to X, y, treatment for compatibility with sklearn

Additional context
When using causalml models in the sklearn pipeline, it is necessary to write adapters
Because pipeline expects the following order: X, y, other arguments, and in fit in causalml the following order: X, treatment, y

[jbbqqf]

Hi! Triaging older issues — I think this one is a candidate for closure as out-of-scope.

Reordering fit(X, treatment, y) → fit(X, y, treatment) would break every meta-learner's public API (BaseTLearner.fit, BaseRLearner.fit, BaseXLearner.fit, etc. all use the (X, treatment, y, ...) signature; see causalml/inference/meta/base.py:16 and corresponding lines in each subclass). Every example notebook, every downstream wrapper, every existing test would have to change. That's a major-version bump that the maintainers would need to schedule, not a drive-by fix.

The standard escape hatch for sklearn Pipeline compatibility is a small adapter that swaps the argument order (a few lines of code per learner). If the discussion is "should causalml ship that adapter," reframing as a fresh meta: add SklearnPipelineCompat wrapper issue with the proposed API would be cleaner than reopening the breaking-rename ask.

(Disclosure: I drafted this comment with help from Claude Code while triaging stale issues; the references above were verified manually against the current master branch.)